INTRODUCTION

Reliability  prediction  is  an  important  method  for  evaluating  a  system design, right from its conceptual stage through development and manufacture and also to assist in controlling changes during the production. This also provides a rational basis for design decisions, choice between alternate concepts, application               of             component            derating         factors,           choosing                      parts                     quality  levels, possibility of redundancies, suggesting environmental control and other related factors. An accurate reliability prediction of a product at the design stage (before it is actually manufactured) is essential for accurate forecasting of supporting and service costs, maintenance planning, warranty costs, marketability etc.

At the early stages of design, reliability is predicted using the parts count method. This is an approximate method, and comparatively easy to perform. This is  generally  used  to  compare  alternate  design  concepts.  As  the  design progresses and more information becomes available, accurate predictions are made using parts stress method.

4.2       SOURCES OF DATA

Reliability prediction is accomplished by generating a reliability model for the system and using appropriate failure rates at part or component levels. The sources for these failure rates are many, such as MIL-HDBK-217 F for electronic components, Non-Electronic Part Data (NPRD), Government Industry Data Exchange Programme (GIDEP) or derivatives from tests on products, or data from devices, which are in use. Appropriate corrections should be applied for getting accurate results. These data are based on extensive in-house tests and feedback gathered by RADC, GIDEP and EXACT.GIDEP    provides     various    data    related     to    engineering,   reliability, maintainability and environment. The Exchange of Authenticated component Test Data (EXACT) mostly operates among European countries.

Major industries manufacturing components and equipment also track field failures and are capable of providing such data, while some manufacturers even conduct component / module life testing.

4.3       OTHER REQUIREMENTS

In addition to failure rate data, Reliability analysis requires the following inputs.

(a)       Part Description :Parts and their applications in the circuit need to be correctly described for any prediction based on part failure rates.

(b)       Environmental Data :These data include the associated natural and induced environments in which the device operates.

(c)        Part Operating Temperature :This includes the internal temperature rise as determined by thermal analysis, junction temperature etc.

(d)       Stress Data :In the case of parts Analysis; operating stress on each part should be analysed  and  appropriate  correction  factors  should  be  applied  to  the failure rate, to account for the effect of applied stress.

4.4       RELIABILITY PREDICTION METHODOLOGIES

There are different approaches of predicting the reliability of the electronic equipment or system depending on the period when the information is required and to what extent information is available. Reliability prediction can be classified into three types:

(1) Feasibility prediction

(2)  Preliminary design prediction

(3)  Detail Design Prediction

Feasibility prediction is intended for use in the conceptual phase of item development. During this phase, the level of design information is restricted to overall aspects of the unit. The configuration data are normally limited to those derived from existing components having functional and operational requirements similar to those of the item being developed.

Parts count and Parts Stress method are the most important methods followed for reliability prediction. In general, Parts Stress Method provides a higher value of equipment reliability as compared to that by the Parts Count Method. Both these techniques are discussed in following sections.