Re-occurring deviations

To obtain an answer to the research questions posed, pro-active safety indicators used in today s chemical process industry were analysed and the deviations upon which they are based were compared with deviations present in the accident trajectories of 70 recent accidents. By comparing these two sets of deviations, it was observed that although re-occurring deviations with no direct perceived safety related consequences were present in the majority of accident trajectories they were nevertheless not addressed by the commonly used pro-active safety indicators. These so-called indirect perceived safety related deviations may therefore be more important indications of a possible accident than was thought previously. 

Based on the results of the analysis the first research question could be answered affirmative. Precursors (defined as re-occurring deviations in an operational process) do exist and can often be obtained from everyday quality, maintenance and reliability data. 

To verify the developed concepts underlying the structured 7-stage protocol in a reactive way, they were applied to an analysis of recent accidents in the Dutch chemical process industry. Despite the limitations in the information available from the accident database, it could be deduced that all accidents were preceded by precursors, and even that similar precursors had led to similar accidents, implying that companies had failed to learn from these re-occurring deviations which were in fact pre-warning signs of impeding accidents. 

Most deviations (82%) belong to the high likelihood category, i.e. they are frequently re-occurring deviations. Most of these deviations (71 %) have low consequences implying that they are signs indicating events that are not directly related to safety consequences. 

It appears that in constructing pro-active Sis, a specific class of data has not been retrieved from the operational business process in chemical process industry. This specific class of data is of re-occurring deviations in daily operations that do not have direct consequences for safety. 

This example shows that the focus was put on deviations that have direct safety related consequences ( high consequence category), such as violation of safety procedures or shortfalls in safety training. However, the normal operation in which this ( indirect safety related) overheating deviation occurred had not been identified as relevant ( low consequence category). Finally this re-occurring deviation caused a similar accident only this time with more severe consequences. 

From this analysis a hypothesis posed as follows can be derived Frequently reoccurring deviations, present in the operational process of an organization, can be identified in the causal path of an accident. Subsequently, a prediction can be made, that the hypothesis will be valid in many cases. The final test of this hypothesis, confirming or disputing it, will be done in Chapter 6, where several accidents will be analysed on re-occurring deviations present in the accident trajectory. 

In the previous Chapter it was shown that most accidents are preceded by deviations in the operational process, e.g. Heinrich (Heinrich, 1959), Turner (Turner, 1978), Leplat (Leplat, 1987), Reason (Reason, 1997), etc. Additionally, it was shown that a specific class of deviations is present which is not covered by current pro-active safety indicators. These deviations are characterised by a high likelihood and low perceived safety related consequences and were defined as precursors and re-occur in the operational process of the organization prior to an accident. In order to find these deviations in a real life operation and to eventually find their underlying causes, the concepts of re-occurring deviation and operational process have to be explained in more detail. The various definitions and concepts derived in this Chapter are necessary to understand the next Chapters, which shows how they are applied in practice. 

Please note that one theory-in-use might differ from other theories in use of the same transformation, because different individuals will hold different images of the prevailing theory-in-use, so they will judge specific situations in a distinct matter. Having set a definition of a deviation, the concept of a re-occurring deviation will be further explained in the following sub-Section. 

In the end of the previous Chapter, a precursor was defined as a re-occurring deviation in the operational process. The concept of a deviation in an operational process was stated in the previous sub-Section. However, the concept of a re-occurring deviation will be explained in this sub-Section. 

Considering both theoretical and practical criteria, a definition for a re-occurring deviation is formulated in this thesis subsequent deviations with equal value to the first deviation, occurring in an equal input, output, or resource, on the lowest aggregation level as recorded in a company. An example to clarify this definition is depicted in Figure 18. 

Figure 18 shows a list of three deviations, at different aggregation levels. At the highest aggregation level (lowest detail), all deviations are equal, on the lowest aggregation level (most detailed level), all three deviations differ. Thus a re-occurring deviation is identified on the lowest aggregation level as present and where all three deviations are superficially equal. 

A re-occurring deviation can be defined as multiple deviations from the transformation s theory-in-use, or espoused theory, equal to each other on the lowest aggregation level as recorded inside a company, all occurring in the same input, output, or resource of the transformation. 

In addition to these problems, several re-occurring deviations (precursors) were present in the operational process. To identify these precursors, deviation data had to be collected and analysed. However, only planning schedules and product quality data were available. Therefore, additional data was obtained mainly by interviews. 

If a second deviation of the same nature occurs simultaneously with the first deviation, does the re-occurring deviation fall into the classification precursor ... 

Is there a minimum number of re-occurring deviations that must occur to qualify both control loops, i.e. single and double as ineffective ... 

Identify precursors Identify individual re-occurring deviations in the operational processes (precursors) of the selected research area. 

In search for similar accidents with the same re-occurring deviations prior to the accident, two cases were identified here. In both cases an explosion occurred during the unloading of a tanker when a pump overheated causing a decomposition of N2O vapour, resulting in an explosion. 

This accident will now be analysed according to the analysis flow chart. The first step is to identify re-occurring deviations prior to the accident. In this case the repeated complaints of the operators about the difficulty in operating the stack damper and the re-occurring flickering of the flames are identified as soft and hard precursors, respectively. 

The deviations, already present, had not led to any serious consequences yet. In fact the existence of a precursor (re-occurring deviation) implies the ineffectiveness of a corresponding control loop. 

Under the stable conditions of the regular way of working re-occurring deviations did not escalate. They remained relatively harmless, until the conditions were changed. Then these harmless deviations could escalate, causing an accident with serious consequences. 

Please note that, some of the identified precursors (re-occurring deviations) are manifestly affected safety barriers. That is to say the precursor itself is a safety barrier and malfunctions repeatedly e.g. the pressure relief valve defect (number 4 in Table 16). Moreover, the precursor may be a process control measure (often tripping of a pressure relief (safety) valve). The presence of these kinds of precursors illustrates what is stated in Chapter 3, that if actors in the operational process don t perceive a deviation as possessing direct safety related consequences, they permit these deviations to exist in the operational process. 

The answer to the question is that in the vast majority of all accidents, re-occurring deviations, which were defined as precursors, are present. That a variety of events were present prior to accidents, was stated by Heinrich (Heinrich, 1931). He combined the common cause model and the descriptive iceberg model, stating that prior to an accident, increased numbers of near misses, errors and recoveries are present. Subsequently, Turner (Turner, 1978) identified a so-called incubation period prior to accidents, in which all sorts of events occur unnoticed or are misinterpreted. However, neither study indicates types, or categories of events that can act as precursors of accidents. 

---