Yeah i'm presuming you've already gone through and had a look at this um the orange text is just so it's easy to see there is some red text red writing that just highlights a few questions the choice of colors doesn't really mean anything it's just at the time i was writing this it was just easier to see orange on a white background versus Black or dark blue as you can see i've used um so the first thing i did was emphasize the overall process so i i've sort of scribbled it down there that the overall
process of a cast accident analysis is sort of these three phases of analysis so we have the physical analysis of the Physical process we've got the controllers inside the control structure and then we have the actual control structure analysis itself now the main emphasis i think i made was that um this bit here was relatively thin on the ground in the report that you provided which you know given it to an early attempt that's perfectly fine but i just wanted to Emphasize that um now when doing any kind of stamp-based analysis there are three broad
phases there's what i call the foundation phase which is where you set up the analysis with information that you need to be able to conduct the analysis to set the scene if you like then there is the modeling phase which is in the case of an accident we create control structure model but we also create a Timeline sequence of events model which is something that you do as part of any accident analysis incident analysis it's not peculiar to cast but it's certainly something that helps to identify elements that need to be considered in the control
structure model but also how the control structure model and the components within it the controllers behaved over time so that's the what i call a modeling phase and then the final Part is the analysis phase and as i say it's physical analysis controller analysis and then control structure analysis now typically you would do the physical analysis first and that would be typically done as part of any accident investigation anyway it's just cast emphasizes that we do do a physical analysis and that's where we can identify things like component failures physical failures Um then we look
at the controllers in the control structure so we'd identify requirements and constraints and control actions related to those controllers uh safety related responsibilities any issues with control and feedback ability so if the actuators were failed or feedback was not provided or delayed things like that very very similar to stpa but the main difference between cast and Sdpa is obviously cast is done in the aftermath of an accident and there may be evidence that's not available and there may be that there's technical information detail that's not available if you're an investigator versus a designer so we
have to recognize the level of information detail that's available and then we'll look at contributor control actions and interactions so these are the control Actions that contribute to the accident that we're looking at being mindful as well that when we're doing a cast analysis it is possible that we can find causes of other accidents that have not happened yet and i know that sounds a bit crazy but it's sometimes you'll identify a control action that might not have resulted in this particular accident but if the circumstances have been different Then that particular control action could
have contributed to a different accident so there's an opportunity there to prevent well understand what happened in this particular accident but also identify potential opportunities that we might want to look at for other future accidents for example and then we look at the uh reasons why uh so looking at actions control actions behaviors of the controllers Uh mental or process model flaws so if it's a human uh you know their mental model may not have been updated with information about the state of the process the chemical process that they're working on uh for various reasons
not just necessarily failures and then context and contextual factors and i think you did a really good job on that in the in the later pages particularly when you were talking about the startup And short startup after a period of uh uh absence so to speak there's the startup phase then we've got the structure so broadly there are six categories to sort of look at as part of a cast analysis but they are they when i say six categories they're sort of areas where you might want to explore but they're not an exhaustive list there
Might be other things that you want to consider that are particular to your particular accident so you identified areas related to communication and coordination and some level of dynamics and changes evolution of the control structure but there was structure architecture and interactions safety management systems safety information system and safety culture they were they were mentioned very sort Of uh loosely they weren't sort of explicitly explored they were sort of mentioned in passing and then of course reasons why they existed in that way and any any context or contextual factors and then finally summarize for each
of these elements uh a summary of how they contribute to the accident Any related recommendations uh recognizing that when you're doing this analysis you're looking at a specific element and you'll make some recommendations and those initial what i call initial candidate recommendations at the end of the analysis of all of it you'll have a big stack of these candidate recommendations that then need to be reviewed together to see how they might uh overcome each other overlap or even Contradict with each other and then they need to be refined into a final set you know solidified
into a final set of recommendations for for execution and of course highlight any unanswered questions that maybe you're not able to answer as part of this investigation but that somebody else needs to answer later on or recognize cannot be answered and then any follow-on actions um because clearly well in an ideal world once we've made Recommendations from an investigation there needs to be some effort to make sure that those recommendations have actually been implemented and that they're actually effective now that's not necessarily the role or job responsibility of the investigation team but it certainly has
that action that function that accountability has to be placed with somebody or an organization That has the ability to make sure those actions have actually taken place and are effective and then if they're not effective then appropriate action is taken to deal with that but this is where it gets a bit of a gray area because you have the people doing the investigation the subject matter experts supporting that investigation and then the ultimate customer of that investigation Recommendation is potentially the plant management and the plant management is not necessarily responsible for things like uh making
sure that you know control of public health and the siting of residential properties is taken care of so you know there is a bit of horse training going on there if you like but in a nutshell on one page that is the sort of high level process that we would Follow and yeah the main observations i've put in orange um i spent about 10 hours plus a brief call which was one to two hours so that's today's call that to one to two hours okay uh are there any questions on that before we sort of
move forward um no no not but okay and so So one of the things you'll notice is i've used different colored highlighters throughout the document and the reason i've done that there is a bit of a a sense to it is that the yellow highlight is to highlight time information so that when i'm going through the report i'm trying to build a timeline in my mind as to when things have happened uh you know You know when people have been involved when actions have taken place and it helps to sort of build this uh timeline
that i was sort of highlighted can you see my mouse moving around by the way yeah yeah you can okay great so when we're thinking about building the timeline i'll always look for time related information and then i'll just highlight it so You know 7th of may early morning 0242 0304 and then later on near to the end you also have a page with the timeline on it uh where is it i'm sure there's a page with more yeah there we go so this has got times on it oh 2 54 04 15 etc and
that helps to give you a sort of an overview of how things evolved over time And where decisions were made which might be relevant to the controller analysis that you do later on so red i used the color red to emphasize the outcomes so the uh in this case there was fatality and injury uh there was a lot loss of life of 12 people 585 people treated in hospital loss of livestock and vegetation so when doing the foundation phase where we define what the losses were I use a red pen to highlight where there is
relevant information then i use the blue pen blue color to highlight to highlight uh relevant sources of information and then green is candidates for inclusion in the control structure model so as part of the modeling phase i'm always thinking you know as i'm reading through this Information what entities what people organizations equipment or procedures might be relevant to the control structure and i'll highlight them in green so that when i come to draw the control structure model i can mark them off as included um so that's sort of you know reviewing information that's how i
sort of structure it myself Um let's have a look so uh yeah so this is the part of the foundation phase so this is where you've defined the system level losses so loss of human life on site offsite happy with those environmental impacts loss of production damage of assets yeah totally happy with those but i've also included these here just to sort of emphasize that you can Make it more broad so that you can define things as injury or death you may want to include on-site or off-site for convenience i know in nuclear in the
nuclear industry in the uk they make a big emphasis on on-site and off-site impacts and i think that's the same as you've got in the chemical process industry and that's why you've done it so you can emphasize injury death on site offsite Uh regulatory action if the regulator says look you know we're not happy with the way things have gone you're not allowed to operate anymore that's the end of the business if you are subject to legal action either from the regulator or from third parties uh you may have your license to operate revoked if
that's appropriate so they're sort of broader Considerations from a loss perspective so it may be that you've got a perfectly functioning site it's not having any kind of environmental impact you've not killed or injured anybody but you've had your license revoked you know so that sort of helps to i mean obviously this is a safety investigation so we're focused on the safety outcomes but the reason i'm emphasizing that is so that you can get A perspective that actually this approach and especially stpa can actually manage other uh other inverted commas risks beyond just safety and
of course we're here to you know serve some customers we don't want to hurt or kill anybody but likewise we want to be able to stay in business so it makes sense to think about those things up front in terms of system hazards i've sort of broken these up a little Bit more to try and get to the real detail of it so um i've highlighted unplanned and i've said well there may be a circumstance where we have a dangerous success where we've planned to do a release because i think that's covered later on that
there are circumstances where a planned release might take place but unfortunately circumstances Allow for harm or death of people now that sounds like a really dodgy situation but the way that we've defined this hazards it doesn't preclude that planned release resulting in an unwanted outcome so that's why i've sort of emphasized it um you talk about release of a toxic substance that's great and then exposure of the public so what i've done there is i've reformulated those to talk about humans exposed to toxic styrene vapor and then release of toxic Styrene vapor and i've reformulated
those and chris clarified them just so that it's uh more sort of distilled and explicit about the control structure for controlling hazard one human exposure is not necessarily the same control structure that controls hazard to the release of toxic substance now there may be some overlaps in those control structures but If we keep them sort of conceptually discrete that means that we might be able to identify some opportunities for improvements or potential causes of the accident we want to be interested in so there's two high-level hazards once we've got those hazards what are the high-level
safety constraints in principle excuse me um that are intended to control those two hazards and it is literally a transpose so humans must not Be exposed to toxic styrene um i've then broken it up a little bit more the facility must not release toxic styrene vapor and if a vapor release and human exposure must be minimized so ideally we don't want to release any toxic styrene vapor but in the circumstances that we do given that we assume that we might not be able to stop the release we must minimize that release Yeah and then minimize
and control human exposure yes then we've got let's have a look emergency response must be provided to minimize vapor release and human exposure and injury or death now that one's interesting because there there will be a site specific emergency response but there will also be a off-site emergency response which is not Necessarily under the control of the plant manager plant operations it is in combination with local government and emergency emergency services for example so you could split that high-level safety constraint into facility management and local government management emergency response the thing also highlighted here is
each one of these four you've highlighted are high level safety constraints But i've suggested providing a unique identifier reference for later using the use in the report so like i've used hazard one and two here to refer to those two hazards i suggest using a specific identifier for these high level safety constraints so that you can see where they're referred to later on in the document because one of the things that we try to do is we try and make sure that we Promote referential integrity and what's the word consistency throughout the report so when
we're talking about a specific hazard we want to make sure we're using that same definition wherever we're referring to it in the report and at some point in that report we might decide to refine it or split it into a different uh set of subset for example so we just want to make sure we keep that clarity of referential integrity does that make sense so you've Got any questions on that uh yeah the planned release in singapore contest when we doing qra cultural risk analysis the terminology used uncontrolled and planned released so chemical plant sometimes
there is a planned release plant means we need to do the like before if anything need to be released uh We have to do like a safety analysis before have to do this part whether we are arranging separate uh scrubbers or [Music] releasing contain one part then released to the approved waste collectors they will come here to quench it out so because planned release we mean that Like this the way because we are planned to release because there is no possible to store in our site or there is no that is when we are planning
we arrange the appropriate control measure then we will release it that is the meaning of that part that's awesome so so in that case what i'd suggest then is is having h2a and h2b or h2 and h3 one is unplanned release and the other is planned release and then you can see the control structure For an unplanned release and the control structure of planned release will have more things in it so when you're when you're doing that planned activity you could take that planned control structure and use it to drive the plan and drive the
the verification validation of that plan with all the various stakeholders and so that would be a starting point for an sdpa for your planned release Really really i appreciate because whenever we're doing like planned release for example uh we need to like punch it out because we're going to shut down the plant for next to 15 days yeah maintenance we need to punch out when we need to purge out we need to fly all the items when we do the flowering we need to inform to the neighboring companies and authorities we Are flying yes yes
because they are flying because this flame is very big yeah and they worried and made chances to the dark smoke is coming the radiation nighttime they are very bright they called the authorities why what they doing the plant so we have to inform the neighbors and oh yeah so now uh control structure for planned release also is good I agree that the point yeah okay um great that's cool okay so uh the safety control structure was developed based on the available information in the high power accident report the reason i've emphasized that uh is because
i just want to say it's really great that you've said that straight away because a lot of people don't necessarily recognize that the control Structure model they're building is based on information provided to them they're not necessarily creating a new one they are creating one based on what exists and so as part of creating something based on what exists you are potentially introducing your own biases in what's included in that model and especially what's excluded and that decision making is not necessarily Explicit but because you've said that specifically it's good it's an indication that you've
actually thought about that and i think that's really important to draw attention to so i think that's really good um so this one then so this is sort of the high level safety control structure which is really great this is a good starting point there's a couple of things on here that i've added just so that as i've been Going through the page through the report i've made a little note to come back and talk about it so one of the first things is uh down at the physical process level so conventionally the control structure
models that you'll see in literature and on the internet and in nancy levison and mit student papers they they typically use the convention of uh vertical sort of hierarchies so at the bottom you have the physical process or the hazards that We're trying to control and then as you move up you get from direct indirect management regulators law and all that kind of stuff yeah so so you've you've captured that sort of uh convention but what i wanted to emphasize is the question of well what what is the actual thing the process the physical situation
that this control structure Actually controls and so i've highlighted you've said there it's a styrene monomer storage tank so yes that tank is part of the control structure and it needs to have some level of control over it to make sure that it doesn't burst or anything it needs to be maintained properly but the the physical process that we're talking about is actually within that Tank so it's the chemical process that we're interested in and the type of sensors that that physical that chemical process has inside that tank is directly related to the type of
process it is and how it can change how it can go outside of control you know whether it needs temperature pressure the type of mixture of the chemical process and the depth of it and all those physical aspects so have a think about well what Is it what is that physical process and what actuators and sensors uh what control actions what feedback do we get about that because i think you like you talk later on um about the the the depth of the fluid and the way different parts of that fluid that that mixture can
react differently dependent on its depth and the temperature profile and things like that and i think that's A really important thing to sort of raise to the surface of this investigation but it's also you know if we have this you know substance in a tank how can we actually control it you know do we need sensors at different depths or do we need to change the depth you know do we need to restrict the physical style of storage so that it is always under control or do we have to accept when the tank is at
a certain depth we don't actually know the Status of it so is that okay or is it not okay do we need to change the design or what can we actually do about it so that's that's the first thing about the chemical process itself then there is the vapor cloud so conceptually we've got this unconstrained uh fluid in the environment how are we well we may not be controlling it directly but we certainly need feedback About it so we've got gas monitors and things like that you might say an actuator could be an on-site alarm
that goes off to warn people that there is actually this vapor in the atmosphere and then of course we've got public health as well so i've i suppose you've put ven qatar purim public and yes they are under the control of local emergency management but it's more what we're trying to Control is public health more principled principled should i say so how do we control their public health well one is we control their location and presence in the proximity of the plant but the other is we provide timely emergency response specific to those kind of
injuries and those those sort of distinct elements Will show different parts of the control structure one is about logistics and coordination one is about medical emergency response now i'm not suggesting that you want to go into that level of detail on this first high level control structure but what i'm trying to sort of emphasize is when you think about that physic what is the physical process we're trying to control You start to see different dynamics different different perspectives of what the physical process is and what the control structure model might look like for each one
so later in the analysis you could add more detail to this control structure model and then examine specifically about logistics of people and then medical aspects of people so that's that's the other thing um on one question sure Okay my understanding each feedback need the control uh uh response is it or no need means not necessarily not necessarily yeah so in in control engineering the ideal control uh control loop is a closed loop control yeah so you've got control actions and feedback and they imply the presence of Actuators and sensors whether they're human or mechanical
whatever they are now there may be circumstances where you don't have closed loop control you may only have actuators and have no sensory feedback you may have situations where you only have sensory feedback and no ability to provide control actions and that can be dynamic especially when we're talking about human organization to human organization interactions you know it's pretty much Voice hand signals eyes and ears you know and and that's dynamic it's it's not necessarily uh and well it's not engineered so it's not necessarily as robust or reliable as you know an engineered system is
communications perhaps but just by thinking in the in the perspective of i ideally would like to see a closed-loop control okay but looking at the real situation i have to recognize i might Have one or the other or not it might be uncontrolled once you know you might not have any sensory feedback at all but as part of your investigation ideally you want closed loop feedback and you'll be looking for it to you know the evidence that you have in front of you might suggest you need closed loop feedback but the real situation there was
no feedback there was no control actions And that's why i've sort of emphasized this link here between the lg polymers management and the general public in the side of the plant you know if they put out an emergency tannoy message or a warning sound siren or something air raid siren to make people in the local area aware that there is this vapor cloud flying around um that might be a control action on behalf of the management to alert people To this presence of this vapor cloud but that control action might be completely ineffective if the
public actually don't know what that means in fact it might have the opposite effect and cause people to congregate around the plant to find out what's going on you know so that allows us to think about what is the control action between here and here if any If there was some was it adequate or not and also was it appropriate for the people were their mental models and their control algorithms actually appropriate for a safe response because if their mental model was oh we hear this siren going off but we don't know what it is
their control actions might be to run away or they might be well let's go and have a look let's get some pictures for Social media you know maybe we could uh get some money for selling videos and pictures of a disaster you know but as part of your as part of your consideration of the high-level safety constraints uh where is it humans must not be exposed to toxic styrene vapor when you're thinking about the plant management one of their control actions will be Alert people on site and train people to know what to do to
evacuate or take emergency cover or whatever it is but then also in the context of public health the people organizations that are in control of public health also need to be aware that there's been a release and know what actions to take and have drilled it xyz but there may be something in in terms of time because Obviously each of these control loops has a time delay in it and so when this is a real time situation where there isn't much time to act there is an implied requirement on lg polymers to provide this control
action to you know the local public and i've not highlighted here that there is actually this link here to workers on site wow how do we tell workers on site what to do You know so this is management this is the chemical process there should be a box down here that says employees how does polymer management deal with employees on site either those employees involved in this process and also employees that are on site but doing something else does that make sense yes so initially what i thought the want to draw the feedback from steering
monomer to the public The feedback go to the public the public feedback to the local emergency the local emergency go to the management and come back so there's a big loop of control action oh yeah yeah absolutely and all of that has time delays also i thought i thought this one is very difficult not the difficult i don't know whether can do this way or not yes in singapore contest uh In the polymer to the public we have the the local legalization is to be they make it like a connection between the local and to
the management yeah okay so anything happen the management uh need to inform unknowns or anything yeah uh if singapore is industrial area is industrial area there is none of the Public people stay inside yeah so government legally say you have to share the information with the neighboring companies yes yes ah okay so i've not included that from here so you you might you won't go to this survey yeah well you could put adjacent uh operators as adjacent companies as well as well as public health you could expand that to include adjacent operator because the thing
at the end actually That's a really good point let's let's just explore that for a minute so now we know effect if anything escalated to the other plant the other plant will cause capture its dominant effect oh god yeah well well say for example you've got a next door neighbor that's doing a bit of a high risk process yeah and they they have to evacuate their plant within five minutes Because of something in your plant is their process capable of being put in a safe state in five minutes yes or do they have to evacuate
and risk another worse situation uh yeah but i'm hoping you can see that just drawing a diagram boxes and arrows is starting to make i don't know about this uh this lg polymer How the surrounding areas or something but if the plant in singapore i have to do this way yeah the controversy will become different yes absolutely this is great yeah the public moved behind to the local emergency it's a local emergency come in the local emergency move the neighboring companies the first person is neighboring companies yeah excellent then then go to the local emergency
then go to the public Yeah well i'll just emphasize so this this interaction here i've i've put it in there because logically lg polymers management they become aware of a problem quicker than anybody else and so it's it's impressive impressed upon them to get the message out to everyone else that there's something not right and that those people that receive that message know what to do and can actually Do something um but so that interaction might be directly you know an alert uh tannoy or alarm or whatever alarm um but also it may be that
that is fulfilled by communicating with local emergency management who then coordinate the public well the question i have is is this lit is this path here actually quick enough to prevent injuries and deaths and gives them enough time to actually do Something or is it necessary that we have a quicker control action uh that's direct as opposed to a third part that you know through another controller if you like well that's something to think about because that's you know i can make these recommendations and suggestions but whether they're actually practicable in the real situation there
may be a very good reason why it's not Like that and i know it might sound a bit dark but there may be circumstances where certain chemicals are released and it doesn't matter if you give them an alert it's too late you know they're going to the diagonal line from lg polymer to public yeah must be enforced by the government yeah yes Yes even because public there is a lot of political politicians and a lot of uh parties involved so they will get the money from that they close them out that's a good point now
yeah but these are my main things it is be uh enforced by the government yes you have to do that is the part that is our one party here also we need to be considered yes um if we were talking about doing a Uh say an audit for example a safety audit as part of looking at this control structure model as an auditor you might want to select that control action and actually do a an independent review of it to make sure it's in place and that it's working and that it meets the requirements of
the government but also the safety assessment for this particular process because this this link Whilst we're looking at this specific chemical process this link represents all of the chemical processes on site potentially and you might have some processes that require faster excuse me faster response or even a specific alert i've worked in particular places that i can't mention where they have specific alerts that mean certain things on site and they mean that certain people on Site nate need to take particular action depending upon what type of alert it is uh you know the severity of
the release of whatever that material is um and so some of those alerts are in place because of regulatory requirements but there may be other alerts in place because of the safety assessment and making sure that they don't overlap or cause problems for each other so that's so that's that physical Process then us that yeah there we go so then i popped out to emergency responders and their coordination and control and their health so two different elements and then outside that there is employee health and then also employees themselves and the actions they take and
like you said there are then next door neighbors and their employees and Their processes and their health that should be thought about but that's across an interface and then on this side i just emphasized two sets of uh so process controls so these are the physical equipments that are involved in this chemical process so we've got chillers and cooling fire sprinklers tank contents control Emergency chemical inhibitors then we've got the actual employees so we've got the director operation the safety head the operations leadership the operators maintainers shift management question mark and then the designers modifiers
of the system itself so i'm not suggesting you take this model and explode it into all of this detail but what i'm sort of emphasizing here as you go through the analysis you can expand certain parts of That control structure look at the detail within it and then move on to the next bit and expand that and have a look at it then one i got one doubt yeah no these branches you put there it will involved in in the stop analysis or cost analysis my understanding cut cost analysis you need to identify the root
cause involved and the impact on this if like this we are putting the cost too Much branches it will explore more it became a stopper is it correct uh not well no not exactly so the main difference between sdpa and cast is the detailed analysis of the control actions and feedback and then the controllers themselves so with cast we're not going to go into that level of detail But there's nothing stopping you doing that if you decided to do that um and what we're talking about here is where sort of where we're emphasizing how flexible
the modeling of the control structure can be and then allowing us to then drill into certain things in a lot more detail than we might otherwise be able to so this is Sort of uh setting the scene for what we're actually gonna drill into and look at how we actually drill into it and look at it in detail is is the analysis in cast or the analysis of sdpa this is more of the modeling stage this sort of sets the scene as to what we're actually going to look at where we're going to search how
we search is the difference between cast and stpa and But the thing to sort of be mindful of is and this is one of the skills that has to be developed from experience really is how detailed do you want to make the control structure model because when you're doing an stpa hazard analysis typically you have access to as much information as you need and so you can make it as detailed as as you need to but when you're doing an accident investigation especially Um in the aftermath of an accident you only have certain bits of
information available you have to go looking for it and it takes time to identify information that may or may not be relevant so it may be that you can only do a very high level control structure to begin with and then as you move through the investigation more information becomes available you can start drilling into certain areas and adding detail as information comes Available or it may be so that's the first thing the second thing is it may be that as you're building the control structure model as part of the investigation you might identify areas
where you have questions and you elect to drill into it and then go and investigate and collect the information that's relevant to it so it can help you drive the direction of your investigation and one example i've used Here is this shift management question mark so the function of shift management who decides who's on the shift what competencies they've got whether they've had adequate rest and food and all that other stuff that function is performed by actual people somewhere and i would argue that with this particular Accident the way it was shut down and started
up given the context how that function of shift management was performed was not adequate in considering how the thing should be started up so that's that's one thing i'd want to explore as part of my investigation whether i include shift management explicitly in the control structure model is you know personal choice really i Mean it may be that when you interview people and you say you know what part do you play in the control structure what roles do you perform what actions did you take what were your beliefs you know what was the timeline sequence
of events they may say well you know the shift management uh you know we were supposed to have ten people on shift and only eight people turned up but we still went ahead with the operation as a silly example The person who made those decisions might not be the person that you're interviewing so you've got sort of a you've got an area that you want to explore but you have to expand your control structure perspective to try and understand the individuals themselves and how they actually work together to perform this particular you know function of
the control structure And that's not necessarily written down that's you know the what they call it work as imagined versus work has done you know people typically try and do the best that they can given what's going on and they may not they may themselves not be equipped to realize that they're setting themselves up for a fall um but again that's you know all problems are leadership problems Uh if they're not equipped to know when they're setting themselves up for a problem that's not necessarily their fault or responsibility it's the layer of leadership above them
you know that's ultimately what their function is so i suppose to try and draw all that together i'm i'm talking about developing a control structure model From the perspective of you know there's been an accident what is actually available in terms of information what have we got available recognizing that some parts might be lost or destroyed and we can't ask questions of it um what uh based on the information we've collected so far what can we see our Deficiencies with the control structure and the interactions between the different layers in that structure but also where
are some of the key questions that we have where do they sit in that control structure and can we get the investigators to go and collect specific information from that part of the control structure um as it existed at the time and that might be through interviews it Might be through physical attendance at the site and collecting physical evidence it might be getting documentary evidence uh you know and that that's why there's always a big emphasis i mean in my experience in aircraft accident investigation um and and facilities and factories it is quarantine information prevent
things being destroyed or lost either You know i'm not suggesting people are corrupt but there are circumstances where people might feel it necessary to destroy evidence not necessarily because they've done anything wrong but because they think you know they might have done something else that's not related to the accident but they don't want to get found out and then criticized for it and so they take action that then undermines the investigation so being able to Quarantine stuff and control evidence is fundamental but what i'm sort of saying is this approach when you're refining and building
the control structure you can actually start to see areas where you might want to ask specific questions and when you've already interviewed somebody you might then be able to go and go back to them and ask them more specific questions based on the refinement of your control Structure model so it is a bit of a it's a creative process it's not linear it's uh sequential potentially and depending how much time you've got how much detail you want to go into you can go really really really deep but again you know how it's it's knowing where
to spend the time knowing where to invest that time Because once you've spent it you can't use it elsewhere um so i've talked about that we've talked about this okay so let's go a bit further up the control structure so there was a mention of the factory inspectorate later on in the report but i wasn't sure how that fitted in to indian government authorities or the local government authorities and then also land use control And legal regulatory standards so there was a mention of the factories act 1948 which covers the manufacture storage import of hazardous
chemicals rules 1989. i i've excuse me i've added those at the top just to sort of emphasize that we can start to look at legislative regulatory and standards compliance but obviously the where the Where they actually apply question mark you know some of those will apply directly to lg polymers management some of them will apply to the equipment designer and manufacturer some of these will apply to government authorities and they'll be decomposed and flown down through legislation and standards directly into the company so this is why i've sort of just put it at the top
there we know that they'll have some kind of effect we Don't know whether they're relevant to this particular accident or not but if you wanted to look at that you could stick that at the top and explore it as a particular topic as part of the investigation now i'd like to say that there is some kind of regulatory body somewhere in here that's already taking care of this that's already making sure that this organization based on the report i read out The legal regulations the two part will take care by uh entrepreneurs the local this
one here local government the state government yeah so factory inspection also and also the standards and the import license the indian government authorities yeah yeah so as part of the investigation as you go up the control structure you could go and explore and interview People within those entities and say you know in the context of legal regulatory standards what was relevant how were they flown down what feedback did you receive that said yes we've complied with it and i think we mentioned earlier on in the uk we have safety case related regulations and so the
lg polymers management would create a safety case Report of some description and that will have a section within it that explicitly says these are the legal regulatory and standards that we're required to comply with this is what we've done about it and then the existence of that safety case report would go back to government authorities to show that they've complied with it now what the government authorities then do with that safety case report i'm not privy to it maybe it's You know hands-off regulation we've told them to do a safety case they've done one tick
in the box you know and also the land use controller yes that is a local government requirement but there are many issues on this my understanding because the bhopal also the same issue on the land issue control There is a chemical plant government allow them people stay near to the high high-risk activities areas absolutely yeah it's the same thing here in 1947 they started the plan 89 they initially the sugar plant molasses stored 89 they change but the government surrounding area government issue the land control to the public to build up their warehouse so If
you do qra the radius is is uh go to the area if you do modeling it reach the public yeah so yeah we cannot control is is even though uh like in singapore uh there is a land constraint yeah if i do qra for my television tank is 1.6 kilometer you say 1.6 kilometer nobody stay means the government will not agree your Project so the control measure need to be standardized to yes how like bow tie in the case they come out the bow tie how they're controlling so yeah that's the legal requirement leave pushing
the people yeah pushing the management side and pushing the management that's right yeah so so that was sort of a Uh you know an initial thought um one of the things that i'd recommend is when you're building your control structure models start with one that's as high level as you can to begin with and then take a copy of it and then modify the copy and start to add detail to it as you move along and then when you come to document the thing in a final report I would make it nice and clean clear-cut
so that you have a working version and then a one for presentations because obviously you can if you really go to town you could explode this and have all manner of stuff in it and that's not necessarily going to help the audience understand what it is you're talking about but also as part of your investigation you might explode should say that you might expand each One of these blocks on a particular control structure model for that bit of the analysis so when we're talking about the physical chemical process we might expand that bit and add
extra detail that's not really relevant to the rest of the model so it is a it is a bit of a challenge trying to manage all of these models together but i i tend to have you know a very big one that's all detailed and then i'll thin it down and expand it again Depending on what i'm trying to do i think i showed you that before and so the next bit i just want to re re-emphasize the hazards and the high-level safety constraints just to sort of emphasize the principles of stamps so we talk
about hazard one and hazard two we talked earlier about planned and unplanned release i've not emphasized that here but you know that's something for you to think about The main thing is we want to control these hazards uh in inverted commas using a control structure so what does that control conceptually look like well we talked about closed loop control so ideally we want control actions and feedback that might not be practically appropriate for certain things especially things that are outside our immediate control such as public health if we're talking about the context of a Plant
management so we want to control of those parts of the hazards what can we control well we can control humans to some extent we can control exposure to some extent can we control whether the vapor is actually toxic or not uh can we control whether the vapor whether it's vapor or not you know maybe there are chemical things that we can do that can turn it from a vapor into a liquid or toxic to non-toxic you know I'm thinking very broad about the concept of control and that sort of leads into the hierarchy of risk
controls you know eliminate reduce and all the other stuff um and then given that we might not be able to control it or we are inadequate controlling it what is the response that we need and that's why there is high level safety constraints that relate to control and then a high level safety constraint that Relates to response so i put that all on one page because it was a bit scribbly on the previous page it was a bit uh where was it it was all sort of embedded yeah so i just put it there to
emphasize that so moving on to this bit so this is where you add more detail to the control structure model and i've i've added the page references to each of these okay and the reason i've done that is Similar to when i mentioned about giving the high level safety constraints a unique identifier the reason i do it on the control structure model is it makes it easier for people reviewing the document to know where that particular bit the control structure is analyzed in the document so i've emphasized distributed control System so dcs i think is
something mentioned and then i've also put polymer level gas detector and then gas detector alarm so and then cctv that was a closed circuit television that was on the outside so this bit of the structure you've got here talks only about the process control system as a concept of control but i've highlighted here that there's Other elements of feedback that are mentioned in the report that are not related to the process control system as a technical system but they are things that you should be included in the control structure model okay um because for example
if the gas detector was failed is the process control system in it on its own adequate to still safely control This process you know is this a safety instrumented system part is that the safety mechanism or is it also built into the process control system so this allows you to sort of conceptually separate out process control only safety instrumentated system control and then other administrative controls such as cctv recognizing that cctv is not going to be monitored Every second by somebody so if that wasn't available would it even make a difference you know it sort
of helps to sort of emphasize that um [Music] so uh the event involved in the uncontrolled release of the styrene vapor so what i've suggested is add detail regarding post release uh including emergency response public notification now you don't necessarily need to do that here because you can Structure your analysis based on we're looking at what's happening inside this boundary here but when we come to the report part of the report where we talk about this that's where we could include the narrative the timing on the emergency public notification aspect so it's how you break
up the report and present the information yeah okay so you'll see the colors for time Yellow for time green for objects time yeah so delta t temperature delta p pressure seven hours twelve hours uh so tank design maintenance age material open to atmosphere modification delta t control design delta t control design contamination delta t control chemical pro control process so these are a list of the proximal Events contributing to the uncontrolled release but they they sound like um a list of relevant information about the process not necessarily events so inadequate design of the storage tank
roof designed many nozzles at the top in terms of events in time are we talking about the fact that the design was created at a particular point in time or Are we talking about the installation and commissioning of it or are we talking about independent review of it was inadequate what what is it about this that's related to a proximal event it's it's relevant information but it's not necessarily relevant to the sequence of events the timeline and i'd argue that this information would come as part of the analysis of the physical process a bit later
on okay So we've got yeah this bit here lockdown period pandemic issues so that is a context related to the control structure and the behavior of the components later on um role and responsibility the controllers role of the styrene physical design in the tank okay so the styrene monomer storage tank is usually designed to protect against runaway polymerization so the reason I've highlighted usually is because that term that implies the use of that word implies that when you've written this you're sort of using your experience based on what you've done and saying well usually we
design the tanks in this way but what about the particular tank we're talking about was was the tank designer of this actual tank Did they usually designed to protect against runaway polymerization no um okay so it's just uh it's just the the reason i'm pointing at it is just to you as an as an analyst you bring your experience to the investigation but we have to be really mindful about uh are we making implicit assumptions about the design or the people who did the Design and the assumptions that they made because it may be and
i don't know but say for argument's sake it was at the time this thing was designed and installed there was no requirement to design it to protect against runaway polymerization and it may be that there was no requirement to do it but it may be that the safety assessment they did at the time that might have generated that requirement did not generate that requirement Because it was a bad safety assessment or because the safety assessment did not consider runaway polymerization for that tank because it was never intended to ever have anything in it that could
have a runaway polymerization and so the original designer um they would not have thought about that and so the design was the design when the chain when the usage of it was then changed it was that change Assessment that should have considered whether it was adequately designed to protect against runaway polymers exactly exactly happen like this they have the another m5 tank they they designed to foster in monomer that tank is designed to faster in one about but this tank is usually initial stage is designed for molasses storage after that they changing to the toxic
storage storage storage did cannot Do the proper emoji the designer not involved they won't do the management duty himself oh ow okay that's an interest in finding an investigation uh they the material is a carbon steel they purchased when korean team takeover they purchased one more tank like m5 stainless steel tank but this tank is a mild steel tank so the mild steel can Steering can allowed to store but you have to monitor regularly yeah but that part they are they missed it out wow so that is a a huge observation there that's also a
huge observation when uh this one uh [Music] when uh because this tank is very big huge because 3000 kl store the steren uh sterling monomers Okay wow also uh sterine monomer is is is a very toxic material that the tank head was also problem liquid phase also problem yeah so if the head phase not kept well not polymerized uh the Inhibitor loss because the the sterile monomer boiling when vaporized it's a volatile compound is going up in the uh to the roof so when the vapor press there is no inhibitor it start to polymerize that
polymerase drop into the liquid the polymer the the the material involves a catalyst to Form the random reaction so is it very critical that's why that usually i i don't know whether i intended put or not but usually it's not designed uh for uh this tank yeah no it's uh that's really good that's really interesting um as i said i was just highlighting it to be mindful of when when you're doing the investigation and Documenting it try to be mindful about your position as an expert in what you know and as an investigator we're looking
at this with uh a uh with a clean sort of mindset you know we we have to be careful about what assertions or assumptions we're making and be clear that it may be that the people at the Time that did this design and specification they the reason they behaved the way they did was because it made sense to them at the time so why did it make sense and what what is the context around that decision making um but that's all i really wanted to emphasize it's not sort of a criticism it's just it's a
mind doing a lot of this stuff is It's getting your mindset getting your perspective right and being very sort of disciplined and it's not easy it is very hard so requirement to provide the physical protection against hazards which is great so i've just highlighted h1 and h2 plus the high level safety constraints and i've also had the added their their refined specific requirements so those high-level safety constraints at the beginning Um when we're talking when we're talking about the specific physical controls uh the tank the equipment that controls the process the actuators the sensors everything
else to do with that they are specific refined requirements based on those high-level safety constraints now they'll depending upon the level of detail Technical information you have available about the plant and you might not have it in the aftermath of an accident and and even in new designs you might have limited access but it's trying to understand how those high level safety constraints actually link with the technical and detail requirements in the safety assessment in the design because that like for example The choice of mild steel versus stainless steel the requirement to have one or
the other uh whilst it's a material requirement materials substance requirements there might be a link to the safety assessment that says you shall have stainless steel or mild you shall have stainless steel because of the potential for an explosion or or whatever and so whilst that when you look at it in isolation it's just a materials Requirement physical equipment physical equipment requirement it does link back to the high-level safety constraint the reason it's this material is because of this high-level safety constraint about preventing preventing release of the equipment so have a you know have a
think about what those specific requirements might be and how they link back to those high-level safety constraints Um so yeah be mindful that accident investigators may not have that detailed technical information and that the design process may have overlooked those specific requirements or the designers made justifiable compromises decisions and that could be due to flawed standards you know so thinking about the control structure from a design control structure Perspective we have a physical bit of equipment the reason that physical equipment is as it is is because of a design process and the people involved in
that design process may have used a standard from 30 years ago where this kind of accident didn't happen and so those standards didn't cater for it but fast forward 30 years later maybe the standard has been updated but it hasn't flowed down onto the physical equipment so there's a there's a gap Between you know the standards of 30 years ago and the standards of today and how they may manifest themselves in reality and it may be that well the 737 max accidents are a great example of this just because the aircraft was certified 40 50
years ago doesn't mean it's actually safe today you know that's a huge assumption that has to be validated and and the control structure Involved in that validation every time a new standard comes out a new regulation is is uh released the management organization has a an implied duty to look at it and check well this new thing that's happened does that does that still hold true for our facility you know we used mild steel back 30 years ago the standard now says use stainless steel why is that change and does that mean we Need to
do something or does it mean that it's okay not to do something so long as we don't go you know doing what we did in this particular situation it is closing that loop across time and generations and it's not easy it's not easy at all uh compatible materials blah blah blah red red red here we go protect against inadvertent release yeah so that was Just emphasizing the unplanned versus planned uh situation um right here we go sirens to alert the public and workers as on-premises so that is a actuator uh that that was uh here
this this this line here is those sirens Emergency response plan and team uh missing or inadequate plant physical controls that might have prevented the accident so this i've put that gray lining just to make sure there's a clear between the different layers of the analysis this is the next sort of layer down uh missing or inadequate plant controls so what i've suggested is later on adding in the analysis for human factors Human biases reasons related to why reasons explanations that these things are missing or inadequate so the the material of the tank being mild steel
instead of stainless steel let's have a look at that you know why was that what the reasons what's the explanation for it were there any missing or issues or flaws inadequacies because and the main point is because thinking About what was inadequate at this stage may poison distract from what was adequate which are dangerous successors so using the word inadequate just be mindful that that word can set us down a particular path of exploration that may may miss other important things that's why i always emphasize when we're talking about failures and success That some failures
are dangerous but also some successes are dangerous and we need to be mindful of you know whether it was adequate but actually it was still dangerous you know because if you comply with a regulatory standard you know compliance is great but is that actually safe and it may be that like i said earlier they were compliant with the standards of the time to use mild steel but it just was not adequate for this Um and then that might have prevented the accident so and this is something that i have a bit of an issue with
with what nancy leveson's written in her guidebook on this but when we list things that were missing or inadequate plant controls that might have prevented the accident we have to be mindful that when we say might have prevented the accident we've been invited to Uh you know use a little bit of hindsight bias so just take care be mindful of hindsight bias and especially building a cast wish list because you know a cast wishlist is something that a term i've come up with where we as investigators where it's post accident oh well you know if
they'd have had stainless steel tanks and more sensors and more feedback and all this other stuff then this Accident would never have happened and it may be that the people that designed it they considered that wishlist but there was some real genuine reasons why it didn't have all of these wonderful features that we think in in aftermath might have been appropriate so we have to be mindful of that uh so think about uh you know building a a cast wish list of controls without acknowledging design constraints decision making which may have been justifiable at the
time it was Made and that's not an easy thing to do that is that is tough um and then this yeah so i've highlighted inadequate inadequate inadequate inadequate so just be mindful of using potentially argument arguably judgmental terms language designers may well have considered it adequate so just be mindful of judgmental language versus judgment of actions in Hindsight and that's not an easy thing to do i'm just sort of emphasizing that physical failures so there was no physical damage except toxic styrene to the atmosphere um but i have a question there about any equipment failures
eg the alarm's not sounding so i think there's a mention that the alarms were not sounding or were delayed or something like that Yeah and the tanks themselves they were corroded and they had issues with their maintenance so is that a physical failure that we need to include in this section because it because again there's physical failures that were present and then physical failures that contributed to the accident so i'm not suggesting we list all Physical failures and issues with the plants that are not relevant but just be mindful about whether there were any that
contributed like if the alarms were not sounding because of a failure you know the electrical wiring was broken or something here the meaning is failure means that the tank is not busted or uh safety wall is not popped or something like this there is nothing okay yeah Okay well it's this uh physical let's go back to the first page so when we talk about physical analysis we've got what components existed what happened with those components so we list the equipment physical controls so that's the tank the actuators sensors the fire alarm the sirens the Process
equipment you know if there's any feedback or other actuators the roles and responsibilities requirements constraints for each of those so like the the siren for example the siren needs to be audible at 1.6 kilometers or two kilometers or whatever it is and it may be that we're not talking About one alarm we're talking about an array of sirens alarms at different areas of the site boundary and does each one of those reach far enough into the surrounding area that they're adequate for this emergency notification so that's just an example of how far you could go
with that piece of equipment and i think you've probably you've got to the point where you've realized actually there's quite a lot of Stuff that could go in this section um that needs to be thought about physical failures with that equipment contributory control actions and then reasons why missing flaws inadequacies context contextual factors so for each one of these bits of equipment you'll go through this list and fill it out essentially and then identify any recommendations or questions Um so let me just go back to there so what i'm saying is this section here where
we talk about from this point from this point forward the role of the styrene tank you've just you've focused on that tank whereas there is all of the other stuff on the outside that needs to be looked at from a physical process perspective as opposed to a controller perspective Okay great uh so we talked about talked about that talked about that talked about those uh we had exceeded half an hour is it okay for you oh gosh yeah i didn't realize the time yeah i'm happy to carry on as long as you know for another
hour or so are you happy to continue because i know it's quite late where you are no no problem guess i'm sitting in the home i can't okay tomorrow sunday here Let's have a look then so uh where you are downloading bottom only one rtd resistance temperature detector so is that a single sensor that might indicate a faulty process control design yes if that's true then there's an action there to explore the process safety case safety assessment yes in the in-sterein monomer every two meter height you have to The design supposed to install one rt
okay you know to avoid the um the layering if you don't have the proper agitation or proper circulation uh each layer varying the temperature if the top layer high temperature polymerization start the bottom is called so which is really happened in this in this m6 tank okay So there's so this uh this point here would uh suggest a recommendation to either explore that safety assessment and the reasons why they decided not to have that any more than one um or it may be that you do that investigation and then add it in this section because
i think the thing to remind as well is this is a as you're doing the investigation you're Capturing information and making it questions and recommendations and actions and then you're going off and doing that work and then coming back and documenting it and so this thing that we see in front of us here this document is a is a working a working report it's not a final report and how you capture those actions and manage the you know that manage the conduct of the investigation and the analysis that you perform it is a Challenge and
it requires a lot of sort of tracking and action taking an action resolution um let's have a look so the next page make explicit yeah there we go make explicit document questions and reference where they're answered so like i've put there there's a question here about well if there's only one rtd that implies that the faulty process control design so Let's have an action to go and explore it so piping modification may indicate faulty process change management so we've got an action there to go and do something um [Music] yeah so this bit here this
is interesting uh the refrigeration system is operated from eight till five in all days by manually Manually control so the first thing is the operation of that refrigerant system is a key safety constraint for controlling the the process so there's a question why was that safety constraint not enforced um and that implies well is it okay that it was done manually is that actually a robust way to control this process and then the next one is pump capacity so the pump capacity of styrene Circulation pump was insufficient so again that might indicate a faulty process
control design or it might be that there's a fault with the actual pump itself you know maybe it had the capacity but it wasn't performing to the right standard because it wasn't maintained properly or something like that so this is design and maintenance control structure exploration So look maintenance issues may indicate faulty process control maintenance design or execution and i always try and you'll probably see from some of my videos i talk about the system life cycle so you've got the operational phase which includes normal operations abnormal and emergency conditions you include maintenance in that
you may include change management but change management is a design phase type Activity so design problem definition solution definition requirements architecture design implementation verification validation test commissioning so in all of that where about are those issues sourced is it that they've existed in design or implementation or maintenance or is it a combination and you can create control structure models specifically for those but just Be mindful that when you do an investigation you may not have access to that information or it might not you know it could be that let's say there was two changes on
the plan one ten years ago one five years ago when they did it ten years ago the process that the people followed may have been different to the process that they followed five years ago and it doesn't necessarily mean that what they did at the time was wrong Because that's the best that they could do but it's just that today it was inadequate and and how do we sort of assess that and make recommendations it's it's it's that mindfulness again be mindful that we don't judge what went on it's about being exploration you know let's
explore what happened and why it made sense uh no monitoring device uh or no monitoring system Absence of dissolved oxygen monitor may indicate faulty process control design again yeah action uh the plant was not run down due to the lockdown period of covet 19 spreading vulnerabilities related to design this could all right okay polymerization runaway polymerization styrene boiling are occurred due to vulnerabilities related to the design okay so this word vulnerabilities at our point in time we're looking at This design and we see these issues and they look like vulnerabilities in the design to us
today based on what we know with hindsight whereas that designed to somebody last week they wouldn't have seen those vulnerabilities i i'd like you know i'm sort of playing devil's advocate i'm just trying to sort of emphasize that two weeks ago if there were vulnerabilities you'd like to think If they knew they'd do something whereas you know that if they didn't know there were vulnerabilities and we only know their vulnerabilities with hindsight then is it appropriate to refer to them as vulnerabilities in the design it's a bit judgmental if if you get where i'm coming
from it's it's a common theme it's you know okay tank was fabricated 50 years ago for Molasses storing purposes so that was a change of use i think you've already said that the engine is yes okay so i'll i'll just jump back to uh page one to the process so when we talk about change of use dynamics changes and evolution that's the topic there that we'd look at so the control structure intended to enforce those high-level safety constraints was changed Back in whenever it was so the question is well why was it changed and why
was it accepted has changed and it may not be it may be they didn't explicitly accept it because they didn't know it was a problem because the process they followed was not very good you know but at least we can ask that question where did we get to so many pages right [Music] no i think yeah we talked about that talked about this yeah there we go this is where we are change of use molasses uh so the plant was in a remote location but the city grew closer what control over land use we've spoken
about that so again this is a category of On the first page dynamics changes and evolution that we can look at [Music] we talked about that oh there we go change of use remote locations we've spoken about that so summary of the role of the physical components in the accident no physical control failed the tank pressure reached that Many of the unsafe interactions were results of tank design flaws um as i meant about i think i've banged on about this a lot already it's not necessarily a flaw as it depends on the perspective of the
speaker um and this is not uh you know nancy leveson in the the guidebook on cast talks about flaws and inadequacies But i don't think it sort of emphasizes that we have to be mindful and respectful that we're looking at something with a fresh pair of eyes and it might be obvious that it's a flaw but that's not how it was for the original designer and that's why when we talk about sdpa hazard analysis versus cast accident analysis when we're talking about control actions that led to an accident in stpa we talk about Uh unsafe
control actions whereas when we're talking about caste we talk about contributory control actions they're the same control actions it's just post accident we can't we can't refer to them as unsafe because the people involved in the accident if they'd have known it was unsafe they wouldn't have done it you'd like to think you know so i won't label that point anymore suggest unique labels references for Recommendations we've already talked about our high level safety constraints i also suggest a robust label and the treatment of them as candidate recommendations for later refinement because as i said
at the beginning when you analyze one bit of the control structure you might have a recommendation that seems perfectly rock solid but then later on you might identify something that potentially Conflicts and it's been free to be able to refine those appropriately uh develop a life extension program uh to monitor all hazardous substance tanks so i've said that's a good question how can the issue be systematic across other parts of the facility businesses and industries so the thing about that is whenever you have an accident or near miss it indicates that the control structure Is
inadequate in some shape or form and it may be that it's inadequate for this particular hazard but it but the control structure at the higher levels is for the entire facility so if we found an issue in this bit of the control structure is it that actually also applies elsewhere and we need to go and have a look and be careful about it and that extends across an entire business potentially at different sites different facilities you Know if you have a global uh process safety management system that requires certain approaches to plant design and analysis
if there's an issue with that process it could affect all of your facilities and you wouldn't necessarily realize until something something bad happened so always think about the bigger systematic perspective of your recommendations So process control system uh so yes cca contributory control action number one so in terms of the order of the analysis um i always recommend and it i think it's in the guidebook as well is always look at where is it so requirements and control actions is the first sort of section and then the second section is reasons why that happened Inside
requirements and control actions we have roles responsibilities requirements and constraints then we have control and feedback ability so this is about you know what control and feedback ability does this controller have what actuators do they have what sensors do they have were there any issues with those uh control action controllers with those actuators and sensors and Then the big part the important part is contributory control actions and interactions so this is essentially stpa step one let's look at the unsafe control actions that occurred uh or were not provided so going back to there we've got
here we go so you've got unsafe control actions there we've already talked about contributed control actions so For future cast use contributory control actions just to try and un unpick that potential bias the use of the word inadequate i think we spoke about that already so not process control system process model flaws they seem to indicate causes of process model flaws for operators mental model so these two here the tank directly vented to the Atmosphere instead of a safe location that's not strictly a process model floor unless we're talking about the people that are observing
the process and they don't know that it's been vented to atmosphere if that makes sense and then gas detectors location at tank installed area so that will lead to a process model floor not themselves being a process model floor if that makes sense So then contribute to control actions what i've highlighted is here is list the specific control actions that the process control uh pcs is capable of providing then identify and assess those that are in income as unsafe taking considerable care to recognize and treat them as contributory control actions and then i've explained the
reason why This word avoids inappropriate unfair disrespectful hindsight bias judgments and also minimizes biasing the investigators and analysts thinking an analysis if that makes sense we talked about that feedback circulating so the aim of circulating so circulating is a type of contributory control action so you can either provide circulation or not provide it or you Could potentially provide too much or too little too early too late too long too short pcs is unable to give correct information to the operator so is that a type of feedback to the operator or is it actually a control
action to the operator and we i have this conversation with with everybody that i work with um when we're talking about human machine interface design because you might say that you know like a A temperature gauge on the screen is that feedback or is that a control action well if if the presence of that temperature gauge on the screen does not make any effects to the human operators decision making or actions when they see it then it is just information it's just feedback and it might not be Relevant to the contributory control actions of the
human operator but if that temperature says you know it's going into the red zone uh or it's it should be showing that it's going into the red zone but it's not for whatever technical reason if the human operator is supposed to take action when it goes into the red zone if that display does not provide that information that contributory control action then the human operator will not Do what they need to do and so there's always a conceptual discussion about is it a control action or is it feedback and it depends on the context but
what matters is we have that conversation and think about it not just decide well it's feedback it's not relevant if that makes sense um so in terms of the analysis of the this uca one the pcs did not assist the overall at the operational personnel Um i've said to list the specific control actions that the pcs is capable of and then sort of go through that analysis in detail i've sort of emphasized here that this bit of the analysis is quite superficial it does not clarify control actions or systematically analyze the control loop or controller
eg actuator sensors control path feedback path control algorithm process model or control Action interactions from other controllers when i say it's a superficial analysis i'm not it's not a criticism of you it's a you know starting point you've done some good work here but it needs some more development and you could go really far into detail and look at the detail of those contributory control actions but also the causes of those contributory control actions um but that Requires a lot more sort of detail and a lot more work there and in terms of uca number
two there was no measurement for some parameters such as dissolved oxygen in liquid space so is that a feedback issue and not a control contributory control action and that might explain why a control action was not provided in terms of dissolved oxygen and headspace oxygen is an essential factor To the effective works so is that actually the provision of dissolved oxygen is that a contributory control action it was not provided to the process and that's why the process went out of control or what are we actually talking about and and who who provides that dissolved
oxygen is it the pcs or is it some human operators physically pouring it into a tank I mean i don't know the details of that but that goes back to this diagram this one so uh is there an operation shift team do they directly interact with what's in the styrene tank by physically placing stuff in there is there some control actions there that we Need to think about does that make sense yeah something uh dissolved control action they never mentioned but uh is it under pcs correct because they go to the rope to measure and
go to the controller give the alarm yes but there is nothing about this this tank okay excuse me that's good um all right there we go Uh uca3 there was no pcs to activate the cooling system yes so that is a contribute to control action definitely yes definitely um it's a refrigeration system only operated during day time by manually so that that's a contributor control action the refrigeration system was not operated by the humans and again that goes back to this diagram here Where do where do we stick the refrigeration system between here and here
in the diagram so the operations team have a control action to do something with the refrigeration system that then interacts with the storage tank but we've not included that control action that actuator coolant provision of refrigeration temperature and pressure that's what i covered in this tapa Ah right okay yeah yeah i covered in the stopper yeah um that's that's cool right so we've done that i've just lost the plot shall we so yeah we talked about that talked about this right zoom in a bit more so there were no gas gas detectors on the top
of the tank to activate the high level alarm yes so the The high level alarm that would be an actuator or control action and that would either go either to the process control system or directly to the human operators um so that so the alarm itself is the actuator the high level alarm noise or display is the control action and that goes to whoever it goes to the fact that there were no gas Detectors means that a sensor feedback was missing or inadequate so that would contribute to some contributed control action perhaps now there's a
bit of a balance here between what actually existed physically in the design so if we have no gas detectors then there are no possible logical control actions or feedback related to those gas Detectors now if there were gas detectors and they were failed they weren't working for whatever reason then it's loss of feedback or it's a loss of contribute to control action control contributed control action not provided so you know let's cons less becomes uh conceptually consistent about whether it's a control whether it's an actuator a sensor whether it's related to control actions Or feedback
now this one is interesting so interesting detail but not relevant to the cca assessment so if any detectors are located at the top of the tank those detectors give early detection to operators for taking proactive action so i think that's an important sort of observation but it's not really relevant to this particular uh contributory control action Uh number five there is no pcs to activate doing of the tvc when the polymerization notified so that's interesting that one um i think it's a key finding why no control action present in the design to deal with that
particular situation and i mean is that is the addition of tbc is that is the pcs capable of doing that or is it a manual process manual process Okay so you need to collect sample and go to qc analysis and come back that part i capture is stopper yeah so in terms of the control structure model then you could add the tbc control of tbc edition and i'd say there's two elements to that one is the actual use of tbc and when it's appropriate but the other is the Supply chain and availability of making sure
tbc is present on site in sufficient quantities to deal with uh whatever situation arises and i think that was talked about later on somewhere i think i remember a discussion about the supply not not the supply chain but the pr some of the operators did not know that this stuff was available on site for them to use and had they known that they might have taken action differently So let's keep going so so that's a really important aspect we could add details to the control structure model uh no information to the operator needs to dose
the tbc if suspect the polymerization occurred in the inadequate schedule of tbc yeah so feedback feedback control action so that's we need to sort of break that up a little bit and then summary of the role of the process control system in the accident Um when we talk about the process control system the pcs are we talking about a a electromechanical software control system or are we using that pcs label to also refer to individual people doing particular physical activities no no this yes the electro means this uh there's a measurement unit Send the signal
to the controller controller will uh analysis then control the items that is a digital controller systems like something is a basic uh process control system okay not manually manually means is it different if you have schedule like you need next sample that schedule is example that control is a different uh if like start in uh tbc newtop analysis yeah you See people take sample then the analysis that uh sent to the uh production teams this is an analysis report then analysis report based on the one they need to control the selling mode number okay but
that process will take a few hours has one analysis they need to collect a sample like 11 o'clock they collect they went there the analysis take one Hour one or two hours then from the report the report reviewed by supriser additional half an hour gone then they reported to the one it's too long that's quite a long feedback loop to know the status of the process too many two three layers involved uh yeah that's that's an interesting yeah if the pcs immediate okay they're like temperature go up the pcs sends the temperature is sent to
the controller controller Will control the flow of a chiller outlet chiller inlet that part i covered in the stopper right okay that's that's good that's cool uh so it was not designed to provide the necessary help to the operators to easily prevent the process from emergency so i've highlighted not designed to provide the necessary help again it's it's just been mindful of What we write in terms of are we being judgmental about the design and the people that made it are we stating facts you know clear facts as opposed to judgment judgment um and as
i said it's not easy it's hard to do it's hard to be that clear it's not easy uh mindful of judgement again this cast analysis does not explore the design control structure uh No we have no adequate study on the design during conversion okay but how do we know this is true we need to explore the design control structure and that might be recommendations for future work you don't necessarily have to do that now you know that there's a question there that needs to be answered and you can make a decision do we need to
go and look at that or not or can we just recommend change the Design have a better design in the future um after the contributor control action has been updated based on previous comments you'll need to revisit what these recommendations are they're quite superficial recommendations and they're not drawn directly from the control contributed control action analysis so i think some of these are good Recommendations but they there's sort of a disconnect between the analysis and the recommendations that have been made question how do we know it was already it was not already compliant being compliant
would fix this compliance is no guarantee of safety so this is um just emphasizing the recommendation about compliance with process safety standards How do we know it was not already compliant and if it was compliant would that have fixed and solved this problem that's not complaints it was not compliant was it because they never follow any because that's why to enhance emoji procedure yeah include and compliance with the one yes there is no process safety standard followed okay that's good that's it's uh so they So they need to follow a safety process but will follow
in this process get them where they need to go yeah yes they should never follow but if the wasa is american standards but they need to follow in the indian one there is the report doesn't mention what is the indian standard for process safety yeah so oh typically i took it it's a Washer washer it's a form all india's so i take it okay so uh operator's competence oh two hours ready is it working for you continue are you oh yeah guys it's getting late are you okay to continue because i'm okay to keep going
for another hour or so if you want to is it like No no it's okay for me but i guess i am worrying about your timing no i'm i'm fine it's saturday afternoon maybe have a comfort break in a little while um okay well this candidate recommendation does not originate from the pcs analysis should it be within this section so this talks about operator competency and that may well be an issue but it's not come from this part of the analysis this you know we've just gone through The pcs analysis and we've identified the recommendation
is it related to that analysis or would it come out elsewhere so it may be that this particular recommendation comes out of the analysis of the uh the humans involved later on to conduct a hazop study perhaps a hazop study was already done but that study missed this particular issue so there's a question there is this an appropriate recommendation do We need to investigate a little bit further there is no uh azab conductor oh so they didn't do well okay no they know process safety also not involved because when you do modifications or change in
use any plan you have to do that stuff okay i i was i was sort of thinking that that may Have already done some type of safety assessment and that might have included a hazop but if they didn't do any kind of safety assessment then i kind of expect oh that's good okay uh operation shift team includes control panel operator and team leader so what i've done here is just draw a basic control structure process control system control panel operator and then a team leader i've sort of emphasized the team Leader might be able to
see the process control system but they may not press any buttons that will be up to the control panel operator so i just i've just included that there just to sort of you know i i always whenever i approach these things i always think in the back of my mind you know what is the control structure involved in this bit of information that we're looking at what does it look like so this is a timeline Oh actually let me just talk about this at the top so dcs dcs operator there were two alarms there's the
ops safety uh ndo i've got night duty officer and then we've got everyone else plant staff public emergency staff we've got a fire hydrant sprinkler we've got the plant safety head and we've got a vapor cloud so this information you see these green things these have come from You know as i said before i use green to highlight uh controllers candidate controllers for the control structure and then i've used blue to emphasize feedback or control actions so i've gone through your timeline yellow indicates times so 2 54 4 15 5 15 6 30. i've looked
at some of the information here this here no alert to the public now this uh this could be a candidate for a contributory control action and That is alarms not provided so we might want to look at that as part of the investigation residents were found running away so you could argue that's a contribute that's a control action you know we need to escape uh which was designed to open automatically in case of fire by heat detection so if this deluge valve was designed to open automatically in Case of a fire by heat detection the
deluge valve is an actuator of some description and it provides uh quenching or fire uh extinguishing capability i think there was some mentioning of cooling capability as well you know using it as a coolant yeah so that is a control that's an actuator and that is a potential control action uh that they manually open the deluge valve so the ap fire service department team Arrived fire sprinklers were made functional uh by opening the deluge valve so the delgi valve controls the fire sprinklers which provides the control action of cooling and extinguishing and that i understand
was done by the ap fire service department so they they will be a controller in the control structure for emergency response of some description And the control action to provide cooling was provided by them the deluge is supposed to opened by the company emergency response team not the ap fire service service the government parties they should not involve they come here they notice the local people no the Operation people struggle to open yeah oh so they did it for them they helped them to do it okay so that's that's interesting because it may be that
the control structure includes the use of fire sprinklers for cooling and that's a recognized mitigation but the mitigation was not uh adequate because they couldn't manually open the valves and it required An external agency to come in and perform that function so there's a potential recommendation there on the fitness for purpose of the valves and the people trying to open them um and recognizing that if we didn't have the emergency services come in and do that that mitigation wouldn't have been available and it could have been worse you know escalation could have been Worse um
but here also there is okay the deluge valve supposed to be provided two actions yeah okay um the polymerization happened uh inside that tank so when polymerization is exothermic reaction temperature shoot up so we need To cool down the temperature polymerization yes and if anything fire happen need to quench that's the due to ts of deluge wall what this m6 tank is insulator ah see that if you have fire okay the religious wall do the do the Duties duties of uh quench the fire but not happen the fire the heat is inside the tank so
you need to cool down okay not cool down so this is this is interesting because for recovery of this or minimizing this accident they used a a piece of equipment which was there for uh fire control fire hazard control um and you could argue and i'm just sort of Moving the idea around you could argue that using that equipment for a purpose which it was not sort of required uh it was fortunate that it worked the question then is well if there was a fire or subsequent fire in the aftermath of this then that fire
mitigation may not be available or fit for purpose which is supposed to be used for so The questions become was the use of fire sprinklers actually considered as part of the original safety assessment i think we said they didn't do one uh but if it was not included then the decision making to use it in the real emergency how was that decision making made and what consideration was given for using a a fire mitigation For a purpose it's not designed and whether it was appropriate or not because there are i don't know about this case
but i'm aware of other situations where uh water is used to quench a fire but the type of fire is such that using water makes it much worse and perhaps even causes explosions and i'm not saying that you know was in this case but you know if there's a If there is a drive to use water in the plant are there certain areas of the plant where it's not a good idea to use water and it might make things worse you know so there's this particular line has quite a few questions i'd like to explore
as part of the investigation and it's an interesting one because it crosses over between work as imagined workers done versus Oh no what do we do this is emergency oh that looks like a good idea let's do it and who is involved in that decision making because i'd like to say that whoever made those decisions was well aware of the potential for using water when it's not appropriate and they knew it was okay to do it if they didn't know it was okay to do it there's a recommendation there for making sure that everybody is
aware of where it is appropriate and Not appropriate to use water to deal with a chemical process like that one suppose i could have said all of that much more concisely it's been a long day so far um but if the serene monomer store in your country the control measure is a different because the temperature is low yeah the temperature is low means I think the melting point is uh minus 17 degree celsius okay so it will be solid yeah increase it out so that is happened in my one of the company at the german
they went to the low level and also creating issues so mastering monomer stern used all the plastic and the polymers but is it is quite nasty Chemicals yeah oh okay uh question oh 6 40 a.m why was the three hours more than three hours response time um that's a question for the investigation why was there a three-hour response time and is that appropriate or not if it's not appropriate which i'd argue it's not appropriate then why was that what does the investigation need to find and what recommendations do we need to Make in terms of
risk associated with this operation uh let's have a look 70 tons of styrene pumped into one spare storage tank therefore considering the risk associated with this operation it was stopped so when we say what risk please explain what other hazards are involved and what does this mean for our control structure model does that mean we need to have more analysis so there's this Bit about bursting of the tank and the risk associated with transferring styrene into another storage tank so at the beginning of the analysis we were talking about uh x uh let's have a
look the two hazards we were talking about were exposed to toxic styrene vapor release of toxic styrene vapor where is this bit about um Where is it where did we get to yeah this bit about uh pumping styrene from one tank to another are there different hazards involved in that that we need to think about from a control structure perspective so one of those is the other thing will affect it yeah so that so that might that particular element might need to add Extra hazards and high level safety constraints to think about uh extra detail
about control structures related to those hazards and then how do you know another one you know so you is another one or term coming the loop means you have to analyze another control structure not necessarily i mean the way i look at it is uh the initial part of the analysis was Set in the context of science uh styrene monomer vapor release so we didn't look at tank bursting or uncontrolled release of tons of styrene into the facility we were only looking down the narrow view of styrene vapor release because that's what contributed to harm
and injuries and deaths of sight whereas this part of the accident they're talking about pumping styrene around so What other hazards might we want to consider as part of the analysis and it may be that we don't need to look at these in any detail because it's not relevant to the accident uh which it's not but with uh in trying to recover the sorry in trying to recover the hazards to do with styrene vapor release that control structure and its behavior Started to pump tyr styrene into a different tank and the hazards associated with that
activity what how were they adequately controlled what was the control structure for those and how do those control structures fit together because like the use of the sprinklers the emergency fire water in that particular case we're talking about uh Using fire fighting sprinklers on hazards that are not firefighting hazards so we're sort of misusing the that bit of the control structure whereas in this case we're talking about bursting of tanks and release of uh 70 tons of styrene uncontrolled release of styrene into the facility that's the hazard that we have to be mindful of when
Transferring uh styrene from one tank to another so the control structure conceptually involved in that activity it involves the same people you know they're the same operators you know in the same functions but the control structure for controlling the hazards of uh tank bursting and release we've not explored those up until this point so we've not got a control structure model for those now i'm not Suggesting we certainly need one it may be that the way that they pump styrene from one tank to the other is a perfectly well understood well practiced activity but if
it was done you know outside of procedures in this occasion as part of the investigation it's certainly something we should look at because it may well be they started doing that operation And then it was stopped so did they stop it because they realized actually we probably shouldn't be doing this because it's actually high risk or was it stopped because of some other normal reason do you see where i'm sort of coming from it's it's not an easy sort of thing to think about because we're almost looking at we're sort of taking a step back
and Looking at well what are the hazards involved in this accident there are the vapor release hazards but then in this case we've got tank bursting hazards or pumping of styrene related hazards but the pumping of styrene or tank bursting was not involved in this accident so do we draw a boundary around it and forget about it or do we take the opportunity to explore it and see whether There is any learning we can have such as if anything on styrene vapor release happens in the future there is a requirement that says do not under
any circumstances try and pump the styrene into a different tank because if you do you might expose yourself to other hazards that are not obvious at the time that you're dealing with styrene vapor release i suppose I'm trying to think of a a silly example would be we need to use the electrical supply from a different part of the facility to stop this accident getting worse and so they go and unplug things in this other part of the facility and plug it into what they're doing to save the day but by unplugging the electrical power
to this other part of the facility they actually start another different accident And if they had known before everything went to that unplugging this would be a problem they would know not to do it okay and so for recommendations for future disaster prevention we might find here today in this analysis it was not a good idea to unplug this other part of the facility and so the recommendation for the future is do not unplug that part of the facility and Everyone needs to know about it does that make sense yes yes makes sense because you're
pumping to the other tank they're creating one more hiding hazard yes the creating hazard and the hiding hazard also is there yes but due to the emergency situations uh yes they because the polymerization happened on the top of the tank not the bottom of the tank so the decision they want to cool polymer to pull out yeah so make more space on the head face Uh to stop the polymerization yes the lesser amount that is the reason but if you the the cold item is go to the another tank it will cause us some impact
on the another tank yeah so it will create another issues yeah so that's the reason they stop it but uh control such a definitely can analysis not that very high level maybe so expected uh Maybe if you're doing like uh what will happen that is analysis separation not here like increasing a lot of analysis well it's it this is the beauty of the control structure model and how you develop it you can sort of postulate questions and then make decisions about well do we want to go and look at that in detail or do we
just want to make a draw a line and put a recommendation That says uh you know the pumping of 70 tons of styrene into a spare tank is okay to do in emergency situations or it's not and the assessment is going to be done to confirm that or we're not going to do it so don't do it in the future well i mean obviously we have to we have to be mindful of you know in emergency situations the people making the decisions to take Emergency actions they it's a chaotic situation there's not much time to
take action and delays might make things worse so it's worth taking the risk and do it taking particular actions to see if it minimizes or reduces the overall impact um i mean as a as a you know an example the the chernobyl uh reactor explosion in 1986 you know that there was graph Radioactive graphite spread all over the place which was deadly to people but to minimize the overall risk to everybody else individuals liquidators were sent in with minimal protection to deal with the problem immediately and if you would if you had the luxury of
time to sit back and do a workplace risk assessment you'd probably say that that's probably not a good thing to do but when something really serious happens You know really painful difficult decisions have to be made yeah and you know ideally we want to look at those decisions in the design phase you know before something negative happens and in in emergency disaster planning you know as part of emergency disaster planning what things can we do what things should we definitely not do because they might make things even worse You know um and the recommendations from
the investigation might be just go and look at that explore the emergency decision making and actions involved in using the second spare tank is it appropriate or not was it appropriate or not what other mitigations do we need but draw a line draw a line through it um analysis is okay to go through Absolutely 2145 the evacuation has been activated so question why was it 18 hours before an evacuation control action was provided uh that's what i'd like to look at uh relevant safety related responsibilities i've said it's generally a good idea to check how
responsibilities align with control actions contributor control actions because it also provides a sanity check for the completeness of the Analysis like if there's no responsibilities what does that mean for a contributory control action so so for example if if in in the case of this evacuation activation if the plan operators were not given a formal responsibility to initiate the evacuation alarm to the general population Then the reason that control action took 18 hours could be explained because they were not allocated that responsibility that could be a quick quick win if they were allocated that responsibility
the next question becomes well let's interview those people how was that responsibility communicated to you how did you understand the importance of it and what explains why there was an 18-hour delay and if the if the thing is well we Didn't know to do it or it wasn't important we had more important things to do that suggests that the allocation of that responsibility was inadequate so the question becomes well who what organization allocated that responsibility and why did it not recognize the potential that frontline operators are going to be rather busy dealing with the investor
dealing with the accident than being able to initiate Emergency evacuation instructions yeah here there is one more part also playing part the decision making so previously i highlighted is a competency for the emergency operator technician yeah the plant is shortage of chemical engineers okay Okay so uh so based on the experience they're running the plant because the plant is very old so those are like a long service become promoter production manager and one so the main chemical engineer are left from this company so those are working there is all experienced experience means in terms of
Plant operation not the academic quality okay that's interesting so uh decision making is too long that's it 8 31 45 evacuation was activated 22 45 22 45 45 154 that is a runaway reaction reach so the the the the temperature slowly clamping up they they thought there is a localized issue they can make it So once the temperature if not cool down the temperature went to more than 200 degrees celsius the tank will burst right if the first is a critical issues lucky not best well it's interesting that the decision to well for consideration the
decision to evacuate should be separate from the decision about temperatures Because the you don't want people directly involved in the physical process to be making decisions about evacuating thousands of people potentially that responsibility should be higher in the control structure and so that they're not you know uh from a human factors and human performance perspective they're not trying to balance the embarrassment of evacuating 10 000 people Versus you know temperature on a temperature gauge between 150 and 200 that's that's not appropriate so who decides and gives orders for the control action for evacuation uh process
model flaws that should come after control action analysis uh this is a statement that does not explain any detail as to what was flawed or why so there needs to be a little bit more Detail there okay contextual factors uh human behavior okay great was the stress fear surprise dimensions to it and again that should be after control action analysis so there's your contributed control actions um did not halt the indication polymerization uh adding the inhibitors immediately so there's two control actions there Uh operation did not know so that's a process model mental model floor
aspect yeah so the control action is this part it was late at 05 15 and it was over 100 degrees so there's context there to think about no processing system processing system control system to dose inhibitor automatically so is that something that needs to be considered technically or other emergency Procedures that need to cover that um cooling system only operated manually during the daytime the operator did not know that the refrigeration needs to start so it's that process model mental model flaw and then if it's only operated during the day time that is the that's
the contributory control actions that we're talking about that is the issue on there's no chemical engineers yeah The people working they know that chemistry background chemistry background uh it's not equal to chemical engineering background okay okay okay chemistry only though though the reactions are blah but the chemical engineers only know uh how the equipment behave to the chemicals stored okay so how much you need rate of cooling or everything these type Of things they have to they know how fast can pull down everything okay uh so process model mental model floor for the people adding
ethel benzene the duty officer did not activate the sprinkler so there's an implied process model mental model flaw here so did they not know to activate it from the control room or what was the you Know why was that control action not provided uh and then this one is showing the inefficiency of emergency communication so where we say inefficiency it's again it's one of those words that's judgmental let's look at it explore it you know from a control action perspective why was there a delay did they have issues with communication equipment or understanding How serious
this situation was so that's a summary of the role the operator is in the accident the last part i've just said this is great background information but it's not relevant to the analysis of the operators okay um recommendations automated process control system for the styrene monitor tank should be implemented i've suggested that that's a bit of a big reach It's not supported by the analysis so far again this is trying to when we're going through the process of the analysis we need to do do it in order and then think about well how does that
analysis generate recommendations and those recommendations are candidates until we refine them later on when we've done all the other analysis now of course when you're doing this analysis In sequence you might identify recommendations that are immediate action recommendations immediate safety relevant recommendations which you'll want to communicate to whoever needs to take action but always be mindful about how they might evolve as you learn more about the situation i've said it does not address all the contributor control actions or their causes Why emergency evacuation call was delayed plus others so these recommendations need to link back
to those uh contributory control actions okay um okay so we've got plant operation management and then a lot of other stuff um i'm just thinking there's quite a lot there to talk through and we've been going two and a half hours Let's let's jump to the next uh towards the end then so uh the structure analysis is missing uh and i've also talked about refining candidate recommendations uh into actual recommendations so when i say that let's just jump back to here so on this first page uh so we've got the three parts the analysis physical
controller and structure and it's these this element here that's Uh missing from the document so far you do talk a bit about this and a bit about this but there isn't a specific section that looks at structure the control structure itself so what you could do is add a section called control structure analysis and then have six or seven headings that look at each of these topics and any others that are relevant because as i say these six topics are uh you know broad Things to look at but there could be other things that are
more specific that don't come under these six categories so have a look at these six and anything else and then include looking at reasons why context contextual factors um this dynamics changes and evolution over time i think you mentioned about the the facility being sold to different businesses so it was acquired and operated by different Companies during its uh during its life so you could look at that and then results of the cast analysis so you've got uh let's zoom in a bit you talk about uh this project was focused on the safety control structures
that belong to lg polymer that's great that you've explained that um if you're doing a full-on cast analysis investigation obviously you'd look at all of the other elements in the control structure as Well beyond lg polymer but if you're a if you work for lg polymer or a company doing this kind of work naturally you're going to stop at the boundaries of your investigation of your uh your responsibility but at least you can look across the interfaces um so recommendations these relate i think you summarize these from earlier in the documents for each of the
components in The control structure yes so yeah so this number one sounds like a reasonable recommendation but it has not originated from the analysis to replace the tank so going back to what i was saying at the beginning when you reference [Music] recommendations and safety constraints and things like that try to give them a reference number so you can use them across the document and then make sure That you can trace so when you when your analysis generates a candidate recommendation give it a reference unique reference number and then when you come to bring the
recommendations together refined make sure you don't miss anything drop anything off okay and likewise with this one you've added replace you recommend to replace the tank That's that might be a perfectly reasonable recommendation but how was it originated from the analysis you know where would it originate from um so no recommendations are not exactly the same as those listed in the main body and it's unclear as to why i you know i don't know why that might be as can you give me a little explanation about that Recommendation uh report exactly the same i suggested
the main body i don't want to confuse one more two recommendations because down there also recommendation here also recommended so i want to follow the same thing okay that's fine that's fine if it's just confused given to two why the two recommendations it's just um in terms of the cast Analysis itself it's just making sure that when we do the analysis and make a recommendation we don't we don't lose those in the detail we capture them somewhere in a list but also if we make any other recommendations even if there you know make sense and
they're logical let's just ask the question how would they how would those recommendations have appeared in the cast analysis is there something that We've missed in the cast analysis that we need to go and explore so where we're talking about uh m6 that the tank was fabricated half a century ago and we recommend to replace the tank that that could have come from the structure analysis or the physical this analysis here so physical context this this tank was 50 years old it wasn't maintained properly recommendation replaced the tank So that recommendation could have come out
of this part of the analysis and likewise changes dynamics changes and evolution the age of the tank the way that it was modified and things like that could generate that recommendation that you've highlighted here so it's a good what i'm trying to say it's a good sanity check that if you if you come up with a recommendation where would it originate from if it Hasn't originated from the analysis do i need to go and add some detail to the analysis okay um a missing requirement from the process control is with regards to management of change
so we need a recommendation in there for improving management of change yes um and then the rest of these for the Process safety department whilst we didn't look at that in a lot of detail in the body of the analysis there is a new requirement that's not in the analysis section which is number 25 procedures for incorporating and using lessons learned need to be established or improved so that's a good recommendation but i would argue that it goes back to controller analysis so for the process safety team One of their roles and responsibilities is to
learn from experience so they need to learn from experience and feed that back into the business into the processes and procedures and everything else but if they have a missing if they have a missing responsibility to do that our recommendation is give them that responsibility and then empower them to be able to do that Learning from experience and one of those recommendations might be learn how to use cast accidents analysis and stpa and do hazops and osha process safety assessments as you've highlighted later on so let's go back to there so so let's go to
here so safety culture there appear to be flaws in the safety culture that should be corrected okay so let's just jump back to Oh what happened then okay let's just jump back to here so safety culture is that one there and that's in the structure analysis i mean you could include it in if there is evidence that particular controllers have safety culture issues obviously you'd include that in context contextual factors but if you're looking at the overall structure safety culture is in there Um uh where are we so we talked about safe culture um root
cause and negligence failure were identified high power committee is compared with the role in respect to recommendations from caste uh so this was i wasn't quite sure i understood this statement the root cause and negligence failure were identified Uh high power committee is compared with the role and respective recommendations from cass i wasn't quite sure i understood that because root cause and negligence failures are not conceptually identified by caste we look for contributory control actions uh physical aspects and structural aspects of the control structure um this root cause and negligence failure Is a from the
investigation report all right okay okay so the cost investigation report is used for fishbone diagram here so now the cost is we are doing so we capture the point we are capturing any additional point of investigation so based on your discussion now we can we can we can explore more and more and identify all the flash and uh uh control uh The root cause may potential causes excellent okay now that's good that's the party possibility that this section i want to compare yeah it's it's something i always suggest uh if you're gonna when you're practicing
how to do cast and you're using existing investigation reports i always recommend once you've finished your cast work and got your candidate recommendations compare those recommendations with what's in the official report and What you'll find is the official report will probably have this many recommendations and you'll have this many recommendations and some of them will be uh one for one you know relations but what you might also find is that there are two or three specific cast recommendations that are more detailed than those that are in the official investigation report but it is always Good
practice to do that and as well if there's a recommendation in the official report that's not captured in your cast like like i mentioned earlier on go back and ask the question well given that the official investigation made a particular uh recommendation where would that recommendation appear in the analysis of my cast because it May be that you've missed something in the cast analysis it may be that there's an area that you didn't dig into as much detail as the official investigation did and so that's an opportunity for you to learn about your application of
cast and how detailed it was okay um so that's that uh Well i've got three hours 30. yeah we're nearly at the end it's page 26 now um so this challenge i think i i think this this table is incomplete i feel like just now when i said yeah i see that one is uh i i incomplete actually analyst yeah that's that's fine i mean it's a draft anyway it's you know work in progress um This is just me conceptually picking a point uh when we talk about sowing the seeds of disaster that metaphor is
very powerful but it i'm not a big fan of it because it implies causality and hindsight bias and so i'm just sort of again just being mindful about language and analogies because these things just roll off our tongue and then we use them in everyday language but when we're talking about accident Analysis we have to be mindful that as an investigator we are we have a duty to sort of step back and try and look at the cause and effect and the uh you know seen the the the context that caused these things to happen
and be mindful that we're not uh judge in unfairly judging these things right if that makes sense um again the same we should have the tank should have had either a flare System or cryogenic system this sort of implies that we're talking about a situation that didn't exist you know they didn't have these things that might have contributed to the accident but but the important question is why they didn't have them you know it may be that an assessment was done that said they didn't need to do it and whether that assessment was right or
wrong is not really the Question it's what they thought was the appropriate action or not and then again the same with the restricted is this a judgment by the investigation if so it's not appropriate temperature measurement is restricted to the bottom zone liquid only is it that it was required to be restricted to the bottom or we think it's restricted as an investigator and we think it wasn't appropriate It might come across that i'm trying to be pedantic and it's not it's just trying to be really clear and mindful about our role as an investigator
and trying not to uh add our own biases and judgments subconsciously and it's not an easy thing to do it is very hard human error is not addressed by the analysis presented so there was no real detail in this Analysis i've seen looking at the human factors aspects which which is fine because it's a draft and there's further work but that's something that will need to be added later on um these three items will need to update the cast analysis for the operations team so what contributor control actions did they provide or not Uh we
didn't analyze those with regards to not analyzing samples with the required periodicity let's you know let's look at that in a bit more detail later on um the contributor control action not adding a tbc contributor control action plan operation department not providing the tbc stock again that's something that should be Looked at and added to the control structure model perhaps uh stalactite formation an action to explore the design change control structure i think we spoke about that and this one has actions to explore these three elements further uh lg polymers was closed during kobit 19
lockdown period as it's a non-essential industry well that may well be true it's not essential from a covid management government Perspective but from a safety perspective it was absolutely essential that those safety processes were under adequate control and regardless of what you know maybe i've been a bit controversial here but this facility this plant has to manage its hazards using a control structure and if other people regulations laws the government come and say hey you need to take these actions Safety management has to consider what those implications are you can't just turn the you know
lock the door and turn the lights off and go home the control structure has to work so process safety above all else there's an action there to explore as many uh as an action there to explore as this might have government issues that span more than this facility this industry you know so your factory next door your Next door neighbor if they decide to just lock the doors and turn the lights out um and then this one the emergency sirens was not activated as a result many inhabitants outside the plant area were engulfed in the
vapor cloud so there's no recommendations on that part but i think that's because you said it's still in development which is fine Yeah yeah um and that's i think if i just zoom out a bit that's the only last thing really is to talk about this so this is a more detailed safety control structure where you've added some control actions and feedback and i think that's really great but the first thing i've highlighted is what is the closed loop for the safety management team You know what information do they have about operations technical and the
storage tank itself because if the plant operation management they are one of their responsibilities is to make sure that their equipment and plant is maintained in a state that is safe now the safety management i would like to think would have an independent audit of some description to make sure that it is Maintained properly that the process control system has uh got appropriate controls in place for the styrene tank that the operation shift team do know and have the availability of tbc inhibitor when they need it you know so this red arrow is just to
sort of emphasize what feedback does the safety management team have to make sure that the instructions they've given to plant operations management have Actually been uh enforced and put into place and and that relates to this action here uh develop the control structures further be clear on operations and management of change structures uh develop analysis further look at the controllers that you develop and the structure so that red section that was missing check and refine candidate Recommendations and then ensure clear traceability between analysis and those recommendations and i think you'll be on i think you'll
be on to a winner um failure failed i just want to emphasize that there's not just failures involved in this accident we should also consider dangerous success i.e things that behaved as designed intended but again that's just a you know we're Used to talking about failure and success failure bad success good and what i'm trying to do is say well there is a class of success which is actually dangerous and bad that we need to be mindful of and that's that's everything really we've not gone through uh which one is this one we've not gone
through Plant operation management or plant process safety management uh so those two controllers but you i mean the theme the themes i've discussed already apply to these as well so if you just have a read through those comments and have a think about it then i'm happy to answer any other questions on it