Causal Discovery in Electronic Circuits and Its Application in Fault Diagnosis
Mohammed Tuhin Rana, Soham Sundar Chakraborty, Murti V. Salapaka · 2023
This article demonstrates that causal discovery approaches can be applied to analog electronic circuits made of Bipolar Junction Transistors (BJTs) to find out the causal relationships among different variables of the circuit. Moreover, the obtained causal relationship structure in the form of a Directed Acyclic Graph (DAG), can be used for diagnosis and analysis of such circuits. First, it is shown that the operation process of a transistor has an inherent notion of causality, which is then exploited to show that the various parameters of a BJT circuit can be expressed in the form of Structural Equation Models (SEM). The results demonstrated using data generated using LTspice establishes that the causal structure of a BJT circuit can be retrieved using data driven causal discovery algorithms. This opens new horizons for analysis and diagnosis of BJT circuits. An example case study of circuit diagnosis is presented to showcase the capability and efficiency of the proposed method.