a b s t r a c tAnalysis of the Fukushima nuclear reactor disaster will show how to learn from failures using multi-models. This type of analysis can enrich the modelling of causal factors, provide insight into policy making and support decisions for resource allocations to prevent such disasters.The analysis presented here is based on a workshop on learning from failures in which participants were first given a brief about the related theory, then an introduction to the analytical techniques that can be used such as FTA, and RBD. They were then given a brief in the form of a narrative of the accident derived from investigation reports and divided into small groups tasked to analyse the disaster and to present their recommendations both orally and in a written report.All the participants were asked to follow a certain presentation format. Firstly there would be a technical account of the sequence of events that would be based on research of greater depth than provided in the initial summary, one related more to the scope of their analysis. This would be followed by a review of the consequences of the accident, a presentation of their multi-model analysis of the event and a summary of generic lessons and recommendations to prevent future failures of such systems. Finally, collective feedback, focused on the generic lessons gained, was offered.

The self-diffusion coefficient, Dll, of gaseous C 0 2 has been measured by monitoring the transport of C i 4 0 2 along a horizontal capillary. The main aim was to study the behaviour of the coefficient in the critical region, but before doing this its variation with density was measured outside the region so as to establish a 'background' against which critical region behaviour could be compared. The non-critical variation of D l l with density has been found to conform closely to Enskog's classical dense-gas theory, a result which resolves disagreement between earlier tracer studies of CO, but appears to conflict with NMR spin-echo measurements, on other substances, which have found no density dependence. In the critical region a large positive divergence, similar to that found for thermal conductivity, has been observed. This again contrasts with the spin-echo studies, in which no critical anomaly whatsoever has been found. Because of the inherent experimental uncertainties the critical region findings, although interesting, are regarded as preliminary only, and in need of corroboration.Der Selbstdiffusionskoeffizient D , von gasfirmigem C 0 2 ist durch die Beobachtung des Transports von C1402 langs einer horizontalen Kapillaren gemessen worden. Das Hauptziel war das Studium des Verhaltens dieses Koefizienten im kritischen Bereich. Seine Abhangigkeit von der Dichte auBerhalb des kritischen Gebietes wurde vorher untersucht, um einen ,,background" zu bekommen, mit dem das Verhalten innerhalb des kritischen Gebietes verglichen werden konnte. Die nicht-kritische Anderung von D , , mit der Dichte geht mit Enskog's klassischer ,,dense-gas"-Theorie konform. Dieses Ergebnis zeigt Unterschiede zu friiheren Untersuchungen mit der Tracermethode und steht auBerdem im Gegensatz zu Ergebnissen von NMR-Spinecho-Untersuchungen an anderen Substanzen, bei denen keine Abhangigkeit von der Dichte gefunden wurde. Im kritischen Gebiet wurde eine groBe positive Divergenz, ahnlich der der Warmeleitfahigkeit, beobachtet. Dieses steht wiederum im Gegensatz zu Spinecho-Untersuchungen, bei denen keine kritische Anomalitat oder dergleichen gefunden wurde. Wegen der experimentellen Unsicherheiten werden die Ergebnisse, obwohl sie interessant sind, als vorlaufig betrachtet und bedurfen noch weiterer Bestatigung.

No abstract

There appears to be an increasing awareness that organizational and managerial factors play a critical role in ensuring plant safety. A detailed study is presented of one branch of industrial management, viz. maintenance management. It has highlighted the nature of some of the links between managerial practices and the reliability of plant equipment and personnel. The causes of several major accidents are analysed using a systems approach, with the aim of understanding how the maintenance function was involved. A key conclusion is that, prior to major accidents, there is often a lack of detailed safety objectives and long-term safety control. Without these it becomes very difficult to assess the adequacy or otherwise of other maintenance management elements such as maintenance strategy, resources or administrative structure. In the absence of tight safety and reliability control and consequent corrective actions a mismatch can develop between the perceptions of management and the actual condition of the plant.

No abstract