Evaluation of the Design

(a) The Failure Modes and Effects Analysis (FMEA). The detailed design is followed by a detailed reliability evaluation. As a minimum the system must be subjected to an FMEA. This is a subjective and nonnumeri­cal analysis that exposes potential failure points The FMEA identifies every component in the system by component number and name It lists the various modes in which the particular component would fail (open, short, closed, stuck, etc.) and lists the failure mechanisms that can induce a particular failure mode. It further identifies the relation a failed component has to system failure and to the failure of other systems.

A sample page of an FMEA is shown in Fig. 11 3. There are many variations of this form, and there should be no
hesitation in adapting the form to suit each analysis. The primary function of the FMEA is to provide an under­standing in depth of how the system reacts to all modes of component failure. It is particularly valuable m serving as evidence of conformity with the single-failure criterion If only a minimum effort can be budgeted on reliability analysis, it is generally best spent on an FMEA.

The FMEA form also has space to develop information on failure rates, application factors, test intervals, and repair times in preparation for the more rigorous mathe­matical model of reliability or availability.

(b) Detailed Reliability Model. The designer now has the information at hand to conduct a more detailed reliability or availability analysis of the system. If the designer adheres closely to the framework of the original simple reliability model, the detailed model should gener­ally be satisfied by the same skeletal block diagram, the exception being only in the number of blocks represented Thus the final model should be just as tractable mathe­matically as the exploratory model on which the design is based. If it is not, the designer should examine the original assumptions, particularly the one on component or channel independence, to see if they have been violated.

The main value m performing the detailed reliability analysis is in making certain that some trivial component does not make an unexpected and unwarranted contribu­tion to unreliability. If such a discovery is made, the problem can usually be rectified by the choice of a better component, more frequent and more thorough testing, or by the judicious use of redundant components

When it has been determined that the contribution of a given component to unreliability is trivial, it may be eliminated from the model or lumped with associated components to simplify the computation.