Observations on the Chemical Safety Board report on the Millard Accident

The analysis and recommendations expressed here have been developed by the IIAR Safety Committee working with IIAR staff


In August of 2010, an accident occurred at the Millard Refrigerated Serviced facility in Theodore, Alabama. Subsequent to the accident, the United States Chemical Safety and Hazard Investigation Board (CSB) investigated the accident and later issued a Safety Bulletin “Key Lessons for Preventing Hydraulic Shock in Industrial Refrigeration Systems”, published in January 2015. A summary of the accident and causes can be found in reviewing the CSB safety bulletin No. 2010-13-A-AL. As a result of the analysis, the CSB safety bulletin recommends certain design and operation practices intended to prevent hydraulic shock, and lessen the consequences of a release should one occur.

While the IIAR does not question the findings of the investigation, the IIAR believes it would be helpful to explore some of the statements and recommendations within the safety bulletin. This response was assembled with the input of numerous industry professionals with a wide range of experience in industrial refrigeration installations, operations, maintenance, and troubleshooting. This document addresses each of topics of interest and is intended to provide the basis for meeting the intent of the CSB recommendations while using alternate methods.

While better engineering controls (i.e. the design of the system and its controls) might have prevented or limited the consequences of the release, IIAR believes that operator training, operational discipline regarding following established procedures, and management of change of existing engineering controls, are at least equally if not more important. The lessons learned from the incident should be universally considered, but IIAR cautions that there are many approaches to safe design that can be considered for refrigeration applications.


In section 3,” Hydraulic Shock”, the CSB bulletin states that “…hydraulic shock events are typically the condensation-induced type and frequently occur in low-temperature ammonia systems.” IIAR respectfully suggests that while hydraulic shock events have not been totally eliminated from the industry, their frequency has declined. Understanding of the causes and prevention of hydraulic shock has improved over the past three decades, and changes have been implemented by the industry. IIAR agrees that the industry should continue to disseminate knowledge and best practices regarding this issue. It is important that owners, operators, contractors, and other stakeholders be knowledgeable about the symptoms of hydraulic shock, and that investigation and corrective action be initiated when such symptoms are observed.


Section 6, paragraph 1 of the CSB bulletin (on lessons learned) states that grouping multiple evaporators to a single set of control valves should be avoided, and is especially important for large capacity evaporators more than 20 tons. IIAR respectfully disagrees that it is unsafe to group multiple evaporators to a single control group if the control valve group and the operating controls are appropriately designed. An appropriately designed control group will safeguard against hydraulic shock events regardless of the volume of the heat exchangers connected to it. The value of 20 tons may have had validity years ago when systems tended to be simpler and smaller. Today, some individual coils could provide much more capacity than 20 tons and conversely, several evaporators grouped together might provide less capacity than 20 tons. Further, the rating of an evaporator is dependent upon many factors including room temperature, suction temperature, air flow, surface area, coil internal volume, moisture load and other factors. While limiting control groups to a single evaporator might seem to provide an abundance of caution, and may enhance the operation or controllability, the design of the control group is the crucial factor in preventing hydraulic shock events resulting from defrost operations. Prudence does dictate that the larger the coil volume on a single control valve group, the greater the importance of ensuring that all necessary steps are taken to prevent hydraulic shock, and the greater the importance of taking corrective action if any symptoms of hydraulic shock are observed.

It is notable that the CSB bulletin does not discuss the use of slow-opening valves. The use of slow-opening gaspowered suction valves has been widespread for many years. These valves incorporate the use of a compensating spring and a pilot gas orifice that slowly balances the pressure on either side of the valve when pilot gas pressure is removed. The valve will not open until the pressure is equalized. A loss of power during a defrost cycle would initiate the automatic equalization of the coil pressure to suction pressure. This is essentially a passive way to ensure that there will not be a sudden introduction of hot gas from a defrosting coil introduced into the system’s suction piping.

Motorized control valves have become increasingly popular to smoothly control the flow of refrigerant. These valves can be programmed to open gradually, and can be made able to slowly close themselves in the event of a power outage. While the use of these valves does not necessarily guarantee that hydraulic shock will be prevented, they do provide a means of protection that are not mentioned in the safety bulletin. Using these types of valves greatly reduce the risk of hydraulic shock, and can provide safety measures that would enable a variety of safe designs.

The CSB safety bulletin makes several suggestions in paragraphs 2 through 4 that presume controls systems are computer or PLC based. The IIAR agrees that if these suggestions can be implemented, they would enhance safety and reduce the risk of hydraulic shock. But there are many systems that use electro-mechanical control for some or all a system’s functions that will not permit the programming features suggested. Electro-mechanical systems can and should be designed to achieve similar goals.

There are also several other methods of defrost risk mitigation:

Low charge (DX) evaporators. These pump out significantly quicker than liquid overfeed coils and the probability of residual cold liquid remaining in the evaporator coils following conclusion of the pump-out cycle is less than it is for liquid overfeed or gravity flooded evaporators.

Hot gas defrost by means of a defrost medium that is not the primary refrigerant. These types of defrost methods use a separate defrost circuit interlaced with the primary refrigerant circuit. The primary refrigerant suction line exiting the evaporator remains open during defrost thereby greatly reducing the probability of liquid hammer of any form.

Automatic ambient air defrost as practiced in alcove and penthouse evaporators in warm climates.

The safety bulletin correctly implies that an ammonia release should be isolated promptly. The bulletin goes on to say that if a release cannot be promptly isolated, then the emergency shut-down switch should be activated. IIAR agrees this should be the default response for most facilities. The IIAR respectfully suggests that while this may be appropriate for most facilities, there are some facilities that have detailed planning, personnel, and equipment in place to facilitate making this decision in real time, as conditions and the situation dictate. Compressors create desirable low pressures within the low-side of refrigeration systems. Continuing to operate a compressor if a release is discovered may help to reduce the amount of refrigerant released. Releases on the high-pressure system may benefit from shutting off compressors. Many facilities develop emergency plans and train their personnel on appropriate implementation of the plans. This might include release mitigation that involves isolation without shutting down the compressors or other valves that automatically close upon activation of the emergency shut-down switch. Facilities which elect to allow operators to make decisions in real time as to whether to shut down the system should clearly state in their emergency plans which scenarios require immediate shutdown if prompt isolation is not possible, and which scenarios allow an operator decision to continue to run while other actions are being taken to terminate the release. If in doubt, the best option is to shut down the system.


In conclusion, the CSB report and this IIAR review should be considered useful references to designers, owners, contractors, and other stakeholders. This review offers some alternate methods that IIAR believes meet the intent of recommendations in CSB Safety Bulletin No. 2010-13-A-AL. The analysis draws on a history of successful practices and technologies that are intended to reduce the risk of hydraulic shock and release mitigation, while maintaining the ability to design and use economical arrangements to safely control refrigerant flow and system control. As with any design, maintenance routine, or emergency plan, the methodologies should be thoroughly reviewed by the designers, owners, and operators of the systems. Training on system operation and function, as well as on the facility emergency plan, should be incorporated at all levels of management and staffing. This is likely the most important lesson to be derived from the observations of the Millard incident.