Abstract
All risk management starts in determining what can happen. Reliable predictive analysis is key. So, we perform process hazard analysis which should result in scenario identification and definition. Apart from material/substance properties, thereby, process conditions and possible deviations and mishaps form inputs. Over the years HAZOP has been the most important tool to identify potential process risks by systematically considering deviations in observables, by determining possible causes and consequences, and, if necessary, suggesting improvements. Drawbacks of HAZOP are known; it is effort intensive while the results lack completeness and are used only once. Repeat of the exercise is required at several stages of process build- up, and when the process is operational, it must be re-conducted periodically.
There have been many past attempts to semi-automate the HAZOP procedure to ease the effort of conducting it, but lately new promising developments have been realized enabling also the use of the results for facilitating operational fault diagnosis. There are several factors enabling and supporting these improvements. In the first place there are now decades of experience with HAZOP. Secondly, system science and artificial intelligence provide methods that even with present-day laptop computing power provide equipment ontological and classifying structuring (computer aided process engineering, CAPE) and enable identification of causal relations by learning causal structures from data. This paper will review the directions in which improved automation of HAZOP is going and how the results, besides for risk analysis and design of preventive and protective measures, also can be used during operations for early warning of upcoming abnormal process situations. The latter will enhance operator's situation awareness and will guide operators more efficiently to causes. Thereby, use is made of advanced methods of process simulation and data treatment. Freeing manpower for HAZOP-ing of operations may allow HAZOPs of non-routine activities such as start-ups, shutdowns, and turnarounds in which a relatively large fraction of accidents happen because of greater uncertainty due to wider ranges of potential upset consequences and reduced control as a result of non-routine activities.
