On the reliability and validity of safety analysis: Dissertation

The use of safety analysis is becoming general in industrial applications.Most of the analyses have been voluntary but, under the safety legislation in some countries, the authorities may nowadays require the use of safety analysis.Simultaneously there has been growing criticism levelled at such an analysis concerning particularly omissions in the identification of accident contributors and inaccuracies in the quantitative assessment of risks.Despite this conflict empirical evaluations of the limitations of safety analysis have been quite uncommon.Such evaluations form the subject matter of this investigation.This investigation evaluated two methods of safety analysis - work safety analysis (WSA) and hazard and operability study (HAZOP).The interanalyst reliability was examined by conducting parallel independent analyses of the same object, the validity of a method in its intended scope - expressed here as coverage - was examined by employing the information obtained by other methods of safety analysis as the reference, and the validity in the sense of accident prevention by taking information on actual accidents as the reference.Three concepts - hazard, deviation and determining factor - were employed in the evaluations.The importance of the ob served disparities was assessed by taking the number of remedial measures implied by individual methods and number of accidents and days of absence as the criteria.The results of the investigation demonstrated that a large amount of valuable information for accident prevention can be obtained safety analysis.The investigation also revealed some obvious deficiencies in the identification of accident contributors.One of the main factors influencing the omissions relates to the nature of the search procedures of WSA and HAZOP, which resemble a heuristic rather than an algorithmic procedure.Several proposals for improving the quality of safety analysis are presented on the basis of the results.These concern, among other things, the search procedures, training of analysts, development of expert systems and the use of safety analysis and its results.Because of the limitations by the constraints of time and resources the results of this investigation and the proposals presented should be considered only as a basis for further research and development.