Failure Concepts in Reliability Theory

The reduction and control of failures in electronics systems is essential due to: 1) the need for maximum safety of personnel on exotic missions; 2) the long unattended life required of space systems; 3) the high dollar costs to buy additional systems and support additional maintenance personnel required to compensate for the unreliability and system down-time. As practiced by advanced organizations, efforts to reduce parts failures by design controls and conventional statistical reliability methods are a necessary condition of reliability, but not a sufficient one. For clarity and background, present reliability and failure rate theory and practices are reviewed briefly and some limitations pointed out. The physics and the mechanisms of actual failures are discussed and it is pointed out that failures follow the laws of cause and effect; that they are therefore intrinsically predictable. The concept of random failures is valid in a statistical sense but it should not discourage efforts to predict and prevent individual failure by all means open to us. A powerful method of failure prevention is a thorough system test prior to use when the test is geared to look for failures where they can be detected. Lack of precision in terminology is part of the problem of failure control; we commonly intermingle references to failures by source, by cause, by end-effect, and by symptom. An improved organization and definition of failures is developed through which the methods, means, and limitations of failure detection, prediction, and prevention can be established.