Data collection and recording guidelines for achieving intelligent diagnostics

Whether using human or machine intelligence, the best decisions are made when using all available information from all relevant sources. In contrast to traditional automatic test equipment (ATE) programming techniques, which stop on failures to perform a diagnosis, we would collect from the entire sequence of electronic tests, and integrate these data with circuit topology and external data (including thermal imaging) to obtain a best informed diagnosis. Novel on-the-fly neural network paradigms provide the capability to make correct assessments from a large history of repair data. Applications with these paradigms require a computer with sufficient horsepower (such as a PC) and high-level programming languages. The circuit topology assessment program can identify “dead” nodes through an entropy calculation, making it possible to perform circuit diagnosis in the absence of good/bad historical information. In theory, this technique could be applied to any automatic test equipment platform, provided that the data collection activities were in place