Estimation Methods for Determining Failure Rates of Component Families from Observed Failure Rates of Units and Yields of Component Families from Observed Yields of Units

Abstract

<p> In the field of communications electronic plug-in units operating together form a system. In the event of failure a plug-in unit can easily be replaced by another. Each unit consists of electronic components soldered onto a board in a particular pattern. A component may be either a single electronic part such as a transistor or a combination of single parts such as an integrated circuit. Electronic components with similar properties have been grouped into families. This reduces the number of parameters to be estimated from the observations available.</p> <p> The method of maximum likelihood is used to estimate the failure rates of component families. The number of unit failures and the number of units in use observed during measured periods of time and the component family makeup of the observed units are used to make the estimates.</p> <p> The probability of an electronic component from a given component from a given family being acceptable after the production process will be referred to as the component yield for that family. Similarly, the probability of a given type of unit being acceptable after the production process will be referred to as the yield for that type of unit. By taking the logarithms of the yields, the estimation problem can be reduced to the linear problem of estimating logarithms of component family yields. Using unit yields, the total number of each type of unit produced, and the component family makeup of those units produced, component family yields are estimated. The method of maximum likelihood is applied directly to the data and the method of weighted least squares is applied to the linearized problem.</p>