Comparison of Reinforcement Learning Algorithms and Multi-Agent Ensemble Algorithms in Acoustic Recognition Tasks

Acoustic fault diagnosis allows for timely prevention of failures, accidents, and reduction of operational costs for autonomously operating equipment. This paper presents a comparative analysis of two approaches to acoustic fault recognition: a reinforcement learning-based algorithm and a multi-agent ensemble method. Both developed algorithms were tested using identical data from the MIMII database, which contains audio recordings of industrial equipment operation. The results demonstrate that the reinforcement learning-based algorithm achieves higher processing speed and fault recognition accuracy, while the multi-agent ensemble algorithm provides better noise resistance. The choice of method depends on the requirements for speed, resources, and operating conditions of the equipment.