Under The Small Sample Reliability Test Method And Data Processing

With the development of science and technology, reliability plays a more and moreimportant role today. Especially in some special field, such as aerospace, military andsome other products with high reliability and long life, reliability has an irreplaceableposition. Reliability test runs through the whole life cycle of products, and provides datasources for research and development of products, mass production, and evaluationremaining life, etc. For some products with high cost, or just put in use, research sampleis not enough, these cases have a characteristics of small sample. So how to predict lifeof products more accurately with some exiting small data, to calculate reliability andother related reliability parameters is particularly important.This paper will solve two kinds of reliability test data processing with smallsamples, the two mainly methods are as follows:The first one is to process accelerated life test (ALT) data, which is with thecharacteristic of small sample. This part will use computer simulation, with the help ofBootstrap method and aiNet method; put forward a new method of small sample dataprocessing. This method based on data expanded and data mining, by using simulationtest instead of real test, makes small sample into appropriate scale sample, and then useclassic least squares estimation(LSE) method to estimate related parameters by linearfitting. At last, gives a sample of electro migration accelerated life test, to verify therationality of this method, and proves that it can improve the accuracy of reliabilityparameters prediction of products which are working under normal stresses.The second one is to reduce the needs of success-failure test number to makeevaluation more accuracy. In some area, reliability compliance test is success-failuretest. This part will introduce a way to make success-failure test into measure test. Thensome typical reliability test specification and corresponding data processing method areput forward, which is mainly used in nonelectric products of aviation. With the help ofthese methods and criterions, reliability compliance test needs less number of tests toevaluate the parameters, so as to reduce the sample size. This way not only can avoidthe defect that traditional success-failure test needs large scale test, but also can give the confidence interval of parameters. At last, one servo system reliability compliance testwas used to verify the validity of this method.