They removed the wrong bolts. The valve was built to a standard that says that cannot happen.

The easy version of La Porte is that three contractors removed the wrong bolts. They did. The harder part, the part that should bother you more than it does, is that the valve they were working on had already been redesigned specifically so that removing the wrong bolts could not do this.

Dusty Day and Shawn Kuhleman went to work on the evening of July 27, 2021, to take a piece of equipment off a valve. Neither of them came home. The job they were sent to do was not supposed to open the process at all.

The LyondellBasell complex in La Porte, Texas, twenty-five miles east of Houston, is the world's third-largest producer of acetic acid. The plant makes it by reacting methanol with carbon monoxide over a catalyst, and one of the catalyst additives is methyl iodide.

Start with what those two do to a person. Acetic acid at industrial strength burns skin and destroys eye tissue at room temperature, and this was not at room temperature. It was 238 degrees Fahrenheit, held liquid by 130 psi, and acetic acid boils at 244. Being sprayed with it is a chemical burn and a scald at once. Methyl iodide is worse in a different way. It boils at 109, so a liquid mixture at 238 degrees does not spray it, it flashes it into vapor the instant it hits air, and that vapor is nearly five times heavier than air, so it sinks and stays at the level where people are standing. NIOSH lists it as a potential occupational carcinogen. Someone caught in that release gets burned by one chemical and breathes the other.

The reactor held roughly 164,000 pounds of that mixture.

Now the hardware. Feeding it was a methanol line, and on that line sat an eight-inch plug valve. A plug valve is the simplest shut-off there is: a tapered plug with a hole through it sits in the valve body, and a quarter turn either lines the hole up with the pipe or turns it broadside. The plug is held in by a cover, and the cover is held on by bolts. Those bolts are the pressure boundary. Bolted on top was an actuator, a pneumatic device that turns the plug so nobody has to do it by hand. The actuator is not part of the pressure boundary. It's a motor sitting on a bracket on top of a pressure vessel.

Two sets of fasteners, inches apart, on the same assembly. One set holds a motor on. The other set holds the plant in.

On Saturday, July 24, 2021, an operations technician found a small leak in the methanol piping upstream of the reactor, traced to a weld. An adjacent unit had gone down anyway, so LyondellBasell made the call every one of us has made: we're down, let's fix the leaking spool while we've got the window.

To cut that pipe out they had to isolate it from the reactor, and the plug valve was the isolation point. Their Energy Isolation Procedure only counted an actuated valve as an approved isolation device if it had a hand jack that could be closed and locked. This one did not. So the plan was to pull the actuator off, slide a pipe tee over the bare stem, run a chain through it and padlock the chain, producing something a lock would physically fit on.

The job existed to satisfy lockout/tagout. That's the whole reason anyone put a wrench on that valve.

LyondellBasell gave the work to Turn2 Specialty Companies, a turnaround contractor already on site, after Turn2 said it had a qualified night crew. There was no procedure for removing the actuator, and neither company trained the crew on how to do it. At 6:45 p.m. the permit was issued, an operator walked three men out and pointed at the actuator, then left. That crew had been repairing heat exchanger tubes as welders and welder helpers, and the CSB found no documentation that any of them had ever installed, repaired or removed a valve actuator.

They stripped the insulation and started on the bracket bolts. Partway through they hit nuts they could not reach with what they had, and the foreman went to his truck for a socket set. He came back, and the crew removed those nuts.

Those were the pressure-retaining nuts, holding the valve cover on. Nothing about them looked different from the bracket bolts and none were labeled. One of the workers explained his reasoning to the CSB in a sentence I've heard some version of on every mechanical job I've stood next to: "it's all connected. As [the actuator] sits up there, it's all connected to one another."

He was right that it was all connected. He had no way to know that some of it was connected to 164,000 pounds of near-boiling acid.

They lifted the actuator clear and set it on the grating. A coupler was still seated on the valve stem, too tight to pull by hand, so they went at it with a pry bar. The valve had been holding itself together on friction alone since those nuts came off. The pry bar and 130 psi finished it. The cover and plug ejected, and the entire contents of the reactor, all 164,000 pounds at 238 degrees, emptied through the opening onto the three men standing under it.

Dusty Day and Shawn Kuhleman died of chemical burns and inhalation injuries. The third man was seriously injured, along with a LyondellBasell responder who went in after them, and twenty-nine other people were taken for medical evaluation.

Work this the fast way and it closes in four questions.

Were the right bolts identified in the design? Yes. The actuator mounting bolts are a distinct set, and removing them is all the job required. Did the crew remove bolts the job did not call for? Yes. Did anyone tell them to? No. Did they then apply mechanical force to a valve they would already unfastened? Yes, with a pry bar.

Unqualified contractors exceeded the scope of the task and disassembled a pressurized valve. Case closed, corrective action is a contractor qualification audit and a stand-down on scope discipline.

I've sat through that meeting. It is not wrong about what happened and it is completely useless, because it cannot answer the only question that matters at the next plant: how was a man on the deck supposed to tell those two sets of fasteners apart? Look at the photo and try it yourself. If you cannot do it from a picture, your crew cannot do it in the dark either.

The CSB did something here that most internal investigations never do. Instead of asking why these men made this mistake, they went looking for whether anyone else had ever made it.

Those two questions behave differently, and the difference is the whole ballgame. "Why did this crew do this" always has an answer waiting. They had not been trained, they did not know which fasteners were which, they reached for a pry bar when the coupler stuck. Every one of those is true. Every one points at the three men on the deck, and the question is built so that it stops the moment you find a human being, which is why so many investigations stop there.

"Has anyone else ever done this" can come back empty. That is what makes it worth asking. If nobody in the industry had ever pulled a plug valve apart by mistake, the crew explanation would have gotten stronger, not weaker, and training really would have been the answer.

They found four.

Puebla, Mexico, 1977. A worker pulling an actuator off a plug valve under a vinyl chloride tank took the pressure-retaining bolts instead. One dead, four severely injured, nearly ninety more burned.

An Amoco plant near New Castle, Delaware, 1980. Mechanics removing an actuator so the valve could not be opened during a cleanup, a safety measure required by procedure, took the cover bolts. Six died.

AkzoNobel, La Porte, Texas, 2013. Same town. A contract worker took the actuator mounting bolts, which also held the cover on, and was burned by pyrophoric butyl ethyl magnesium. The company's own investigation named the absence of qualified supervision and of a job-specific procedure.

ExxonMobil's Baton Rouge refinery, 2016, is the one that matters most, because it is the near-miss that was not a miss. An operator removing an inoperable gearbox took the top-cap bolts, then put a pipe wrench on the valve. It came apart and released isobutane, which found an ignition source within thirty seconds and severely burned four workers. The CSB investigated and published a public safety bulletin in September 2017 telling the whole industry to survey their valves, find the ones where actuating hardware attaches to pressure-retaining components, and consider upgrading to a newer design. ExxonMobil did the survey and found fifteen susceptible valves in that one alkylation unit.

The warning went out to everybody. Four years later three more men were standing under a valve nobody had marked, in a town that had already had one of these.

Now the part I could not get past. LyondellBasell's valve was the newer design. The CSB says so explicitly, and quotes API Standard 599: the design "shall ensure the stem cannot be ejected from the valve by removal of actuator mounting hardware," and valves capable of mounting actuators "shall be capable of doing so without removal of any pressure-containing parts."

The valve complied with the clause written to prevent this. Two men are dead anyway.

That's not a paradox, it is a gap in what the clause promises. API 599 guaranteed that removing the actuator mounting hardware could not eject the stem. It said nothing about a man who removes something else by mistake, because nothing on the valve told him which was which. The standard engineered out the correct procedure performed on a wrong assumption, not the wrong procedure, and the wrong procedure is what five crews across forty-four years actually did. Neither API 599 nor ASME B16.34 required anyone to distinguish a pressure-retaining fastener from one that is not. A LyondellBasell worker put it plainly: "Nothing was marked. None of the bolts were labeled 'remove' or 'don't remove.'"

Trevor Kletz had proposed the fix decades earlier, in a book whose whole thesis is try to change situations, not people: paint the bolts that are safe to undo under pressure green and the others red. The CSB quotes him. It costs a can of paint. After the incident LyondellBasell marked the valve cover fasteners in that unit with tamper-resistant metal tape, which the report shows as a blue X across the bolts. That is the fix, and it had been available the entire forty-four years.

So in May 2023 the CSB wrote to the three bodies that control this, API, ASME and the Valve Manufacturers Association: mark the valves already out there, and design the new ones so this cannot happen. API published a new edition of Standard 599 in October 2025 carrying a warning about these incidents, a labeling requirement for pressure-retaining fasteners and a strengthened design requirement, and in January 2026 the CSB closed that recommendation. Real movement, and it took two and a half years.

But read what API declined. The two pieces aimed at existing plug valves, write procedures for the ones in your plant and mark the ones in your plant, were left out, on the grounds that Standard 599 governs new manufacture and does not address valves already in service. Defensible as a statement about what a manufacturing standard is for. It also means the fix reaches valves not yet built. The recommendations to ASME and to the valve manufacturers, the ones that would have reached further, are both still open three years on.

Which puts the marking of every actuated plug valve currently bolted into your plant exactly where it was on July 27, 2021. With you.

Three moves here are worth stealing.

The first is recurrence analysis, and it's the cheapest, most underused move in incident investigation. The count is what does the work. When the same mistake turns up across five companies, five crews, four countries and five decades, the variable that stayed constant is not the people, it is the hardware, and you cannot reach that conclusion by interviewing the crew no matter how well you interview them. You can run it this week: take the incident in front of you, strip the names out, describe it as a mechanism in one sentence, then search your own history, your corporate database and the public bulletins for that mechanism. Two hits means you have a class, and a class has a design answer.

The second is a set of three questions the process safety author Roy Sanders published in 1996 about an inadvertent valve disassembly. Note the year. The CSB structured its whole analysis around them. Is the risk too high to do this work while the equipment is full of highly hazardous liquid? Do written procedures exist detailing this job and its precautions? Were the mechanics trained in those procedures? At LyondellBasell the answers ran no, no and no.

That first question deserves an argument rather than an easy answer, and I don't think it has a clean one. De-inventorying a reactor is expensive, takes days, and carries its own hazards in the draining and purging. Set the bar at "de-inventory for any work on a valve in hazardous service" and you'll be down constantly and create fresh exposure every time. Set it at "the actuator isn't part of the pressure boundary, so proceed live" and you get La Porte. The CSB does not resolve it either. I'd like to know where you've drawn the line.

The third is what the CSB reached for once they were past the crew: the hierarchy of controls, applied through Prevention through Design. Marking the bolts is an administrative control and it works, which is why it is in the checklist below. But the real answer is a valve that cannot come apart in a person's hands: recessed fasteners, captive hardware, a bracket that mounts to the flange instead of the cover. Design out the possibility and no crew's training records matter.

Two things should follow you back to your own plant, and neither is about contractor qualification.

The first is that the job existed because of a safety procedure. Nobody was cutting a corner. The energy isolation procedure required a lockable isolation device, this valve could not accept a lock as configured, so somebody planned a modification to make it lockable, and that modification, executed to satisfy lockout/tagout, is what breached the process. I've written the clause that says the device has to be capable of accepting a lock. I've never once asked what a crew would have to physically take apart to make that sentence true on a valve that does not already comply.

The second is that "simple task" is a risk classification, and at most plants it is an unwritten one. Both companies called the actuator removal simple. That judgment did all the work: it is why there was no procedure, no training, no risk assessment and no LyondellBasell employee standing there. The label gets applied by whoever plans the job, based on how the task sounds, and once it lands every safeguard the plant owns is quietly waived without anybody signing anything.

When enforcement caught up, most of it did not stick. OSHA cited LyondellBasell in January 2022 with proposed penalties of $54,612 and the company contested. In June 2024 a judge vacated the citations covering safe work practices, energy control, mechanical integrity and management of change, and affirmed exactly one item, the failure to inform the contractor of known toxic release hazards, at $13,653. That decision went to Commission review in August 2024 and is still pending, so it is not even final. Five years on, two men dead, one contested citation.

Save this next part or screenshot it, and take ten minutes with it at your next toolbox meeting.

1. Go look at one actuated plug valve in hazardous service and try to tell the bolts apart. Don't pull the drawing, don't ask the engineer, don't call the vendor. Stand in front of it the way a millwright would at 7 p.m. and point at which fasteners hold the actuator and which hold the pressure boundary. If you cannot do it in thirty seconds from the outside, neither can the crew you are about to send, and the fix is a paint pen and a tag this week, not a capital project next year.

2. Decide, in writing, which valve work requires the process to be de-inventoried. This is the check I expect people to argue with, and it should be argued about, at your site, with operations in the room. What matters is that somebody with authority writes the line down and signs it, because right now that call is made silently, job by job, by whoever writes the permit. An unwritten line moves every time production is behind.

3. Find out who at your site is allowed to call a job "simple," and what that word turns off. When a task gets called routine, which requirements drop away: the written procedure, the JSA, the training verification, the owner's rep on site? Then take the last ten contractor jobs under that label and check whether any touched a pressure boundary. That is the audit, and it takes an afternoon.

The investigators at La Porte did not need a better theory of why contractors take shortcuts. They needed to count how many times this had happened before, and then to notice that the answer had been sitting in a can of paint since 1977.

Which of your safeguards would survive being described as a simple task? Reply and tell me. I read every one.

Before Human Error is a weekly teardown of industrial incidents for EHS and operations leaders, built from public investigation findings. Subscribe here on LinkedIn to get each issue. The inbox edition, plus the Process-Safety Teardown Checklist, is linked in the first comment.

Sources: U.S. Chemical Safety and Hazard Investigation Board, Fatal Release of Acetic Acid and Methyl Iodide Mixture at LyondellBasell La Porte Complex, Report No. 2021-05-I-TX, 25 May 2023. CSB Safety Bulletin 2016-02-I-LA (ExxonMobil Baton Rouge, September 2017). CSB recommendation status summaries, 2021-05-I-TX R4, R5 and R6. OSHRC Docket 22-0153, ALJ Calhoun, 27 June 2024. OSHA citation, January 2022. ABC13 Houston and KHOU, July 2021. Kletz, An Engineer’s View of Human Error, 3rd ed., p.175. Sanders & Spier, Process Safety Progress 15(4), 1996.