Abstract :
Developing meaningful reliability models from failure data often requires a measure of subjective artistry. Data, especially field data, often lacks some underlying detail to directly identify varying failure modes, environmental conditions, or other factors. The analyst must be mindful that unknown factors may be encoded in the failure data but avoid over analyzing every subtle feature in the data. It is desirable to derive an objective approach that would both yield a reasonable overall reliability model and identify distinct, salient signatures within a set of data. This paper reports the findings of a feasibility study that explored the use of an artificial life (ALife) framework to objectively construct reliability models that both model overall reliability and identify these distinct and salient signatures.
