Testing and Analysis Terms
ANOVA stands for analysis of variance, a method by which the source of variability is identified. This method is widely used in industry to help identify the source of potential problems in the production process and identify whether variation in measured output values is due to variability between various manufacturing processes, or within them. By varying the factors in a predetermined pattern and analyzing the output, one can use statistical techniques to make an accurate assessment as to the cause of variation in a manufacturing process.
Competing failure modes
A model whereby items that fail due to more than one failure mode can be represented as a series reliability system with each block representing a failure mode. The failure modes are considered to be "competing" amongst each other to see which one will cause the item to fail.
A graphical representation of the possible solutions to the likelihood ratio equation. This is employed to determine confidence bounds as well as make comparisons between two different data sets.
A technique that uses the performance (degradation) measurements of a product over time to predict the point at which each unit in the sample is expected to fail. This analysis is useful for tests performed on products with very high reliability, where it is not possible to test the units to failure under normal conditions.
A testing methodology, the goal of which is to show that a product meets a given reliability target (e.g., that it has a reliability of at least 90% at 100 hours). The term is often used to refer to "pass/fail" tests that involve observing the number of units that have failed by a given time, rather than recording the failure times for each unit (i.e., reliability testing).
A measure of the relative contribution of a component’s contribution to the overall system’s reliability. The importance measure of a component is equivalent to the first partial derivative of the component reliability with respect to the system reliability.
Life data analysis
The statistical analysis of failure and usage data performed in order to mathematically model the reliability and failure characteristics of a product.
Monte Carlo simulation
A method of generating values from a known distribution for the purposes of experimentation. This is accomplished by generating uniform random variables and using them in an inverse reliability equation to produce failure times that would conform to the desired input distribution.
A method of analysis that allows the user to characterize failure data without assuming an underlying failure distribution. This avoids the potentially large errors brought about by making incorrect assumptions about the distribution. However, the confidence bounds associated with nonparametric analysis are usually much wider than those calculated via parametric analysis. Additionally, predictions outside the range of the observations are not possible.
see Probability plotting paper
A type of plot that linearizes a distribution’s cdf, allowing the user to manually plot failure time vs. estimated unreliability. Provided that the plotted points fall on a relatively straight line (thus indicating that the chosen distribution is a good fit), the parameter estimates can be obtained from scales on the plot. This is a crude, time-consuming method of fitting a distribution to failure data, but it was practically the only method available prior to the widespread use of computers.
Probability plotting paper
A specially designed type of graph paper that allows the user to plot failure time vs. unreliability as a linear function. Plotting paper constructions varies from distribution to distribution. Probability plotting papers that have been generated by ReliaSoft's software are available on the Web at http://www.weibull.com/GPaper/index.htm.
Recurrent event data analysis
A testing methodology where the observed events are dependent and not identically distributed (such as repairable system data), or where the analyst is interested in modeling the number of occurrences of events over time rather than the length of time prior to the first event (as in life data analysis).
A testing methodology in which test units are tested consecutively instead of simultaneously.
The stocking of spare units or components based on the anticipated number of failures for a given mission or length of operation.
SPRT stands for sequential probability ratio test. This is a type of accept/reject sequential testing in which accept/reject boundaries are defined by the user and units are sequentially tested until either the accept boundary or the reject boundary have been reached and a decision is made about the suitability of the units.
A method by which the probability of failure of an item is calculated by superimposing the distribution of the item’s strength over the distribution of the stress it will encounter during normal usage.
A "pass/fail" demonstration test, where a product is shown to meet a specified reliability requirement if all units survive the test.