Gradient-free method for multiple soft-fault diagnosis in linear analog circuits

Abstract Fault diagnosis is essential in various engineering fields, including mechanical, electrical, and military systems. Analog circuits are fundamental components of nearly every engineering solution. A well-designed diagnostic process can enhance equipment performance, reduce the risk of failure, shorten service times, and ensure safe operation. The paper concerns soft-fault diagnosis in analog electronic circuits. It proposes an original method for identifying multiple faults in the time domain. The approach avoids the time-consuming sensitivity analyses typically required by multiple fault identification procedures. Combining the concept of deflation with the Nelder-Mead algorithm enables the identification of distinct sets of components, as well as multiple parameter values within each set that meet the diagnostic test. Numerous simulations and laboratory tests confirm the method’s effectiveness. The paper includes representative results of these tests. The technique, after modification, can also be used to diagnose nonlinear circuits and can act as an auxiliary tool in circuit synthesis and design for testability.