In the other chapters of this handbook we have been focusing on how to mitigate the consequences of gas explosions. If an accidental explosion has occurred, it is important to find out the causes of the accident and how to avoid similar accidents. The key factors of successful accident investigation will depend on the emergency plan and the expertise of the investigators.

The objectives of this chapter are to:

• List the main activities after an explosion accident and indicate how a systematic investigation into causes can be performed.
• Discuss what type of damage can be used to indicate the chain of events.
• Present FLACS as a tool for accident investigation.
  

14.1 Activities after an Accident

When an accident occurs, the first phase of actions consists of

i) rescue and help
ii) consequence reducing action, e.g. fire-fighting
iii) safe shut-down of the processes involved.

In case of a large accident, task force teams or investigation committees will be involved in the next phase to continue the operation of the facility and to analyse the accident. For minor event, this activity is often done internally by the responsible company. Figure 14.1 shows the main objectives of this type of work. As shown in the figure, the co-ordinating function is important since some of these activities will have different priorities or objectives. For instance repair of equipment and documentation of the damage may be contradictory, if not co-ordinated. By cleaning-up, damage indicators may be lost. After an accident a common reaction is to start cleaning up without documentation of the damage in mind. A person is therefore needed to chair the investigation committee and to co-ordinate all the activities.

Investigating the cause of an accident may require a lot of resources and time. Figure 14.2 shows some typical activities in a systematic investigation into the cause of an accident.

The objective of an explosion analysis would be to calculate backwards from observed damage and from eye-witness accounts: the operating data, the likely i) gas cloud, ii) ignition source and iii) release source. It is also important to point out what can be done to avoid similar accidents in the future.

Useful references for such work is:

i) Baker et al. (1983): Methods for calculating the effect of explosions from damage indicators.
ii) Lees (1980): General information on loss prevention in the process industries.
iii) Kuchta (1984): Data for combustible substances.

These sources of information should be available in accident investigations.

To perform the analysis will require highly qualified personnel. Baker et al. (1983) recommended that a team should include an explosion expert, a structural expert and plant operation experts. These experts shall be involved immediately after the accident. Most of the useful damage indicators will otherwise be lost.

Many companies are reluctant to publish accident reports. This attitude may be understood, but Kletz (1991) argues that there is a moral obligation to publish information which can prevent other accidents.

14.2 Damage Indicators

Documentation of the damage has to start immediately, and should be done by an explosion expert and a structural response expert(s). Take many photographs, both of the area view and the specific damages. Use a professional photographer and make systematic records of locations and directions of all the photos taken.

Organise a fragment map, showing the original position of the fragments and where they landed. Fragments can be a good indicator of where the explosion occurred and of the magnitude of the explosion. Figure 14.3 shows the trajectories of four parts of motor casings from an actual case. The casing of Motor A flew up to 15 m from its original position.

The fragments of the Motor A casing tell us that combustible gas has intruded into Motor A and that part of the explosion has been an explosion under Casing A. This explosion has most likely been the initial explosion and damage also tells us that the probable ignition location was under the casing or near Motor A.

Deflection of a ductile structure is another damage indicator. In Figure 14.3 the motor casing was deflected. For Motor B, the explosion load must be from the outside. Pipes, as shown in Figure 14.4, or panels that have deflected can be used to estimate the loads from the explosion (Baker et al., 1983).

Damaged ordinary window glass can be used to estimate blast wave, i.e. size of cloud and maximum pressure in an explosion area, as discussed in Chapter 7. Size of the window, thickness of glass, type of glass and percentage of windows broken should be recorded.

14.3 The FLACS Code

The FLACS code, as described in Chapter 12, is a useful tool in accident investigation. The FLACS code is a numerical code that predicts explosion pressure as function of time for different types of scenarios. FLACS takes into account the interaction between flame and geometry. In FLACS simulations, scenario parameters such as:

• size and fuel concentration in the combustible cloud
• type of fuel
• location of ignition point
  

can be defined.

The FLACS code has been used in investigations following several explosion accidents, e.g. "West Vanguard" (NOU 1986:16), Piper Alpha (Lord Cullen, 1990) and Beek (Salvesen and van Wingerden, 1993).

By performing FLACS simulations of different scenarios and comparing the estimated damage with observed damage, a good picture of what actually happened can be obtained.

14.4 Guidelines

• Know the company's emergency plan and procedure for accident investigation.
  
• Accident investigation has to be co-ordinated with other activities.
  
• Ask for help from experienced investigators (explosion and structural experts) as soon as possible.
  
• Use tools like FLACS or µFlacs.
  
• Make an open report of the findings from the investigation.