Study Of Reliability Estimation Method For Onboard Computer System

Reliability is one of the important capability in the onboard computer system, which has been an important guideline to evaluate system quality in many application fields. It may to say that, only a safety and dependable system can ensure all of the work to be run normally. The testing and evaluation for onboard computer system reliability is a quantitative computation. An effective method of reliability evaluation can not only enhance work effectiveness and reduce testing time, but also reflect real work conditions.Traditional estimation methods for onboard computer system reliability are combination model and testing method. But fault injection is a new estimation method for onboard computer system reliability. Fault injection is a method of injecting faults into system by person, which can accelerate system failure. Fault injection technology has been classified into hardware fault injection, simulation fault injection and software fault injection. By comparison, software fault injection method takes advantages of low cost, easy operation, no corruption to object system. Hardware fault injection technology is adopted to estimate onboard system in old days, which is the study about fault tolerant mechanism especially, and isn't the study about system reliability. Since the item make onboard computer system as background, software fault injection system simulation platform SFIS is built, and interruption method is adopted to inject faults into the system. In SFIS software fault injection system, break point interruption method is adopted to inject transient faults and step interruption method is adopted to inject permanent faults.The study of reliability estimation in this paper is based on software fault injection. In order to estimate reliability of the system after fault injection, first, reliability model of onboard system is studied. After software and hardware model is summarized, computing method for three model redundant reliability of onboard system is purposed. Meanwhile the failure correlation in software model is analyzed, and software reliability model frame based on Markov renewal process is purposed. Second, Markov chain is analyzed in details. The influences about fault transmission are described in Markov, which change the supposition of"faults will induce to failure"when computing reliability. Path probability computation method for onboard system reliability based on statistical usage testing is purposed, which is identical to the results of computation method purposed by James A.Whittaker. But the idea of path probability is a new view for reliability study. Reliability testing flow of onboard computer system is built, and the process of reliability estimation for onboard computer system is described in details. Third, operational profile is study in this paper, and operational profile states graph description is purposed. So that, the test cased selection became easily, which made more failure present under testing. Forth, reliability estimation of adaptive testing algorithm is purposed, and the result of reliability estimation deviation by adaptive testing algorithm is smaller than that by random testing method with equal probability operational profile, and that with general testing method with special operational profile. So that, the value computed by adaptive testing method is near to real value.