Root cause failure analysis methods

This article provides an introduction to the various methods most commonly used to investigate the root causes of technical failures in industrial processes. Indeed, for failures which have critical consequences for the availability of industrial plants, it is essential to identify their root causes in order to be able to eliminate them. As there is no international standard specifying the terminology and approach to Root Cause Failure Analysis (RCFA), the first section will present and clarify the different notions of root causes and failures. The only standard on root causes in the broadest sense was published in 2015 by the IEC in the document IEC62740 – Root cause analysis (RCA) – Root Cause Analysis (RCA). In this article, the term "root cause" will be used instead of "initial cause". A classification of the different causes (technical, organizational and human factors) will also be proposed. As the RCFA method is, by definition, a deductive method which determines the root cause from the symptoms, this section will be devoted to defining the notion of causality, and will present the hierarchical and tree-like methods of describing industrial processes which are essential for graphically representing the causal chain. It will emphasize the need to identify the level associated with a root cause, which depends on the maintenance policy adopted (standard exchange or in situ repair). To illustrate the approaches that can be implemented, the procedure followed by the investigators to identify the root cause that led to the fire and explosion on the ground of Boeing 737 flight 120 of China Airlines in 2007 will be presented in the second section. The Japan Transport Safety Board investigators' report identified the primary cause as human error and design error. In the remainder of this article, this example will be used as an illustration to present the principles of the methods and tools detailed in the third section. This third section will recall that RCFA only concerns "post-mortem" analyses relating to a failure that has just been observed. Logic trees and a detailed typology will then be proposed to facilitate the search for root causes. An inventory will be drawn up of the information that needs to be gathered before root cause search tools can be used. This section will present tools based on the collective judgment of experts, such as brainstorming and the Delphi method, emphasizing that they enable consensus to be reached on the nature of the root cause. She will then describe the tools used in the field of quality: the QQOQCP, 5P and Ishikawa methods. As the discipline of operational safety offers a very wide range of methods, this section will present the principle of three methods: fault trees, illustrated by an example, the Apollo approach, which enables the causal chain to be visualized graphically,...