Accident data shortcomings

The analytical studies of accidents can be performed either by a conservative approach or by a best estimate approach. In the first case, conservative assumptions are adopted for initial and boundary conditions and for the various elements of the evaluation (correlations, parameters, equipment availability, etc.). Apart from the obvious advantages (for safety) of this approach, it, however, frequently leads to a completely unrealistic description of the real accident sequence, with a distorted timing of the events and the masking of interesting phenomena (see also Chapter 27). Because of these shortcomings and the current maturity of best estimate codes, they should be used in a safety analysis in combination with a reasonably conservative selection of input data... 

For obvious reasons, the measurements of fission product behavior inside and outside the affected plants which were performed during and after the accidents show considerable shortcomings and are by far are not as complete and reliable as measurements performed in the course of planned experiments. For this reason, the results reported have to be critically evaluated prior to their application in fission product transport and behavior models moreover, most of these data can only be used in very limited areas. But they are the only results which have been obtained in full-scale reactor accidents and, therefore, they may be able to contribute to our understanding of the processes in question and of relevant countermeasures. 

Data show that wire strikes have been the most common type of aerial agriculture accident since at least 1985 consistently accounting for about one-third of human factors related accidents. While this statistic alone suggests fundamental shortcomings in methods adopted by the industiy to address the problem in the past, more concerning is the statistic that almost all of the wire strikes involve wires that the pilot knew were there and had previously avoided, but due to a range of well-known human error conditions, such as excessive task load, distraction, fatigue, etc., the pilots failed to avoid the wires. 

One shortcoming of the Maycock et aL (1991) and the Forsyth et al. (1995) studies was that experience was measured in months of driving rather than the more commonly accepted measure of exposure miles or kilometers of driving. However, in an analysis of data collected in the Netherlands approximately ten years later, and using accidents per billion kilometers driven, Vlakveld (2004) obtained a very similar pattern of results, which are presented in Figure 6-7. Another advantage of the Vlakveld results as presented in Figure 6-7 is that the line plots are of actual empirical crash data rather than of functions derived from a data-based model. 

We will review different measures or indicators of SHE performance and their merits and shortcomings when applied in SHE management. Part IV starts in Chapter 16 with an overview of SHE performance indicators. The following chapter presents different indicators using data about accidents and losses as the main input. Chapter 18 presents various indirect indicators based on data about the process that may result in loss. In Chapter 19, finally, we will look into SHE performance indicators, where the focus is on the nnderlying contributing factors and basic causes. 

---