Abstract :
A system of classifying failures is proposed as a starting point for increasing the amount of science in the art of reliability engineering. Failures are classified as: type I, lack of qualification, type II, human error, and type III, wear out. Subsets are proposed covering the majority of failures observed in practice. Usefulness of the classification is illustrated by relating specific methods of failure control to each type of failure. Validity of current reliability prediction techniques is discussed. The suggestion is made that validity and accuracy would be improved if new predictive methods were used which are more directly related to the distinct type of failure being examined.
