720 MECHANICAL DESIGN AND PACKAGING 



Thus, from Equation 13-17 



0.9 = e-^' 

 or 



r = 0.035 (system failures per hour) 



T = - = 28.5 hours (mean time-to-failure). 



Reliability Prediction. In weapons system analysis it is often 

 necessary to effect a reliability prediction on equipment which has not yet 

 been built. Reliability predictions are most useful in comparing two or 

 more alternative systems under consideration, because of the limitations of 

 present prediction techniques; however, they are also of some value as an 

 indication of the reliability any given system can be expected to attain. 

 Reliability predictions are useful as a basis for contractual negotiations of 

 reliability goals or specifications. Further, a prediction will often point out 

 the portions of the equipment which will require special, concentrated 

 reliability efforts. However, in order for a prediction to have any value at 

 all it must be arrived at in a sound and meaningful manner. 



In many areas of science and engineering, an object's future behavior can 

 be predicted with great accuracy. Unfortunately, this is not yet the case in 

 predicting the reliability of electronic equipment. Reliability engineering 

 is still a young science and its prediction techniques are, for the most part, 

 still unproven. Coupled with this uncertainty of prediction techniques is 

 the generally poor quality of the past failure data on which predictions 

 must be based. This deficiency in the accuracy and completeness of 

 available data can be seen to be of overriding importance, when it is 

 considered that the only manner in which it is possible to predict future 

 behavior is by utilizing past data and experience. 



The product rule prediction technique is the one most widely used in 

 reliability work today. Basically, the product rule states that the proba- 

 bility of a system operating satisfactorily is equal to the product of the 

 operating probabilities of all the independent components that go to make 

 up the system. This probability rule holds only when the performance of 

 every component is entirely independent of the performance of every other 

 component. This condition is very rarely encountered in real life since few 

 electrical components operate independently of their associated compo- 

 nents; however, the product rule does give a good first approximation of 

 the equipment's potential reliability. Further, the product rule also 

 assumes that the failure of any single component will result in a complete 

 system failure. Since this is not completely true, the product rule tends to 

 give a pessimistic estimate of reliability. 



