Life Vs. Lab
In the laboratory, a process for making triallyl cyanurate (TAC) used by a Georgia chemical plant in April 2004 worked perfectly. However, when they attempted to make it in the production reactor, the process radically overheated and released a toxic and flammable vapor cloud that forced the evacuation of nearly 200 families, states a U.S. Chemical Safety and Hazard Investigation Board report. Police officers and ambulance personnel responding to the emergency were some of the 154 people who required decontamination and treatment at the hospital for exposure to the toxic vapor.
Released on the second anniversary of the emergency, the CSB report concluded that better process design, engineering and hazard analysis would likely have prevented the runaway chemical reaction and toxic vapor release. The report also cited inadequate emergency planning by the facility, city and county as contributing to the severity of the event.
Mark Kaszniak, a CSB chemical incident investigator, said that the company involved did not understand the reactive chemistry of the product they were attempting to make.
"They had done a number of tests to improve product yield, and to use less expensive raw materials," Kaszniak said. "This was a new product for them.&rdquo
The accident occurred during the company's first attempt to make a production-scale batch of TAC, a chemical used in rubber manufacturing. CSB's report pointed specifically to published reports of two previous runaway reactions and subsequent fires at other companies that occurred during attempts to produce TAC. The technique and chemicals used by the Georgia company were similar to the two earlier, failed attempt. Luckily, Kaszniak said, the chemicals released in Georgia did not ignite. The company did not thoroughly review the literature to identify reactive hazards associated with these chemicals, nor did they do any chemical reactivity testing to determine if any hazards were present, he said.
Vapor released by the runaway reaction primarily consisted of allyl alcohol, an extremely flammable liquid with a flashpoint of 70 degrees F. It is toxic and poses a severe inhalation hazard, high skin absorption hazard and high ingestion hazard. It is corrosive to the eyes and a severe skin irritant. The vapor also contained highly flammable and toxic allyl chloride.
The small specialty chemical manufacturer arranged to make TAC for a New Jersey chemical company under a tolling arrangement. The Georgia company was selected because it had a 4,000-gallon, glass lined reactor and associated equipment needed for the chemical synthesis available. The New Jersey company also concluded that they had the technical resources capable of developing the conceptual procedure into a full-scale production process.
In rubber manufacturing, TAC is used as a modifier that makes products more pliable, Kaszniak said. In the chemical development process known as "scale up," small quantities of cyanuric chloride, allyl alcohol and a catalyst were first combined in the lab to confirm the basic TAC "recipe." Next they varied the quantities of each chemical and experimented with mixing rates and temperature controls to refine the process.
"What they were doing in the lab was checking to see if the chemistry worked," he said. "They were using an expired patent so they didn't have to worry about patent infringement. Obviously when you make a product you want as close to 100 percent yield as possible to reduce the amount of waste that is generated and make more profits." Once the laboratory-scale batch procedure was finalized, they increased the total quantity of each chemical at the same ratio as the laboratory batch up to a 30- and then 100-gallon reactor vessel. Finally, they went to the full, 4,000-gallon production batch size.
The company anticipated a significant exothermic or heat producing reaction, the report states. In the laboratory this was easily controlled by an incremental chemical addition sequence and additon of caustic soda to neutralize the mixture at each increment, Kaszniak said. As controlling the reaction temperature in the two small test reactors was easy, they thought that simply attaching a large chiller system to the production reactor would control the heat released by the full scale exothermic reaction.
What was not anticipated was an exothermic decomposition reaction. Controlling the temperature of the reacting chemical was significantly more difficult in the 40,000-gallon production reactor than in the 30-and 100-gallon laboratory test reactors. The heat removal capacity of a reactor equipped with an external jacket is directly proportional to the ratio of the jacketed surface area to the reactor volume.
"This surface-to-volume ratio decreases as the reactor volume increases, thus the ability to remove excess heat may be significantly less in a large production reactor compared to the bench-scale reactor," the report states. An uncontrolled runaway decomposition reaction results when the temperature of the reacting mass gets too high.
There was also a big material handling issue, Kasniak said. Two ingredients of the batch were liquids and one was a powder. The liquids could be easily pumped into the reactor, but the increased production batch size made loading the powder more difficult; many supersacks of powdered material would need to be loaded into the reactor. The powder, when dumped into the reactor through the reactor manway, generates dust to which the employees would be exposed.
"In their first production run they decided to add the entire quanity of the powder into the reactor first, followed by all the liquid ingredients, rather than in the incremental portions as had been done previously," he said. "They did not consider how this change might affect the process."
A short time after loading the catalyst and allyl alcohol, the operators noticed that the reactor temperature had increased from 32 degrees F to about 72 degrees F, presumably due to the addition of the warm allyl alcohol. Ten minutes later, the operators noted that the temperature had already climbed to 103 degrees F. The temperature continued to increase rapidly to 118 degrees F. With all the chemicals already in the reactor, the operators had no way to control the reaction. They could only stand by and watch the temperature climb. The last observed reactor temperature was 124 degrees F.
Unknown to the engineers and operators, the process rapidly approaching the temperature at which the exothermic decomposition reaction occurs, the report states. Rapidly increasing pressure in the reactor caused first the manway gasket to blow out followed shortly thereafter by the reactor's rupture disc. Dense, white vapor immediately began to spray out of the manway and the discharge pipe of the rupture disc, which vented directly to the atmosphere near ground level, rather than to a scrubber or catch tank.
The CSB report also cited deficiencies in the company's emergency planning. Use of allyl alcohol would have significantly increased the quantity of flammable liquid stored on site, the report states. Company management told the CSB that they assumed that the local fire department would provide all emergency response in the event of a significant release, even though their training only covered very small releases. However, the fire department told investigators that their discussions with management clearly explained that they were not qualified or equipped to respond to a toxic chemical release and that the company would have to make provisions for that.
Management provided a product safety bulletin to the fire department and informed them of their intent to handle allyl alcohol. They also agreed to notify the fire department after the allyl alcohol isotainer arrived on site, but before connecting it to the reactor. The fire department agreed to send a representative to become familiar with the placement of the isotainer and to discuss emergency response before the production run started.
"However, the company did not notify the fire department when the allyl alcohol arrived, so the fire department site visit did not occur," the report states.
The report goes on to say emergency response to the vapor release by city and county agencies was flawed in several key areas. A call informing the county 911 Emergency Management Center that a "chemical release involving allyl alcohol" was occurring reached the city fire department simply as a "hazmat spill inside the building." The dispatcher incorrectly told the firefighters that the material spilled was "alloy" alcohol. Responding firefighters unknowingly drove through the dense vapor cloud spreading east from the plant, the acrid odor immediately irritated their eyes and nostrils.
"Since they don't have a lot of chemical releases in that area, their notification procedures for handling them weren't very well developed," Kaszniak said. "The notification protocols in use were adequate, it's just that people did not have a lot of experience using them."
Emergency responders entered the affected neighborhoods to alert the residents to evacuate the area, the report states. None of the responding city police knew the identities of the chemicals being released or their hazards. They did not have appropriate safety protective equipment (e.g., respirators) needed to enter the neighborhoods, but there was no other available community evacuation method (such as a community evacuation siren, or reverse 911 dialing.) One officer was instructed to go door-to-door to alert residents because his vehicle PA system was not working. A request from one police officer for information about protective equipment went unanswered. An ambulance crew answering a 911 call in the area was also overcome, the EMT experienced a severe reaction.
After the toxic vapor forced the unprotected police officers to retreat, firefighters wearing breathing apparatus were called in to complete the evacuation.
At least 154 people had to be decontaminated and treated for chemical exposure at a local hospital. Thirteen police officers were also treated. Five people were hospitalized for toxic chemical exposure. One worker at the plant suffered minor chemical burns.
Firefighters set up unmanned water cannons and directed the flow on top of and around the reactor in an attempt to absorb the releasing vapor and cool the reactor. As the vapors being released were water soluble, the water spray was effective in knocking them down, but because the reactor was a glass-lined steel vessel, the water spray was not effective in cooling the reacting chemicals inside. The water spray was continued in spite of information that runoff was entering the storm-water drainage canal flowing into a nearby creek. The result was a fish kill extending almost seven miles downsteam of the facility.
The CSB report found that the company did not implement the EPA Risk Management Program (40 CFR 68) prior to receiving the allyl alcohol. This regulation required comprehensive engineering analyses of the process, emergency planning, a pre-startup safety review and coordination with the local community before they received the chemical on site.
The company also did not implement the OSHA Process Safety Management standard (29 CFR 1910.119) before using the allyl alcohol. Although they kept the allyl alcohol quantity in the reactor below the threshold limit, they incorrectly concluded the allyl alcohol in the connected isotainer was exempt.
"These are what OSHA and EPA call performance based standards and a lot of interpretation goes on with them," Kaszniak said. The company mistakenly interpreted the standards as not applying to them if they kept batch sizes below a certain amount.
According to Kaszniak, the potential for tragedy was far greater than the outcome. Neither the company nor the emergency responders had test equipment capable of measuring the toxic chemical concentrations in the air, yet the OSHA permissible exposure limit (PEL) for ally alcohol is only 2.0 parts per million. Furthermore, the vapors escaping from the reactor were well within their flammable range. Again, luckily, they did not ignite, Kaszniak said.
"Flammability and toxicity were a big concern," Kaszniak said. "When you have untrained and unprotected employees trying to do mitigation activities that are beyond the level of their hazmat training it's easy for somebody to do something wrong. The chemical release in Georgia resulted in 154 people going to the hospital for medical treatment. It could have been much worse."