SIMULATION TECHNIQUES IN SYSTEM RELIABILITY EVALUATION

Abstract

The recent past has witnessed rapid developments in the area of complex system designs, such as in the process and Aircraft industries, communications, transportations and many other ambitious areas encompassing space program , Nuclear power plant technology, besides missiles and allied defence systems and other strategic systems. This poses a new problem of choosing the best system design out of the several possible designs alternative even before actually the system is commissioned. This also necessitates selection of highly reliable safety and control systems so as to provide the means of minimizing or containing hazard to human lives and capital loss to the organisation. This dissertation is presented with a view to solve this challenging problem. However, the author makes use of a very versatile technique of Monte Carlo Simulation in preference to widely used analytic approaches which sometimes may not only be difficult but unwildly or infeasible for the analysis of large, complex and maintained systems. The idea behind this type of selection is to bring out the immense potentialities and utility of simulation technique over the analytic approaches. The simulation technique in the area of reliability have not found favour with analysts primary perhaps due to the -11- lack of appropriate or due to the state of art in the area. One thing that is basic to simulation technique is the Fault-tree analysis. This is also relatively a new development for system reliability studies. Besides offering the complete state of art on Fault-tree development a new methodology is presented on automatic generation of the tree logic for the desired system. Anew form of tree graph is suggested in the form of a digraph representation which is easily adaptable to any other techniques. A more general case with increasing failure rates of components is considered. This rightly reflects the philosophy of deterioration model in reliability study. System failure is envisaged with respect to the various operating conditions of the components. In general, failure of stand-by and demand components differ from that of operational units. In most of the existing codes, the important differentiation is not widely accepted. This in the opinion of the author is not realistic. The basic formulation of Monte Carlo method in this regard has been derived. A specific reference is made to Nuclear Safety Systems whose failure is revealed only when the random demand for its operation appears. Throughout the attempt has been to consider system components as maintained. A periodic inspection is -inassumed to detect the fault with stand-by components however, complete inspection is not feasible with demand units. A realistic repair strategy for the system has been considered in this thesis where simultaneous repair of all failed components may not at all be feasible. In general, the system model is considered to have the number of repairmen less than the number of components. This measure results into component waiting time during a repair operation. A cumulative account of man hours wasted on unsuccessful repair is recorded in respect of each component. Time-to-repair of a component is random however, a definite maximum time is allowed for this purpose. A component is declared as failed if the repair is not accomplished within the maximum limit of repair time and is assumed to have been replaced with identical new component, The component is treated as -bad-as-old for reliability evaluation however system availability improves. The whole procedure is illustrated with a system example. The complete computer algorithm supported by computer flow charts are provided at all stages of computation. The appropriate results of studies on practical system have been also reported in the thesis, which might be quite helpful to reliability analysts in deciding upon the course or mode of analysis. The simulation approach is further .extended to a complex system consisting of s-dependent components. Acommon mode failure event is also possible in this analysis. In brief, the present thesis provides a powerful general computer code with such wide considerations as mentioned earlier to reliability analysts and is applicable to any specific system conditions which have not been considered till now. The users of this computer code however will be required to specify the input data and with the change of system, only a change in the Fault-tree logic of sub program called here FALTRI will be necessary. Because" this makes it a useful tool to those who are engaged in the continuous task of developing a better designs and Plans in order to see the effects of conceived modifications in system design on the overall reliability of the system.