Dependability of complex industrial systems — Operational study

In this issue, which follows on from [S 8 250v2] and [S 8 251] , we take a look at aspects relating to operational safety studies and the conduct of a predictive or operational study.

Dependability data are essential for all predictive and mainly quantitative studies. They are of two types: event-driven and reliability-driven.

Event data are obtained from statistical studies of accidents and full-scale experiments. They therefore concern the "macroscopic" aspect and provide estimates of the behavior of an entire system under certain circumstances (large number of indiscernible or non-quantifiable events). They are particularly useful for assessing risks (probability/severity of consequences) and therefore safety.

Reliability data, on the other hand, is obtained by testing basic system components under given conditions (discernible, quantifiable events). They are therefore "microscopic" and are essential for the predictive methods described in the [S 8 251] file, which are widely used.

In large, high-performance companies, the use of CMMS (computer-aided maintenance management) software that includes modules for collecting feedback enables analysis of all the parameters associated with the circumstances in which failures occur, and the time spent on corrective or preventive maintenance.

Feedback data collection systems are used to assess the performance of operational systems using quantitative indicators such as MTTF (Mean Time To Failure), failure rates, repair times and uptime.