A data fusion method for circuit fault diagnosis is presented based on Bayes decision theory.By measuring the temperature and voltage of circuit components

the prior probabilities and conditional probabilities of different sensors to every circuit component are obtained.The target probability values of fault types and attributes for the components are calculated via two-level data fusion with Bayes conditioned probability formula

thus the fault component is found according to maximum probability value.Comparing the diagnosis results based on separate sensor to multi-sensor

it is shown that the later not only improves the accurate rate of fault recognition but reduces the probability of uncertainty.Tests indicate that this method is effective for circuit fault diagnosis.