Comparison of Classical and "Cause Consequence Diagrams" Recursive Operability Analysis: the T2 Laboratories Accident

On December 19, 2007, a powerful explosion and fire killed four employees and destroyed T2 Laboratories, Inc., a chemical manufacturer in Jacksonville, Florida. In the accident 32 people were injured, including 4 employees and 28 members of the public who were working in the surroundings. Debris was found up to one mile away, and the explosion damaged buildings within one quarter mile of the facility. After the accident, the question which arose was: could an even simplified risk analysis prevent such a tragedy? It is widely accepted that performing a detailed Quantitative Risk Analysis (QRA) is a complex and time consuming task because of all the steps which must be carried out: 1) hazards identification; 2) frequency estimation; 3) accident consequence evaluation; 4) individual and societal risk calculation. Specifically, Hazard Identification (HI) and Frequency Estimation (FE) represent two fundamental activities since: 1) not identified hazards can remain hidden until the occurrence of the related accidents; 2) the probabilistically quantification of the hazardous plant states frequencies, e.g. through Fault Tree Analysis (FTA), helps to support decisions making on risk reduction. Particularly, since generating FTs is a time consuming task, the Recursive Operability Analysis (ROA) has been ideated. ROA, both in its classical and revised version (called Recursive Operability Analysis - Cause Consequence Diagrams, ROA-CCD), is based on a procedure which allows collecting plant perturbations data in a structured way. The aim of this work is to apply both the classical and the new ROA-CCD analysis on the T2 Laboratories chemical plant (with particular reference to the reactor node) in order to identify all the possible top events and states of plant unavailability. In this way, it has been evidenced that even a simplified but reliable risk analysis could have been unearthed clearly all plant criticalities. Moreover, the results concerning the risk quantification have been critically analyzed showing that ROA-CCD achieves the same results, in terms of Minimal Cut Sets, of its classical version with a lower effort. This goal is fulfilled by avoiding the unnecessary subdivision of the plant into nodes prior to perform the analysis; in this way, considering only the process variables related to the "key piece of equipment" (in this case, the reactor), records in the ROA format are drastically reduced.