This article presents a new method for the classification of machine failures using an example of selected generating sets. Measurements and an analysis of the electrical parameters, such as the phase-to-phase voltages at the terminals of a synchronous generator, armature current, and voltage and excitation current of a synchronous generator, are the basis for determining the failure symptoms. The existing energy quality coefficients are adopted as symptoms for the assessment of failures in the monitored generating set. We assume in this method that the description of the input−output relationship is in the form of a black box and use the binary diagnostics matrix (BDM) to investigate the failure−symptom relationships between the inputs (intentional failures) and outputs (failures symptoms = fault-sensitive power quality (PQ) coefficients). The method presented in this article enables the detection and classification of both electrical damage in a synchronous generator and mechanical damage in a diesel engine. It is anticipated that further work and development of the method will focus on the implementation of the algorithm in the form of software into a miniature IoT module for the automatic classification of failures.