In Weibull analysis, the units for age depend entirely upon part usage and the failure mode under consideration. Product lifetimes can be measured in hours, miles, cycles or any other metric that applies to a period of successful operation for a particular product. For example, the age of an automobile tire is likely to be measured in the number of miles or kilometers for which the tire has been used. The age of a burner and turbine is likely to be measured in either the amount of time spent operating at a high temperature or the number of cold-to-hot-to-cold cycles. Thus, component age can be measured in distance, time, mission cycles, duty cycles, number of revolutions, etc., depending upon the failure mode in question.

The best results from Weibull analysis are achieved when each failure mode is analysed separately and the time origin and scale for the age of the component has been attentively considered. Because the best data analysis methods cannot improve bad data, thoroughly investigate data sources to find the root cause of reported difficulties, keeping in mind that a single part can have many failure modes. If the data set contains a mixture of failure modes, tag individual data points to indicate the appropriate failure mode. After the life data is manually entered or automatically imported into Weibull software, distributions can then be fitted to each failure mode.

Although the failure mode generally dictates the most appropriate unit for age, uncertainty about the best age parameter may occasionally exist. For such situations, Weibull probability plots can easily be generated for each alternative age parameter. The best age parameter would then be the one used in the Weibull probability plot that most closely fits the data points to a straight line. Weibull software often provides for automatic selection of the best distribution and optimizes the scale for the life data being analysed.

Because Weibull probability plots usually provide significant knowledge from very little data, graphing what is viewed as “bad” data can even be informative. When operating data is not available or obtainable, for example, the age parameter can be based on calendar intervals. For a failed furnace, the most appropriate age parameter would probably be either operating hours or operating cycles; however, the only data available may be initial shipment and return dates. Although using calendar time for the age parameter may result in a poorer fit and increased uncertainty, a measure of the goodness of fit can easily be calculated to determine if the resulting Weibull probability plot is accurate enough to provide valuable analysis.

When material characteristics such as creep, stress rupture and fatigue are considered, the age parameter is often stress, load or temperature. Although these parameters do not truly indicate age, the resulting Weibull probability plots are interpreted as if they were component ages. Prior to collecting component age for any probability plot, however, ensure that: