Screen Shot 2017 05 15 at 10.52.08 AMAt about 14:35 on 9/29/12, an explosion and fire occurred in an intermediate tank that temporarily stored bottom liquid from the glacial acrylic acid rectifying column. The fire then spread to the nearby equipment and buildings such as acrylic acid tanks, toluene tank and fire engines.

Fatalities: 1 (firefighter)

Severely injured: 5 (2 firefighters and 3 employees)

Moderately injured: 13 (8 firefighters, 1 police officer and 4 employees)

Lightly injured: 18 (14 firefighters, 1 police officer and 3 employees) Total: 37 persons

Overview of Equipment Involved

The equipment involved was the acrylic acid intermediate tank V-3138, which had a nominal capacity of 70m3. It was an insulated (75mm thick) cone roof type storage tank. The tank was installed in November 1985. It is used as an intermediate tank for temporarily storing the withdrawal liquid from the rectifying column when, for example, the rectifying column stopped.

During normal operation, the bottom liquid is fed directly into the recovery column without passing through V-3138 and therefore the intermediate tank is kept stagnant.

There was a cooling water coil inside V-3138 that serves to prevent the freezing of acrylic acid and cooling the liquid that was fed into V-3138. The amount of liquid necessary to fully cover the top of the coil is 25m3.

Although acrylic acid is a flammable liquid, its vapor does not burn when it has an oxygen concentration of 8% by volume or less. Therefore, a mixed gas (referred to as “M-Gas”) consisting of 7% oxygen and 93% nitrogen by volume is fed into the tank for sealing purpose.

Liquid stored in V-3138 is circulated through pump P-3138C back into the same tank at two (2) locations:

  1. liquid level gauge nozzle set near the lower side of V-3138’s wall (referred to as “Recycle to Level Gauge”), and
  2. nozzle set at V-3138’s top (referred to as “Recycle to Top”).

Initially, when the V-3138 was installed in 1985, there was only the "Recycle to Top" line for pump circulation. In order to prevent the precipitates accumulation and results in incorrect reading at the level gauge, the "Recycle to Level Gauge" line was installed in 2009.

Sequence of Events

The committee has divided the sequence of events leading up to the accident into four stages and investigated the accident details and causes.

(1) Before started the storing operation in V-3138 (From September 21 to about 9:30 on September 25, 2012)

After the Total Power Shutdown Work was completed, cooling water was commissioned to pass through V-3138’s coil and similarly the M-Gas sealing was also commissioned. Circulation from pump P-3138C via Recycle to Level Gauge was commissioned too.

There was no abnormality observed in the operations of the crude acrylic acid production facility. The correct amount of inhibitors was also fed into the crude acrylic acid obtained from the purification process in accordance with the operation standards.

Operations of the rectifying column (T-6108) and the recovery column (T-6701) in the glacial acrylic acid production facility were started on September 21. The bottom liquid of T-6108 was fed directly into T-6701.

On September 24, another rectifying column (T-5108) was started. The bottom liquid of T-5108 was fed to the T-6701 via V-3138. During this operation, the liquid volume in V-3138 was maintained constantly at about 10m3.

The transfer piping of T-5108 bottom liquid was steam jacketed to prevent plugging due to precipitation. The temperature of the bottom liquid as it entered V-3138 was estimated at about 100°C, based on steam temperature and the length of the jacketed piping.

Inhibitors were fed into T-5108 in accordance with the operation standards.


(2) During storing operation in V-3138 (From about 9:30 on September 25 to about 14:00 on September 28, 2012)

At about 9:30 on September 25, the feed of bottom liquid from V-3138 to T-6701 was stopped and commenced to build up the liquid volume in V-3138. This was to prepare a capacity load up test in the recovery column which was scheduled at a later date.

There was no particular change in the crude acrylic acid fed to T-5108 or in the inhibitors.

The operation was running normally according to the operation standards.

The volume of liquid in V-3138 reached the planned 60m3 at about 14:00 on September 28, approximately 77 hours after the commencement of storage. During this period, the circulation of Recycle to Top was not commissioned. Figure 4-3 shows the V-3138 liquid level trend during storing operation.

The coil cooled the liquid in the bottom of V-3138 but was unable to cool the liquid above the top of the coil effectively. This had created an uneven temperature distribution in the vertical direction of V-3138 liquid. It is presumed that this relatively high temperature section has led to the reaction of forming acrylic acid dimer (diacrylic acid, referred to as “DAA”). With the heat of this reaction, V-3138 liquid temperature has increased steadily. However, V-3138 was NOT equipped with any thermometer and it was NOT possible to detect the temperature has increased.


(3) After storing operation in V-3138 (From about 14:00 to about 14:10 on September 28, 2012)

At about 14:00 on September 28, the liquid volume in V-3138 reached 60m3. Thereafter, T-5108 bottom liquid was switched back to feed T-6701 directly without passing through V-3138. Even then, the circulation of "Recycle to Top" was still not commissioned.

Consequently, the liquid above the top of the coil was not cooled and remained at a relatively high temperature. In order to simulate the temperature distribution in the V-3138 liquid, a 3-D model of V-3138 was developed.

The analysis showed that if "Recycle to Top" was not commissioned, the liquid just below surface remained at a relatively high temperature, average about 87°C immediately after reaching a volume of 60m³.

(4) From liquid holding in intermediate tank (V-3138) to explosion and fire (From about 14:10 on September 28 to about 14:35 on September 29, 2012)

It is presumed that the reaction of DAA formation accelerated in the liquid stored in V-3138 and the liquid temperature continued to increase from the heat of DAA formation.

In a laboratory experiment conducted under adiabatic condition, the reaction was confirmed to proceed until the composition has reached to approximately 40wt% of acrylic acid (AA), 60wt% of DAA and others. If the reaction is allowed to proceed under adiabatic conditions up till these compositions, the amount of reaction heat is able to increase the temperature of the V-3138 entire liquid volume by about 40°C.

It is also presumed that this continued rise in temperature has started the polymerization of acrylic acid and the temperature of V-3138 liquid rose even more rapidly by the heat of polymerization.

At about 13:20 on September 29, an operator discovered white smoke (acrylic acid vapor) coming out from V-3138 vent. It is estimated that the high temperature section of V-3138 liquid was about 160°C at that time. This temperature was estimated by confirming water vapor is visible as white smoke in the laboratory test and then by using Aspen Plus to confirm the vapor pressure with the composition of the V-3138 liquid.

The V-3138 liquid level gauge reading (pressure differential type) has shown increasing trend. It started to rise at about 13:20, reached the maximum liquid volume setting at about 13:30 and exceeded the gauge’s out-of-range limit (84.8m3) at about 13:40. This trend has shown the V-3138 internal pressure has increased. The increasing pressure is presumed to have occurred because the vent’s discharge capacity has exceeded by the increased amount of evaporated vapor resulted from accelerated polymerization.

The V-3138 pressure continued to rise thereafter and cracks appeared at the shell plate. It is estimated the pressure had built up to 0.24~0.29MPaG (42 psi) based on tank structural analysis at the time the cracks appeared. Independently, based on the adiabatic polymerization experiment data, the V-3138 internal pressure was also estimated to reach about 0.27MPaG (39 psi). Also, the V-3138 liquid was estimated to be about 230~240°C (464°F) in the high-temperature zone. The reaction data of the adiabatic reaction test is explained in Appendix 1 and the structural analysis results will be discussed in Section 4.2.1.

The V-3138 contents started to leak from the cracks and caused the V-3138 pressure dropped drastically.

At about 14:35 on September 29, the V-3138 liquid started to boil violently due to the drastic drop in pressure, while the liquid temperature still remained at high temperature. These resulted in Boiling Liquid Expanding Vapor Explosion (BLEVE) inside V-3138 and subsequently ruptured the V-3138.

The scattered contents ignited and caused the fire. Based on the distance of scattered debris, it is estimated that the tank internal pressure at the time of explosion was about 0.45~0.64MPaG (93 psi). The possible ignition sources were sparks generated from the impact of metals or from broken electrical wires during the explosion.

The explosion has also damaged the nearby tanks. Thus, the leaked acrylic acid and toluene from these tanks caused the fire to spread further. Table 4-1 summarized the major events from V-3138 liquid reached its boiling point till the explosion and fire occurred.


CLICK HERE for the full report

View 's profile on LinkedIn

Training Course: 

"Safety" for

Process Safety Managers 

September 18-22, 2017
Cincinnati, OH

CLICK HERE for more


 LinkedIn Group Button



safteng man copy


I am Speaking 2017

My Friend Jonathan Zimmerman

and I are presenting

720 - Process Safety Management

Best Practice - Safe Work Permit

Management System

June 22, 7:45am - 9:00am



an unpaid endorsement

kemkey logo

an unpaid endorsement


an unpaid endorsement



 organdonor website reputation