Sensory information fusion–based fault diagnostics for complex electronic systems

Abstract

As the structure of complex electronic systems is large, there is inherent uncertainty, so it is often inefficient or infeasible to perform detailed fault diagnostics directly. Fault diagnostics for complex electronic systems need to be performed based on the sensor monitoring data, so in this article, sensory information fusion is introduced to minimize diagnostic uncertainty and computational complexity. Sensory information fusion combines the sensor data by measuring the mutuality of the information and dynamically selecting the most diagnostic decision-relevant sensor subset. This article proposes a sensory information fusion–based diagnostic methodology and framework, which is made up of a dynamical sensor selection that combines sensory information and detailed fault diagnostics based on the sensory information fusion results. A numerical example is conducted to illustrate the proposed sensory information fusion–based diagnostics. The diagnostic results as well as comparisons with different training data sizes confirm that the proposed sensory information fusion–based diagnostics is superior when dealing with uncertain and incomplete information. Furthermore, the sensory information fusion–based diagnostics is shown to simplify the diagnostic complexity and accurately and efficiently solve the diagnostic problems for complex electronic systems. The proposed methodology is also flexible enough to accommodate other fault diagnostic methods so as to perform applicable diagnostics for other complex electronic systems of varying sizes.