Stochastic Models for Promoting and Testing System Reliability Evolution

Abstract : Many systems and systems-of-systems function in sequential-stage fashion, and are constantly on when operative, but are failure-susceptible. Communication systems, power generation and transmission, and vehicular transportation systems tend to fall into this category. We propose a reliability growth model for such systems that is based on design defect removal under a Test-Fix-Test (TFT) protocol: a system is assembled and put under test, for example for a fixed mission time, or multiple thereof. If the system fails during the test time its failure source in some stage is diagnosed, the stage is re-designed, and the new prototype system reassembled (system design is 'fixed') and the system is re-tested. The test (TFT) process is repeated until a pre-determined test period elapses with no failures. This is analogous to the run-test criteria analyzed for one-shot devices 1. In this model we also allow for occasional defective re-design: response to a test failure can actually (and realistically) increase the number of failure-generating design defects. Our model allows quick numerical assessment of TFT operating characteristics, given defining parameter values. It thus provides a planning tool for test designers.