Event

A powerful explosion and subsequent fire occurred at T2 Laboratories, Inc., a chemical manufacturer at Jacksonville, Florida, on December 19, 2007 [1]. The explosion killed four employees of T2 and injured four employees of T2 and 28 members of the public, who were working in surrounding businesses. The explosion occurred at the production of the 175th batch of methylcyclopentadienyl manganese tricarbonyl (MCMT) in a 2450‐gal (9.27‐m³) batch reactor. Following the incident, T2 has ceased its production operations.

Event Timeline

The process operator had an outside operator call the owners of T2 to report a cooling problem and requested they return to the site at 13.23 h. Upon their return, one of the two owners went to the control room to assist. The bursting disk broke, the reactor burst, and its contents exploded a few minutes later, at 13.33 h, killing the owner and the process operator and two outside operators who were exiting the reactor area.

Cause of the Accident

A runaway reaction occurred during the first step (the metalation step) of the MCMT process. A loss of sufficient cooling during the process likely resulted in a runaway reaction, leading to uncontrollable pressure and temperature rise in the reactor. The pressure rise in the reactor resulted in an explosion comparable to the explosion of 1400 pounds of TNT. The reactor contents, coming into contact with the atmosphere, ignited subsequently.

T2 Laboratories, Inc.

T2 was a small privately owned corporation that began operations in 1996. A chemical engineer and a chemist founded T2 as a solvent blending business and co‐owned it until the incident. From 1996 to 2001, T2 operated from a warehouse located in a mixed‐used industrial and residential area of downtown Jacksonville. T2 blended and sold printing‐industry solvents; it also blended premanufactured MCMT to specified concentrations for Advanced Fuel Development Technologies, Inc., a third‐party distributor.

In 2001, T2 leased a five‐acre site in a North Jacksonville industrial area and began constructing an MCMT plant. In January 2004, T2 began producing MCMT in a batch reactor. MCMT production was the primary business operation by December, 2007. On the day of the accident, T2 employed 12 people.

MCMT

MCMT is an organomanganese compound used as an octane‐increasing gasoline additive. Apart from blending, T2 manufactured MCMT themselves and sold it under the name Ecotane. Their process consisted of three steps that occurred sequentially within a single‐batch reactor.

The Metalation Step of the 175th Batch

The manufacturing procedure used by T2 started with the charging of sodium, methylcyclopentadiene dimer, and dimethyl diglycol ether (purity 95% by weight) to the reactor. According to a recipe given by the Chemical Safety Board (CSB), the percentages by weight of these three compounds were, respectively, 10.59, 43.55, and 42.58. The first step is called the metalation step, because sodium, a metal, is one of the two reactants. The other reactant is methylcyclopentadiene dimer. Dimethyl diglycol ether was used as a solvent in this process. Sodium metal arrived packed in mineral oil to prevent oxidation and limit moisture contact with the metal. Some mineral oil, 3.28% by weight, was transferred into the reactor with the sodium. The mineral oil does not participate in or interfere with the reaction. The three aforementioned percentages by weight and 3.28% by weight add up to 100.00% by weight.

At the 175th batch, the operator began heating the batch to melt the sodium and initiate the metalation reaction at about 11.00 h on December 19, 2007. The metalation reaction is a reaction between sodium and cyclopentadiene. Cyclopentadiene dimer was added to the reactor. The name cyclopentadiene dimer indicates that two cyclopentadiene molecules have combined. First, the cyclopentadiene dimer was converted into cyclopentadiene monomer. That means that each combination of two molecules is split up into two single molecules. Next, cyclopentadiene reacted with sodium. Hydrogen gas was a by‐product of this reaction. This gas was vented to the atmosphere. The pressure in the closed reactor was 53 psig (3.65 baro). Once sodium melted, at 210 °F (98.9 °C), the process operator likely started the agitator of the closed reactor. Heat from the reaction and the heating system (hot oil in a coil in the reactor) continued raising the reactor temperature. At a reaction temperature of about 300 °F (148.9 °C), the process operator likely turned off the heating system as specified in the manufacturing procedure, but heat from the reaction continued increasing the temperature of the reactor contents.

At a temperature of about 360 °F (182.2 °C), the process operator likely started cooling, as specified in the manufacturing procedure. A cooling jacket covered the lower three quarters of the reactor. City water was passed into the jacket at the bottom of the jacket and allowed to boil; steam from the boiling water vented to the atmosphere through an open pipe connected to the top of the jacket. Thus, the temperature of the water in the jacket rose to 100 °C. However, at the 175th batch, the temperature of the reactor contents continued to increase. The pressure increase caused by the temperature increase led to the explosion.

Chemistry of the Runaway Reaction

The CSB observed two exothermic reactions using the T2 recipe. The first reaction occurred at approximately 350 °F (176.6 °C) and was the desired reaction between sodium and MCPD. The second reaction was more energetic than the first one and occurred when the temperature exceeded 390 °F (198.8 °C). The second reaction was between sodium and the solvent dimethyl diglycol ether.

Remarks

A continuously running cooling system can be part of intrinsic process safeguarding as it does not need activation. The continuously running cooling system failed at T2 for unknown reasons. Extrinsic process safeguarding, i.e. a back‐up system for the continuously running cooling system, was not in place. That is, redundancy was not practiced. See Chapter 1 for a discussion of the concepts of intrinsic continuous process safeguarding and extrinsic process safeguarding.

The temperature margin between the desired reaction and the undesired reaction is 22.2 K. That margin is acceptable.

It is likely that T2 management was not aware of the occurrence of an undesired reaction at 198.8 °C.

The reactor was equipped with a 4 in. bursting disk that appeared to be inadequate.

Event

Two explosions occurred at the Moerdijk Site of Shell at Moerdijk in The Netherlands at approximately 22.48 h on June 3, 2014 [3]. Both explosions occurred in the hydrogenation section of the SMPO2 plant. The first explosion concerned the bursting of Reactor No. 2 and the second explosion concerned the bursting of the separating vessel of Reactor No. 1. The exploding contents of the reactor and the separating vessel took fire. Two persons working in the vicinity of the plant were wounded. The explosions caused extensive damage.

The direct cause of the accident has been unambiguously established. Ethyl benzene (EB) reacted unforeseen with a catalyst. The possibility of the chemical reaction remained unnoticed and could at one time develop into a runaway reaction causing a pressure increase and the bursting of Reactor No. 2 and the separating vessel of Reactor No. 1.

Shell SMPO Plants

The acronym SMPO stands for Styrene Monomer Propene Oxide. The name SMPO2 refers to the second plant at the Moerdijk Site. Both styrene and propene oxide are products of an SMPO plant. Styrene is used to manufacture polystyrene, an insulating material. Propene oxide is used to manufacture propene glycol, a material used in food, cosmetics, and pharmaceutical products. The raw materials are EB, propene, hydrogen, and oxygen. The intermediate product methyl phenyl ketone (MPK) is converted into methyl phenyl carbinol (MPC) in the hydrogenation unit of an SMPO plant. This conversion is accomplished by the reaction between liquid MPK and hydrogen gas. The reaction occurs in the presence of a catalyst. A catalyst accelerates the rate of a chemical reaction without itself being changed chemically. The pressure at this process is rather high, e.g. 23 bar in the SMPO2 plant. The temperature at this process is also elevated. MPC, the product of the hydrogenation unit, is subsequently converted into styrene.

Shell exploits five SMPO plants worldwide. There are two plants at the Moerdijk Site, two plants at Seraya in Singapore, and one plant at Ninghai in China. The plants were built between 1979 and 2005.

The Hydrogenation Plant of SMPO2

The hydrogenation plant of SMPO2 is indicated as Unit 4800. The main components of this plant are two reactors, two separating vessels, one circulation pump, and one heat exchanger. The runaway reaction occurred in Reactor No. 2. The heart of the hydrogenation plant is the set of two reactors. They are continuous trickle‐bed reactors. Figure 8.2 schematically depicts Reactor No. 2 and its separating vessel. A short description of how the plant is normally operating is given. MPK, the liquid to be converted into MPC, flows concurrently with hydrogen gas from the top to the bottom of the reactor. A distributor at the top of the reactor produces a shower out of the two‐phase flow. The reactor has been filled with catalyst particles, the pellets. The liquid trickles down the catalyst bed. Each pellet is surrounded by a liquid film. The chemical reaction between the liquid and the gas occurs at the surface of the catalyst. The diameter of Reactor No. 1 is 2.8 m and the diameter of Reactor No. 2 is 1.7 m. The height of the reactors is twice to thrice the diameter. The pellets are small cylinders having, typically, a diameter of 3.2 mm and a height of 3.4 mm. The two‐phase flow leaving each reactor enters a horizontal separating vessel. The liquid collects at the bottom of the vessel, while the gas leaves the vessel at the top.

The Timeline and the Process Conditions of the Accident

The MSPO2 plant had been shut down on May 25, 2014 for maintenance. The main purpose of the stop was the replacement of the pellets in the two reactors. The new pellets were in place on June 3, 2014, and it was planned to start up Unit 4800. A step preceding the normal operation is the reduction of the catalyst. The catalyst contains among other compounds copper oxide, copper chromite, and copper chromate. These and other compounds contain oxygen. The reduction step comprises chemical reactions between these and other compounds on the one hand and hydrogen on the other hand. Hydrogen reacts with oxygen bound in the catalyst. The reduction was carried out by passing EB and hydrogen gas concurrently from the top of the reactors to the bottom of the reactors. Hydrogen reacted with the catalyst while EB did not. EB served as a carrier for the heat of the reduction reaction, and it was recycled and cooled in an indirect heat exchanger. The reduction reaction was carried out at 130 °C. Warming the pellets was a step preceding the reduction of the catalyst. This was carried out by passing EB and nitrogen gas concurrently from the top to the bottom of the reactors. The accident occurred during this warming step. EB served as a heat carrier to warm the pellets to 130 °C. Recycled EB was warmed in an indirect heat exchanger. The circulation with EB was started at 18.20 h. The EB flows to the two reactors were independent from each other. 250 kg h⁻¹ of nitrogen gas passed through the two reactors in series and was subsequently vented. Reactor No. 1 was the first reactor entered by nitrogen. Warming of EB started at 20.15 h. The EB flow to Reactor No. 1 was stationary at 88 t h⁻¹ between 20.15 and 22.48 h, the time of the explosions. The temperature of Reactor No. 1 increased gradually in this period and the three thermocouples along the height of the reactor indicated 120 °C at 22.48 h. The EB flow to Reactor No. 2 varied strongly between 2 and 30 t h⁻¹ between 20.56 and 22.48 h. Compared to the warming of the pellets in Reactor No. 1, the warming of the pellets in Reactor No. 2 was delayed, e.g. the thermocouple at the bottom of the reactor indicated 60 °C at 22.48 h.

The Interruption of the Nitrogen Flow

The EB flow to both reactors had been started at 18.20 h. EB was collected in the separating vessels after having passed the reactors. The levels in the separating vessels varied strongly. The flow of nitrogen to the ventline was automatically closed at 22.15 h because of too high a level in the separating vessel of Reactor No. 2. The reason for this action is the necessity to avoid liquid being entrained into the ventline. Shortly after 22.15 h, the level in the separating vessel of Reactor No. 2 decreased. The operator could then have restored the nitrogen flow manually. However, the operator did not restore the nitrogen flow. Thus, as from 22.15 h, the nitrogen flow was reduced to zero. The system pressure rose to 7.8 bar, the nitrogen supply pressure.

The Cause of the Accident

Shell calculated that the EB flow to Reactor No. 2 should have been 16 t h⁻¹ and the nitrogen flow 1700 kg h⁻¹. It is likely that the following scenario has caused the accident. There were dry zones in the catalyst bed of Reactor No. 2 during the warming of the pellets because of the varying EB flow and the small nitrogen flow. In these dry zones, there was EB on the surface of the catalyst but no flow of EB. A reaction occurred between EB and the catalyst in those dry zones. The heat of that reaction could not be carried away and, subsequently, hot spots developed in the bed. The temperature in those hot spots rose to 180 °C or to an even higher temperature. Reactions can occur at that temperature between, e.g. EB and copper oxide. The latter reactions developed into a runaway reaction leading to the destruction of Reactor No. 2 and the separating vessel of Reactor No. 1. The bottom temperature of Reactor No. 2 was 60 °C at 22.48 h, the time of the explosion. It is thus improbable that the explosion started in the bottom of Reactor No. 2. Likely, the explosion started in the upper part of Reactor No. 2. The high pressure caused by the explosion could more easily reach the separator of Reactor No. 1 than the separator of Reactor No. 2. There was a line between the top of Reactor No. 2 and the separator of Reactor No. 1. The relatively cold bed of Reactor No. 2 was between the upper part of Reactor No. 2 and its separator.

History of the Catalyst

The first catalyst bought by Shell is the Cu‐1808T catalyst. This catalyst typically contains (% by weight):

It was supplied to the first plant, built in 1979 at the Moerdijk Site. According to the report of the Dutch Safety Board, it was established by means of experiments that the catalyst did not react with EB up to a temperature of 130 °C [4]. 130 °C is the temperature at which the catalyst is reduced. The first plant had a liquid‐full hydrogenation reactor. Further plants were equipped with trickle‐bed hydrogenation reactors. The catalyst Cu‐1808T appeared to be not fully satisfactory for trickle‐bed reactors. Moreover, the catalyst was expensive. Because of these two reasons, Shell tested three catalysts of different suppliers in the period 1999–2002. Regarding the comparison, the focus was on regular production. The G22‐2 catalyst of a new supplier was selected as an alternative for the Cu‐1808T catalyst. The catalyst G22‐2 is a mixture of copper oxide, copper chromite, barium chromate, and silicon dioxide. It typically contains (% by weight):

Shell labeled the new catalyst as a “drop‐in” catalyst, i.e. neither plant equipment nor manufacturing procedures needed modification. The possibility of a chemical reaction between the G22‐2 catalyst and EB was not checked.

Shell Experiments After the Accident

The chemical reactions between catalyst and EB were studied experimentally. Three different catalyst samples were warmed with EB and the evolution of carbon dioxide was checked. The evolution of carbon dioxide was considered a proof that, yes, chemical reactions between catalyst and EB had occurred. Reactions between catalyst Cu‐1808T and EB started at 150 °C. Reactions between two versions of catalyst G22‐2 and EB started at 100 °C. The first of these two versions concerned deliveries to Shell before and including 2010 and the second regarded deliveries to Shell after 2010. There was no significant difference between the latter two versions in the test.

Shell Explanation of the Experimental Results

The different behavior of catalysts Cu‐1808T and G22‐2 is caused by the difference in Cr(VI) content. Cr(VI) stands for six‐valent chromium. Oxygen bound to Cr(VI) is more reactive vis‐à‐vis EB than oxygen bound to trivalent chromium. Cu‐1808T contains 0.2–0.3% by weight of Cr(VI) and the two versions of G22‐2 contain between 2.4% and 5.1% by weight (13 measurements of samples between 2004 and 2015). G22‐2 tested between 1999 and 2002 contained 0.1–0.2% by weight of Cr(VI). The G22‐2 deliveries to Shell started, at some point in time between 2002 and 2004, to contain substantially more Cr(VI). The Cr(VI) content of the catalyst G22‐2 was not part of the specification. The supplier indicated the Cr(VI) content on the MSDS (Material Safety and Data Sheet). The MSDS was sent to Shell; however, the fact that the Cr(VI) content was higher than in 1999–2002 was not explicitly communicated to Shell by the supplier. The reason is that the Cr(VI) content was not part of the specification. Shell did not notice the higher Cr(VI) content.

Modifications Implemented by Shell After the Accident

The step of warming the pellets is now carried out by passing warm nitrogen gas through the catalyst beds. EB is no longer used to warm the catalyst.

The step of reducing the catalyst is now carried out by passing a mixture of hydrogen gas and nitrogen gas through the catalyst beds. EB is no longer used at the reduction step.

Remarks

According to the report of the Dutch Safety Board, Shell had concluded that EB was inert vis‐à‐vis the first catalyst, i.e. Cu‐1808T, at temperatures up to 130 °C in 1977 [4]. EB did not react with Cu‐1808T up to 130 °C. However, for such a conclusion, it would have been necessary to check the possibility of such a reaction at 150 °C. Such testing provides a safety margin of 20 K.

The catalyst manufacturer advised Shell in writing in 2010 and 2013 to reduce the catalyst in the gaseous phase. The catalyst manufacturer did not explicitly exclude other reduction procedures. Up to the accident in 2014, Shell reduced the catalyst in SMPO2 with a two‐phase flow, i.e. a concurrent flow of hydrogen and EB from the top of the trickle‐bed reactor to its bottom.

Working with a two‐phase flow at both warming and reduction of the catalyst proceeds faster than working in the gaseous phase. However, the catalyst is replaced once in 3 or 4 years. Saving time here is relatively insignificant.

The accident happened during the warming step of the catalyst preceding the reduction of the catalyst. The catalyst is fit‐for‐use after the reduction step. The thought may come up whether it would be wise to ask the catalyst manufacturer to reduce the catalyst and to deliver a fit‐for‐use product. The reasoning here is that the catalyst manufacturer is knowledgeable concerning the manufacture of the catalyst, whereas the catalyst user is knowledgeable regarding the application. An aspect here is that the reduced catalyst reacts with oxygen in air. Thus, a reduced catalyst would have to be transported with precautions from the manufacturer to the user. The choice has therefore been made to reduce the catalyst in situ [5].

Event

An accident occurred at the salt furnace of the second melamine plant (Melaf‐2) of DSM Melamine Europe (DME) at Geleen in The Netherlands on April 1, 2003 [6]. A mixture of natural gas and air exploded in the furnace. The explosion lifted the roof of the vertical cylindrical furnace. Three employees of DME were standing on the roof, fell into the hot furnace, and lost their lives. Two further employees were slightly injured by the pressure wave. The furnace was damaged.

Stamicarbon Melamine Plants

Melamine (2,4,6‐triamino‐1,3,5‐triazine) is produced from urea. It is used in the fabrication of melamine‐formaldehyde resins for laminating and adhesive applications. Melamine is also used as cross‐linker in heat‐cured and high‐solid paint systems.

There are several processes for the manufacture of melamine. DME uses the Stamicarbon process, which will be described shortly. The conversion of urea into melamine is carried out in the gas phase in a well‐mixed continuous tank reactor at 7 bar and 400 °C with the aid of a silica‐alumina catalyst. Ammonia and carbon dioxide are by‐products of the chemical reaction. Urea is added to the lower part of the reactor as a melt having a temperature of 135 °C. The silica‐alumina catalyst is fluidized in the reactor by gaseous ammonia having a temperature of 150 °C. Melamine is recovered from the reactor outlet gas by water quench and crystallization. The chemical reaction is endothermic, meaning that heat must be added to the reactor. Molten salt is circulated through heating coils in the reactor and provides the reaction heat. The molten salt thereby cools down and is reheated indirectly in a furnace by the combustion of gas. The accident happened at that furnace.

Melamine has a melting point of 350 °C. The material is, after the reaction, present in the plant as a powder. Hence, the plant parts after the reactor suffer from incrustations. These incrustations are removed by steam approximately once per fortnight. The cleaning procedure lasts approximately 12 h. Urea is not fed to the reactor when the reactor is being cleaned. Gaseous ammonia flows through the reactor. The reactor is kept hot during the cleaning step by circulating molten salt through the coil in the reactor.

The accident happened while the plant was being cleaned.

The Furnace of Melaf‐2

See Figure 8.3. The furnace is a vertical cylindrical vessel having a height of 14 m and a diameter of 5 m. Natural gas or fuel‐gas enter at the roof of the furnace and are burnt with air, which is also entering at the roof of the furnace. There is a coil in the furnace, which is close to the wall and through which the molten salt circulates. The salt is heated by the gas flame while flowing through the coil and is then returned to the reactor where it provides the heat needed for the chemical reaction indirectly. The furnace was started up for the first time in 1999.

The salt circulating through the system was a mixture of potassium nitrate (55%) and sodium nitrite (45%). The salt had, at the time of explosion, a temperature of approximately 350 °C, the melting point of melamine.

The furnace load is low during the cleaning step. The cleaning step also provides the possibility to maintain the plant. Two maintenance activities were planned concerning the furnace. Two out of three filters should be cleaned and a leakage in the furnace roof should be repaired.

The two control valves 2051 and 2052 (see Figure 8.3) have the same function. This design had been chosen because of process safety reasons. If one control valve fails to function and stays in the open position, the other control valve is still active and performs the process control task. We see an example of redundancy.

Description of the Accident

The line related to natural gas should be considered only. The line related to fuel‐gas did not play a role in the accident. The accident is related to the cleaning of the filters. The upper two filters in Figure 8.3 had to be cleaned. The cleaning procedure begins with the closing of the valve marked BL (Battery Limit). Natural gas cannot flow to the furnace anymore. First, the procedure for the left filter will be treated. By opening and closing valves, the gas in the line can be displaced by admitting nitrogen and leading the gas to a safe location. This design is called a design with block and bleeder. It is then possible to remove the filter, clean it, and reinstall it. Next, the procedure for the right filter will be discussed. The design for the right filter is not a design with block and bleeder. The control valves 2051 and 2052 were manually locked in the open position to enable the displacement of the gas with nitrogen into the furnace. It is then possible to remove the filter, clean it, and reinstall it. It was allowed to lock the control valves 2051 and 2052 in the open position for the displacement of the gas with nitrogen.

The line was filled with nitrogen after the reinstallment of the two filters. The nitrogen was removed by admitting natural gas through the manually controlled valve marked BL. Subsequently, the gaseous mixture was passed on into the furnace. The valves 2051 and 2052 were, during this admission of natural gas, still locked in the open position. Fan K2001 was started manually and locally shortly after the removal of nitrogen by the admission of natural gas. The explosion occurred a few seconds after the start of the fan. It was not allowed to lock the two control valves in the open position while admitting natural gas to the furnace. Instead, the two valves should, during the restart of the furnace, no longer have been controlled manually but by the process control system. The reason that the operators kept the two valves under manual control is that it was, when the two valves were controlled by the process control system, difficult to ignite the gas to restart the furnace because it was initially mixed with nitrogen. The ignition had to be activated several times and each time the ignition failed the furnace had to be flushed with air to displace inflammable material from the furnace. This was considered a roundabout and time‐consuming way of starting the furnace.

The roof of the furnace having a diameter of 3.80 m was lifted by the explosion and pushed against the roof of the structure in which the furnace was installed. It fell back on the furnace and then made an angle of approximately 45° with the horizontal.

A Closer Look at the Explosion

About 26 nm³ of natural gas had been admitted to the furnace. This corresponds with 59 m³ at 345 °C and atmospheric pressure. The furnace volume was approximately 250 m³. The natural gas concentration in the furnace would hence have been approximately 24% by volume if the gaseous mixture in the furnace would have been well mixed. The upper and lower explosion limits (UEL and LEL) are, respectively, 19.9% and 3.7% by volume at 345 °C and atmospheric pressure. This means that the natural gas content of the gas in the furnace would be somewhat greater than the UEL if the furnace contents would have been well mixed. It also means that the natural gas content of the gas in parts of the furnace could have been in between the UEL and the LEL if the gaseous mixture in the furnace would not have been well mixed.

Fan K2001 was started shortly after the natural gas entered the furnace. The specific mass of hot air in the furnace is 0.44 times the specific mass of air at ambient temperature. The specific mass of natural gas at ambient temperature is 0.55 times the specific mass of air at ambient temperature. Thus, the relatively cold natural gas collected at the bottom of the furnace. An oxygen content of 0% in the flue gas duct has actually been measured, proving that air in the flue gas duct was replaced by natural gas. The natural gas bubble at the bottom of the furnace was heated by the coils in the furnace and acquired buoyancy. The natural gas/air mixture moved upward and collected against the hot roof. Air was mixed with this natural gas/air mixture in a short time when fan K2001 was started. This caused a natural gas concentration in the gaseous mixture or part of the gaseous mixture between the UEL and LEL. It has been estimated that the temperature of the ceramic burner tip was between 800 and 1000 °C. Ignition at the burner tip, possibly in combination with self‐ignition in the vicinity of the burner tip, is considered the most probable mechanism.

It has been estimated that the maximum pressure in the furnace as a result of the explosion was between 0.6 and 1.5 bar gauge.

The Role of Fan K2001

Fan K2001 was activated shortly after the nitrogen in the natural gas line had been displaced by natural gas. The explosion occurred a few seconds after fan K2001 had been activated. Normally, the fan was kept running when the filters were being cleaned. The fan did not run during the cleaning of the filters on April 1, 2003, because, simultaneously with the cleaning of the filters, the roof was repaired. The starting of the fan will have enhanced the mixing of the gas in the furnace.

Background Information Concerning the Explosion

Three gas filters can be distinguished in Figure 8.3. It was envisaged in the design stage that the three filters would rarely have to be cleaned. Because of the use of fuel‐gas, two out of three filters had to be cleaned more often than expected. The filters had to be cleaned every 2 months. Due to this circumstance, the prescribed procedure for start‐up had not been followed on April 1, 2003.

Modifications Implemented by DSM After the Accident

It was decided to stop using fuel‐gas in the furnace of Melaf‐2. This considerably reduced the cleaning frequency of the filters. The line‐up of the filter after the mixer was changed. A design of block and bleeder was implemented. Nitrogen in the line can now be displaced to a safe location. A parallel spare filter for the filter after the mixer was installed. Finally, a physical coupling between valve BL and the process control system was made. This physical coupling makes it impossible to lock valves 2051 and 2052 in the open position when gas admission is possible. Several extrinsic safety measures were implemented additionally. For example, the gas supply is closed when fan K2001 is not running.

Remarks

The heart of the matter in this case is that instructions were not followed. That made it possible that natural gas entered the hot furnace while the furnace was not operating. A relatively unsatisfactory design seduced the staff and the operators to not adhere to the procedures. The fact that fan K2001 was not running at the time of the admission of the gas made it worse.

The roof of the furnace could have been a no‐go area when the furnace is functioning. Functioning includes start‐up.

Event

An explosion occurred at the Dow Company's chemical plant at King's Lynn, Norfolk, Great Britain, at approximately 17.10 h on June 27, 1976 [7]. The explosion caused the death of one Dow employee and extensive damage to the plant and buildings on the site.

The explosion concerned approximately 1300 kg of Zoalene, a poultry feed additive, which had been left inside a closed contact dryer for a period of 27 h after the drying operation had been completed. According to the report [7], the explosion was a detonation. The Zoalene had a temperature between 120 °C and 130 °C at the end of the drying operation. Under these circumstances, the product began to decompose with the evolution of heat. The evolution of heat per unit of time exceeded the heat loss to the surroundings per unit of time. This caused a temperature increase of the product. The decomposition rate increased due to the temperature increase. The runaway reaction ultimately led to detonation. The energy released in this event was approximately equivalent to the energy released by 200–300 lb (90.8–136.2 kg) of TNT (trinitrotoluene) on detonation.

The Product

The chemical name of Zoalene is 3,5‐dinitro‐o‐toluamide, an aromatic compound. The material is used to prevent coccidiosis infections of poultry. The purity of the material is minimum 98% by weight. The material is very slightly soluble in water.

The Equipment

The product had been dried batchwise in a contact dryer. The moisture removed from the material by the drying operation was water. The dryer was a double‐coned glass‐lined steel dryer equipped with a jacket. The dryer was located at the ground floor of a building. During the drying operation, steam was supplied to the jacket, whereas vacuum was maintained in the dryer. Evaporated water could be condensed overhead and collected in a catchpot. The dryer rotated about its horizontal axis while drying was taking place. The rotational speed could be varied between 1 and 11 rpm. The dryer was charged manually and discharged by means of a conveyor. The diameter of the dryer was 1.8 m and the height was 2.34 m. See Figure 8.4.

The Process

The specific Zoalene drying operation was the tenth drying operation of a campaign to rework off‐spec material. Dow decided to redry between 75 and 92 metric tons of this material. The purity was 96–98% by weight and it could be brought on‐spec by redrying. Wall incrustations formed during the drying operations. To remove incrustations from the walls of the dryer, water had been added. The dryer was half filled with water. The dryer was then rotated and heated until the sides were clean. Such a cleaning operation was carried out at one out of approximately three drying operations. More Zoalene was added to make up a batch of approximately 1300 kg of the product. The drying started at 15.00 h on June 25. The dryer was shut down at 14.00 h on June 26. The steam to the jacket was turned off and the vacuum released. The dryer was left with the charging opening lid clamped loosely in place. The explosion occurred at approximately 17.00 h on June 27, which means that the material had resided in the dryer after the drying operation for 27 h. At other drying operations, the batch was cooled by passing cold water through the jacket for about 40 min. The batch was normally cooled in order to permit immediate handling. However, in this case, cooling did not take place.

The Process Conditions

Steam having a pressure of about 40 psig (2.7 barg) was admitted to the jacket, resulting in a jacket temperature of 130–140 °C. A vacuum of about 25 in. Hg (97.7 mbar absolute pressure) was applied to the dryer. The water boiling point at this pressure is 45.5 °C. The rotational speed of the dryer was increased stepwise until it reached, after 4 h, 10 rpm. The water flow to the catchpot was an indication of the progress of drying. The endpoint was reached when the water flow came to a halt.

Product Safety Characteristics

After the accident, Dow established that typical production samples of Zoalene would self‐heat to decomposition if held under near adiabatic conditions at 120–125 °C for 24 h.

The Royal Armaments Research and Development Establishment (RARDE) at Woolwich, Great Britain, established the explosive potential associated with Zoalene.

Remarks

Explosive properties of an aromatic compound having two nitro groups in the molecule could have been surmised.

The material should not have been left for 27 h in a closed process vessel at an elevated temperature.

As a general rule, reworking off‐spec material entails more risks than regular production.