A transform technique for obtaining reliability distributions for multi-component systems

by Jing Yu

Institution: Ohio University
Department: Industrial and Manufacturing Systems Engineering (Engineering)
Degree: MS
Year: 1990
Keywords: Engineering, Industrial; Transform Technique; Obtaining Reliability Distributions; Multi-Component Systems
Record ID: 1639210
Full text PDF: http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1183476239


A transform technique to obtain the reliability uncertainty distributions for multi- component systems is introduced and a PC software package using a histogram approximation algorithm (step by step uniform approximation), for both the Mellin transform and the Laplace transform, has been developed. With the histogram approximation method, the Mellin and the Laplace transforms as well as the inverses of the Mellin and Laplace transforms for a variety of distribution shapes can be obtained by using the algorithm developed in this thesis. The PC computer software package for the system reliability uncertainty analysis using the transform technique has been written in C language (Turbo C) and contains approximately 2800 lines of code. The algorithm and the software package have been verified and validated, and have been proven to be accurate, user- friendly and efficient. In order to demonstrate the usefulness of this technique and the software package a system reliability uncertainty analysis has been conducted for different systems and it was found that the use of the Central Limit Theorem represents an acceptable and close approximation for the log-transform of the unreliability (or unavailability) for parallel systems and for the reliability (or availability) for series systems. The reliability distribution characteristic of a complex system appears to be dependent on the overall system configuration. If the overall system is a parallel system, the system will have the characteristic of a parallel system, and vice versa. It was also found that in a parallel system consisting of multiple components not only the unreliability is reduced (i.e. the reliability is increased) but also the reliability uncertainty dispersion is reduced significantly. However, it was found that in the case of a multi-component series system the reliability uncertainty dispersion may not be significantly reduced. It is envisioned that the developed transform technique can be applied efficiently and successfully to determine the overall system reliability uncertainty distribution in systems such as nuclear power generation systems, mechanical and electronic consumer products, provided that one can assume that the reliability uncertainty probability distributions of the individual components are independent each other.