Printed circuit board diagnosis using artificial neural networksand circuit magnetic fields

Testing of electronic systems using conventional testing methods has become more difficult and costly as these systems have become more complex and compact. Conventional testing methods and systems often require lengthy analysis to define testing strategies. These test systems may require lengthy test periods, complex stimulus and measurement instrumentation as well as complicated fixturing. The results are often ambiguous and require further interpretation. This paper presents an exploration of a “non-intrusive” test method based on interpreting changes in the magnetic field close to a Printed Circuit Board (PCB). Currents moving between devices on the PCB produce these magnetic fields. Changes of the PCB operational status due to faults cause changes in the associated magnetic field pattern that can be interpreted by Artificial Neural Networks (ANNs) for fault identification. An apparatus to collect magnetic field measurements is described along with some problems of collecting data. Typical magnetic field patterns for “known-good” and faulted PCBs are presented. Possible extensions of the method are discussed. This paper resulted from internally funded work at Southwest Research Institute (SwRI) concerning non-intrusive diagnostic techniques