Downloaded from https://www.everyspec.com AMWP 706-10 In this situation, a competent statistician may be reliable at a certain level of tempera- ougjht to be consulted, One may even wish ture. It may also be reliable at a certain to consult the statistician before planning vibration level. However, a combination of the experiments or trying to let historical thew aiame environmental levels may cause data, serious degradation in reliability, Analysis of variance pwrmits the effects of these inter- actions to be evaluated, 2.7 ANALV81S OF VARIANCE Three sources of variations in reliability During Kystem design and development, it are considered:
(I) vartiations caused by may be desirable to establish the relation- each environment acting singly, (2) van- ship between reliabUlity and speciflc environ- ations caused by combinations of environ- mental parameters (Refas. 2 and 6). Also, ments, and (3) a remainder (the residual the system desilnera may wish to determine error) which is caused by slight variations if changes in environmental factors or com- In the production processes and test equip- binations of them have an Important im- ment fluctuations.
The residual error Is used pact on reliability, It Is possible, by careful as a standard saginst which the other sources experimental design, to obtain a considerable of variability are compared to determine amount of information, oven with smaller their statistical signlflcance, sample sizes. Two techniques will be dis- cussed-the analysis of variance and rogres- 2.7.1 STATISTICAL EXPLANATIONS sion analysis, Just as the name implies, Analysis of Analysis of variance permits the effects of Variance (ANOVA) analyzes the variances of individual or combinations of several an. a "et of data to see if some effects are real, vironmental factors on reliability to be de.
or just due to random sampling effects, termined. By use of regression analysis, an Three categories for measurements are in- equation can be derived which relates re- volved: liability to environmental parameters. I. Factoen (e, heat treatment, supply It is very difficult to design a test that voltage, humidity) covers the entire range of environments that an equipment experiences In practice (Refs, 2, Levels within a factor (e.o., high 6 and 16), However, if the monk important voltage, usual voltage, low voltage) parameters can be Isolat.d, a t,,t can be de- signed around them alone with ad !!,a other ?, Replication within levels (e.g., 10 parameters being ignored.
During the test, memureinent, for each voltage). the test parameters 2re allowed to asume a range of values which simulate the operation. hach measurement Is of a performance char- al environment of the equipment, The be- acteristic such as strength or tirne-to-failure, havior of the test units under varioup com- binations of the test parameters is observed, Just to get a broad picture of what Is in- It is then possible to use analysai of varl-..
volved, consider the following experiment on to determine the effect of each parameter Fomc radio receivers, Ilhi time-to-first- acting singly or in combination, failurn Is to be measured for each receiver, The ability to analyze the effects of corn- I. h b'ctor,. T1here are 2 factors: binations of parameters is a very useful part of analysis of 'variance, A component a Ambient temperature 2-. .8 .. ,