3.1 Analyses

After the accident, investigators analyzed the chemical residues in the blender and the remains of the blender itself. These analyses are discussed in detail in Appendix A. Appendix A also includes information on the injuries suffered by the victims of the explosion.

The chemical analysis showed the presence of sodium, potassium, and aluminum, consistent with the blender contents. Phenol and methylphenol compounds were also found in large amounts. These compounds were probably due to the insulating material remnants which were originally located in the annulus between the outer wall of the blender and the outer wall of the water-glycol jacket. These compounds could also be derivatives of the benzaldehyde that was added to the blender contents.

The analysis of the remains of the blender showed extensive erosion damage in parts of the interior and around several openings. Ejection of heated material from the inside of the blender is the likely cause of the erosion. The blender also showed extensive damage attributed to impact when the blender was propelled about 40 feet due to the explosion. The investigators looked in particular for possible damaged areas where water could have entered the blender. Metallurgical examination of the water-cooled graphite seal for the vacuum tube showed grooves that could have allowed water to leak into the blender.

Evaluation of injuries suffered by the victims indicate that the explosion at the Napp facility was a deflagration rather than a detonation. A deflagration releases energy at a lower rate, generates lower overpressures, and is less destructive than a detonation.

3.2 Significant Facts

Other facts considered by the accident investigation team in determining the causes of the accident are listed below.

Napp's Analysis of Hazards of Blending Operation

• The GPA was blended on one previous occasion by Napp in July 1992, using the same PK-125 blender. The manufacturer of the blender states that the blender is not designed to mix water reactive substances.
• Technic provided Napp with MSDSs for components of the GPA and an MSDS for the GPA.
• The MSDSs noted that aluminum and sodium hydrosulfite are water reactive.
• The MSDS for the GPA recommends "flooding with water" to control auto-ignition of the material.
• Napp's "new product review" procedure was conducted primarily based on information included in the MSDSs. Review was conducted individually (i.e., not by a team) by four managers of Napp to determine equipment suitability and health and safety concerns.
• Napp management was aware of the water reactivity of the materials in the blending operation.

Sources of Water and Heat

Literature review and a limited laboratory study of the hazards of the GPA mixture revealed that small quantities of water were capable of inducing a runaway reaction at relatively low temperatures. Further, the presence of aluminum powder in the mixture provided a substantial increase in the amount of heat released during decomposition (see Appendix B).

The following were possible sources of water and heat in or around the blender at the time of the accident, which were evaluated by the JCAIT as potential initiators of the event.

• The seal of the intensifier bar was water-cooled.
• Water was observed on internal surfaces of the blender prior to charging; a drying procedure was subsequently performed.
• Water was used to clean the blender.
• Water was noted in the liquid feed tank and feed line filter just before the feed line entered blender.
• The coolant in the blender jacket was a water/glycol mixture.
• Water was seen in the packing gland area/bearings of blender prior to the blending operation.
• Although the blender was blanketed with nitrogen, the ports were opened and atmospheric humidity could have been a source of water.
• The intensifier bar moving at a high speed through dry powders my have been a source of frictional heat.
• Particulate matter between bearing surfaces at the intensifier bar shaft in the packing gland area may have been a source of frictional heat.

Deviations from Napp's Expected Maintenance and Operating Procedures

• During a routine maintenance procedure, water was noted between a packing gland and bearings inside the PK-125 blender. Because this location was inside the vacuum section of the intensifier bar, no water should have been present. The water was dried and the gland re-packed.
• At 3:30 a.m. on April 20, two operators and a foreman found that the vacuum head area inside the PK-125 blender was "wet." Water (droplets) was observed on the intensifier bar shaft connection as well as on the inside walls of the blender. The moisture was removed and the blender dried.
• The quantities of the ingredients were recalculated because Napp lacked the proper amount of potassium carbonate.
• Operators noted that the level of dry ingredients was above the level of the intensifier bar and up to the middle of the vacuum head. According to Patterson-Kelley, the level should be no higher than the middle of the intensifier bar.
• At 10:00 a.m. on April 20, operators detected a vanilla-like odor in the liquid feed tank (used to contain liquids prior to injection into the blender).
• Water was observed in the liquid feed tank and its filter system. All parts of the injection system external to the blender were cleaned and dried. Because the powdered ingredients were already in the blender, the components of the injection system inside the blender could not be cleaned or dried without contaminating the charged material.
• At 2:30 p.m. on April 20, operators were unable to inject benzaldehyde, the sole liquid component of GPA, into the blender. For approximately 14 hours, operators attempted to clear the liquid feed line and re-inject the benzaldehyde, while intermittently operating the intensifier bar and rotating the blender. The intensifier bar was used far longer than specified in the Napp procedure, in order to properly blend material.
• At 7:00 p.m. on April 20, an operator smelled a "rotten egg" odor inside the PK-125 room. The operator assumed it was from open drums which originally contained the raw materials which had been charged into the blender.
• Two distinct liquids, assumed by Napp employees to be benzaldehyde and drops of water or IPA, were seen in the vacuum separator bowl. No analysis to determine the components of the liquid was performed.
• At 9:30 p.m. on April 20, a maintenance employee noted an unpleasant odor inside the PK-125 room.
• By 4:30 a.m. on April 21, bubbling was reportedly observed on the surface of the materials in the blender.

• At approximately 6:00 a.m. on April 21, workers were verbally notified to leave the building. Arriving day shift employees were directed to remain in the rear parking lot of the building.

Conditions at the Time of Employee Reentry to Unload Blender

• Bubbling (in an area approximately eight inches in diameter) and smoking were noted on the surface of the material inside the blender approximately three hours earlier.
• The blending operation is normally conducted at atmospheric pressure. However, an exhaust nipple was relieving pressure from the nitrogen blanket and off-gas products of the batch from within the blender. Blender could not be tumbled and cooled due to the open vent.
• A strong sulfur smell was noted in the blending area.
• With the exception of 12 employees which were directed to assist in the off-loading of the blender, the facility remained evacuated.

Lack of Notification of Local Community Immediately Preceding Accident

• Napp solely used verbal notifications to evacuate the building. Use of an internal alarm would have automatically notified local emergency responders.
• During the attempt to remove the contents of the blender, Napp did not charge fire hoses that were laid out and held by Napp employees in case of ignition. The fire water system included a device that would automatically notify local fire officials if a pre-set flow rate was met. Since the system was not utilized, no alarm was initiated.
• The Napp facility was located in a residential area, with homes and other businesses nearby. Napp did not provide any warning to local community of an on-going chemical emergency.

4.0 Causes of the Accident

4.1 Possible Causes of a Chemical Reaction

The reports of unusual odors in the blending room and building, bubbling and smoking noted on the surface of the material in the blender, and pressure buildup in the blender all suggest that the explosion and fire were triggered by an unwanted and uncontrolled chemical reaction occurring in the blender.

The GPA was successfully blended by Napp and other companies without incident on several occasions prior to this accident. Based on this history and an evaluation of the chemical components of the GPA, the JCAIT believes the raw materials by themselves will not normally react with each other in the absence of an outside initiator such as water and/or heat. The JCAIT eliminated outside sources, including sabotage and "Acts of God," as possible causes, as there was no evidence to support these possibilities.

The predominant reactions taking place most likely were:

The exothermic reaction of sodium hydrosulfite with water;

The exothermic reaction of aluminum powder with water;

The exothermic thermal decomposition of sodium hydrosulfite, which would have been initiated by heat from the exothermic reactions with water; and

The exothermic oxidation of hot aluminum powder, which would have been initiated when air contacted the blender contents.

The oxidation of hot aluminum powder could have resulted in a deflagration. The chemical reactions of sodium hydrosulfite, aluminum, and benzaldehyde are discussed in detail in Appendix B. Brief descriptions of other accidents in the past involving sodium hydrosulfite or involving aluminum powder are presented in Appendix C.

Reaction Initiation

Based upon the nature of the chemicals in the blender at the time of the accident, and the circumstances of the accident, the JCAIT investigated possible sources of water and heat as the initiator of the reaction. The following table identifies the possible sources of heat and water in the blender system. These sources are discussed in more detail following the table.

Based upon eyewitness testimony, the reaction seemed to be localized (as a hot spot) rather than a generalized reaction of the materials in the blender. The eight-inch wide bubbling area noted in the Description of the Accident also suggests that the hot spot was more near the center of the blended material rather than along the sides of the blender. The JCAIT therefore considered the various sources of water/heat, as discussed below, to determine which ones would likely have led to a reaction in the location observed.

Inadequate drying of blender before use. Water was used to clean the inside of the blender before it was used. If water remained in the blender, it would have reacted with the raw materials. However, before the raw materials were added, water found in blender was drained and fully dried. While the bubbling noted towards the middle of the blender reveals that the reactions did not take place at the walls of the blender (where any moisture would have been located prior to use), it is likely that any moisture in the blender before it was used would have been distributed throughout the batch.

Leak of coolant from blender jacket into the raw materials. A water/glycol mixture is used in the outer jacket of the blender to cool or heat the blender. A breach in the integrity of the blender wall could have allowed the water/glycol mixture to migrate from the jacket into the contents of the blender. The JCAIT conducted metallurgical analysis of the blender after the accident. This analysis revealed one crack near the off-loading port; however, the crack was considered to be from impact damage. (See Appendix A for additional details.)

Moisture from raw materials. Moisture present in any of the raw materials could have been sufficient to initiate a reaction. However, no signs of reactions (odors, heat) were detected by the operators charging the blender with raw materials. Also, Napp performed a quality assurance check on the raw materials, and no moisture was noted.

Moisture in nitrogen. Nitrogen was used throughout the blending process to inert the headspace in the blender and to prevent atmospheric moisture from reacting with the materials in the blender. Any moisture in the nitrogen would have been carried into the blender, and it would likely have caused reactions on the surface of the material in the blender. However, this nitrogen source included a filter to extract moisture from the supply. There was no evidence that suggests moisture was in the nitrogen.

Atmospheric humidity. Atmospheric humidity is known to react with sodium hydrosulfite and aluminum. Interviews revealed that a nitrogen blanket was placed on the contents of the blender when the blender was opened to the atmosphere, minimizing the effects of atmospheric moisture on the feed materials.

Water used to clean liquid feed line. Some information indicated that operators may have used water or steam to clear the liquid feed line in an effort to add the benzaldehyde. In this case, water would have been injected into the contents of the blender. However, the JCAIT confirmed that efforts to clean the liquid feed line after the initial attempt to inject benzaldehyde did not involve water or steam.

Water in liquid feed line or tank. Water was noted in the liquid feed tank and line up to the in-line filter near the line's entry into the blender, prior to the attempts to inject benzaldehyde. The maintenance employees were not able to clean and dry the portion of the injection system that is inside the blender because the raw materials were charged into the blender. Several drying/vacuum procedures were performed prior to charging, yet the liquid feed line within the blender was not checked to ensure that it was dry. Any water left in this portion of the line could have been deposited in the blender as the operators attempted to add benzaldehyde. However, given that operators were not able to inject the benzaldehyde into the blender it is unknown if any water in the feed line actually entered the blender.

Water-cooled seal failure. Analysis of the graphite water-cooled seal for the intensifier bar after the explosion revealed wear patterns that could have allowed water to leak through the seal. Had the seal leaked, it is likely that the water would have been deposited in the material being blended. However, analysis did not determine whether the seal failed during the blending of the GPA or during a previous use of the blender by Napp. Before the blending of the GPA, a Napp maintenance worker noticed water near a packing gland in the blender, which could be a sign of failure of the seal. However, even though it is unlikely- based on previous cleaning operations- it is possible that water used to clean the blender could also have been the source of the water near the packing gland.

Heat generated by shear of intensifier bar moving through dry metal powder. The heat generated by the friction between the intensifier bar moving through the dry metal powders may have been adequate to initiate or contribute to the thermal decomposition of the sodium hydrosulfite. Thermal decomposition of sodium hydrosulfite is exothermic; therefore, once the decomposition was initiated, it could have continued. The intensifier bar was only used while mixing the blender contents. Overfilling the blender may have reduced the efficiency of the blending operation; therefore, heat generated would not have been distributed throughout the raw materials within the blender.

Heat generated due to particulate matter between the bearing surfaces inside the packing gland for the intensifier bar shaft. The area inside the packing gland for the intensifier bar shaft contains bearing surfaces. Had any particulate matter been between those bearings , the particulate matter and the bearings could have become excessively hot due to friction. With this area in contact with the raw materials inside the blender, heat generated due to the friction would be transferred to the raw materials. If this heat transfer is sufficient it may have been enough to initiate the sodium hydrosulfite.

4.2 Most Likely Causes of Chemical Reaction

Based on the witness testimony, physical evidence, and analysis, the JCAIT has determined that the reaction and explosion of the sodium hydrosulfite and aluminum powder was most likely initiated by two mechanisms: water introduced that initiated the exothermic decomposition of sodium hydrosulfite; and heat which caused the sodium hydrosulfite to decompose. The JCAIT believes the most likely sources of water are the graphite water-cooled seal and the liquid feed line. The most likely source of heat was the shear of the intensifier bar moving at a high rate of speed through the dry powder.

Water from failed seal and liquid feed line. The JCAIT estimates that only a small amount of water was needed to initiate a reaction. If a large amount of water was injected into the material in the blender, the JCAIT believes a large hydrogen gas bubble would have been formed, causing a detonation with greater energy then was released in this accident. However, if a small amount of water was injected into the material in the blender, the reaction would begin a series of exothermic reactions over a longer period of time (see Appendix B). This theory is consistent with the findings of the accident investigation.

Through interviews and review of the blueprints of the blender, the JCAIT believes that if the water-cooled seal for the intensifier bar failed, it is likely that the water would have leaked into the material in the blender. Analysis of the seal after the accident revealed grooves that may have been deep enough to allow water to leak into the blender.

Additionally, the JCAIT can not rule out that any water in the liquid feed line may have been delivered into the blender when attempts were made to add benzaldehyde.

Heat generated by intensifier bar. The heat generated by the friction of the blades of the intensifier bar running through the dry chemicals in the blender may have been sufficient to initiate or contribute to the thermal decomposition of the sodium hydrosulfite. Thermal decomposition of sodium hydrosulfite is exothermic; therefore, it is possible that once the decomposition was initiated, sodium hydrosulfite in the blender continued to decompose, generating more heat. This reaction would have continued to create a "hot spot" in the material in the area below the intensifier bar. As the material was being removed by the Napp employees, the hot spot ignited, setting off the rest of the material in the blender.

4.3 Root Causes and Contributing Factors

Root causes are the underlying prime reasons, such as failure of particular management systems, that allow faulty design, inadequate training, or deficiencies in maintenance to exist. These, in turn, lead to unsafe acts or conditions which can result in an accident. Contributing factors are reasons that, by themselves, do not lead to the conditions that ultimately caused the event; however, these factors facilitated the occurrence of the event. The root causes and contributing factors of this event have broad application to a variety of situations and should be considered lessons for the chemical processing industries which operate similar processes, especially the tolling industry.

Immediately following the accident, members of the JCAIT collected and recorded the details on the event and the circumstances leading up to the event, interviewed witnesses, and collected, photographed (see Figures 4-24) and analyzed physical evidence and documentation. In the following months, the JCAIT conducted engineering analyses of this information using elements of Events and Causal Factors and Hazard-Barrier-Target techniques and professional judgement to determine the root causes and contributing factors, and to generate recommendations to prevent a recurrence. The JCAIT concludes that the root causes and contributing factors of this accident are:

?An inadequate hazards analysis was conducted and appropriate preventive actions were not taken.

Through Napp's accident report and interviews with Napp employees, the JCAIT learned that most of the pre-operation hazards assessment as part of the "New Product Review" was based upon the information presented in the MSDSs for the GPA and its ingredients. MSDSs can provide adequate chemical hazards information but not necessarily process hazards information. For example, the information presented in the MSDS for the GPA was for a typical package size (up to one 55 gallon drum), not for the quantity being blended (22 drums).

Under the New Product Review Procedure using the MSDSs, Napp noted that aluminum, sodium hydrosulfite, and GPA were water reactive. However, this procedure and the MSDSs did not reveal or address accident history, identify and account for all of the potential sources of water, ways to eliminate or control these sources (engineering safeguards or procedures), recognition ofwater contamination of the raw materials or GPA, the immediate steps necessary to stop or handle an unwanted reaction inside the blender, and the proper technology and design of the equipment necessary to safely and effectively blend water reactive substances. In addition, the procedure did not identify that heat could also have an adverse affect on the substances in the blender. Without this information, appropriate prevention actions were not taken.

Napp successfully conducted the blending operation previously (1992) using the same process, procedures, and equipment and was unaware of any major accidents involving this product or process. However, Napp did not appear to know about, or at least consider the consequences of, past accidents that have occurred involving sodium hydrosulfite or aluminum powder.

The lack of an adequate process hazards analysis led to a lack of knowledge or understanding that small amounts of water or heat could trigger a self-sustaining exothermic reaction at relatively low temperatures. This exothermic reaction lead to the catastrophic deflagration and fire. Napp removed sources of water in the blending room and took steps to ensure the blender and raw materials were kept dry. However, other sources of water were still present and Napp did not fully eliminate the possibility that water could contaminate the blending process and operation. Water was used to cool the mechanical seal in the intensifier bar and a water/glycol mixture was used in the heating/cooling jacket. Evidence also suggests that water may have been present in the liquid feed tank and piping system since they were not cleaned and dried prior to the startup of blending operations. This equipment was cleaned and dried during the blending operation. However, since the blender was already loaded, the liquid injection piping inside the blender could not be cleaned and dried. In addition, the intensifier bar offered a potential source of heat input from bearings and mixing shear (see equipment selection, below).

The JCAIT notes that OSHA allows the collection of the MSDSs to suffice for compliance with its information collection requirements for process safety information, provided the MSDSs contain information to the extent that enable the employer and employee involved in operating processes to identify and understand the hazards posed by these processes. Such specific information criteria include: reactivity data, thermal and chemical stability data, and hazardous effects of inadvertent mixing of different chemicals that could foreseeably occur. A review of MSDSs alone for highly hazardous processes in lieu of a formal process hazard analysis would not meet OSHA's requirements. Industry may not clearly understand this distinction which may have contributed to less than adequate hazards analyses since MSDSs were relied upon to conduct the company's new product review. As a result, thermal and chemical stability as well as inadvertent mixing of chemicals were not adequately addressed in the review process.

?Standard operating procedures and training were less than adequate.

Napp's standard operating procedures (SOPs) failed to adequately address emergency shutdown, including conditions under which emergency shutdown is required and the assignment of shutdown responsibility to qualified operators to ensure that emergency shutdown is executed in a safe and timely manner. Also, the SOPs did not address operating limits, including the consequences of deviations and steps required to correct deviations. Consequently, employees could not have been trained on these critical steps, hampering their ability to properly execute the blending process under the conditions occurring on April 20 and 21.

Napp did not recognize or understand the significance of the abnormal situation beginning on April 20 and culminating in the explosion and fire on April 21. According to Napp's procedures, the blending portion of the process should occur in less than an hour. However, employees attempted to correct a variety of deviations, including unusual odors, bubbling, pressure buildup, and difficulty adding the liquid portion, while the blending operation continued for many more hours than normal. The odors, bubbling on the surface of the contents of the blender, buildup of pressure and venting of gases from the blender all signaled that an undesired reaction had been initiated and was ongoing. There were no operating procedures to address the deviations observed, the corrective actions to be taken, or conditions under which an emergency shutdown should be triggered.

?The decision to re-enter the facility and off-load the blender was based on inadequate information.

The lack of a complete understanding of the chemical and process hazards led to a lack of knowledge of the significance of the conditions present at the time employees re-entered the facility to off-load the blender. Once the contents of the blender began reacting, the reactions were self-sustaining. As described above, the reaction had progressed to a point where employees were evacuated because of the sulfur smell coming from the blender. At the time the decision was made to unload the blender, Napp was aware of, and concerned for, the strong possibility of a fire. However, there is no evidence to suggest that Napp was aware that off-loading the blender may have exacerbated the reaction mechanisms by exposing the contents to air or that the contents could violently erupt and deflagrate. Any attempt to stop the reactions by smothering or providing a nitrogen blanket would have been ineffective (see the discussion of chemistry in Appendix B).

?The equipment selected for the GPA blending process was inappropriate.

Napp took steps to eliminate or control sources of water but elected to use blending equipment that incorporated sources of water in the design (water cooled seal, water/glycol jacket). Although regular maintenance of seals and cooling jackets helps to prevent failure and leakage, the possibility of a malfunction still exists, allowing water to contaminate the blender. The Aluminum Association, in their brochure Recommendations for Storage and Handling of Aluminum Powders and Paste, recommends "In mixing aluminum powder with other dry ingredients, frictional heat should be avoided. The best type of blender for a dry mixing operation is one that contains no moving parts, but rather effects a tumbling action such as a conical blender." The PK-125 blender, as previously noted, was a closed blender which, for the batch blended, contained a high speed intensifier bar for mixing. Shear from the movement of the intensifier bar through the dry powders may have been a source of frictional heat. In addition, particulate matter between bearing surfaces at the intensifier bar shaft in the packing gland area may have been a source of frictional heat.

?Communications between Napp and Technic were inadequate.

Inadequate communication between Napp and Technic also contributed to the lack of a complete understanding of the process hazards and their consequences. There is no standard or delineation of responsibilities in the toll manufacturing/blending industry which specifically assigns the responsibilities for input into hazard reviews at the toll manufacturer's facility.

Technic, Inc. is the patent holder of the GPA and is in the best position to know the hazards of its product. In this tolling operation, Technic contracted with Napp to perform the blending. Technic provided Napp with MSDSs for the raw materials and the finished product and a batch recipe ticket. As noted above, MSDSs provide product hazard information but generally do not provide process hazard assessment information. This investigation showed that Napp did not possess all the information necessary to make sound judgements regarding the responses to deviations from the procedures in the blending process. Napp is in the chemical processing industry; however, Napp did not have in-house expertise regarding the GPA, nor did it have experience in working with water reactive materials. Although some of its employees had previously used the ingredients of GPA, Napp blended the GPA only one other time prior to this event (the previous blend was performed in 1992).

It should be noted that the MSDSs for sodium hydrosulfite and aluminum powder also gave contradictory emergency response directions. For sodium hydrosulfite, copious amounts of water should be used, while the MSDS for aluminum clearly stated "avoid water," "reacts with water," "exothermic reaction with water . . . to generate hydrogen and heat." The MSDS for the GPA advised the use of "A water spray . . . to extinguish fire." The JCAIT investigators have located numerous scientific references, which were presumably readily available to Napp management personnel as well, indicating that powdered aluminum that is moistened or wetted becomes a very serious fire hazard. The only information that Napp used to determine the emergency response procedures for handling emergencies involving the GPA was the MSDSs. However, the MSDSs do not provide sufficient information to guide a response to an uncontrolled reaction or fire, given the significant quantity of material in the blender. The recommendation in the MSDS for "small fires" was to flood with water; however, a small fire was not defined, and the amount of water necessary to flood the fire was not specified. For a fire involving an agent that is reactive with water, the addition of an inappropriate amount of water as part of an emergency response can have tragic consequences.

?The training of fire brigade members and emergency responders was inadequate.

Eight of the twelve employees who were inside the building immediately preceding the explosion (during the unloading of the GPA) were also members of the Napp fire brigade and trained to handle fire hoses. Napp was concerned about the potential for fire by arranging for hoses and personnel to be ready. However, employee training records indicate that the employees standing by with fire hoses were not trained to deal with chemical fires or emergency response operations involving chemical fires. Consequently, the lack of training of the fire brigade members and emergency responders may have contributed to the consequences since the personnel present had no training to recognize, understand, and assist with a potentially significant emergency situation.

Link to Previous Section: Introduction, Table of Contents, Background and Description of the Accident.

Link to Next Section: Recommendations, Outcomes of OSHA/Napp Technologies Settlement, Appendices