Optimal Control of Performance Risk for Large Potentially Dangerous Systems (LPDS)

The paper considers the safety problem for large potentially dangerous systems (LPDS). Disruption of their normal operations may lead to casualties, ecological and property damage. Solution to the above problem is sought in the framework of risk control of LPDS during their normal operation, based on the principle of preventive actions. Risk is described as the product of conditional probability of failure and the overall consequences of such failure. Methods of brining down risk analysis problems to reliability problems are presented. They are based on the following: assessments of “cost of life” (as economic equivalent of casualty); simultaneous optimization of the LPDS and its safety subsystem (expansion of the object of optimization). Such an approach allows unification and merging of structural reliability theory and probabilistic risk analysis. A quantitative method of damage size (the first component of risk) assessment is described, based on computer modeling of a full group of scenarios of a structural failure developing into a full blown LPDS catastrophe. As a result of modeling, the destruction zones and the character, size and probabilities of all kinds of damage (casualties, ecological damage, loss of property) are assessed. It is proposed, as the main method of securing LPDS integrity and safety, to equip each LPDS with suitable monitoring/inspection/maintenance systems, designed as an instrument for controlling the second component of risk (conditional probability of failure), on the basis of a three-level (warning-alarm-failure) control policy. In the outlined format maintenance/repair is considered as optimal control of random degradation and renewal functions, interaction of which forms a certain regeneration process. Analysis of this process allows defining the optimal triggering levels of deterioration parameters or risk that minimize total expenditures of LPDS performance while ensuring its safety. The problem formulated above naturally embodies all existing maintenance methods (based on admissible performance time, rate of failure and on actual and prognosed system condition). Further, the problem of optimal cessation of performance is solved. It allows convoluting a multi-parameter problem into a one-parameter problem and defining the ultimate permissible level of conditional probability of failure. The described methods of risk analysis and control were used in residual lifetime monitoring systems for oil pumping aggregates and for main oil pipe line segments repair prioritization.