Parameter-Optimized LSTM via DBO Algorithm for Industrial Process Fault Diagnosis

The swarm intelligence optimization algorithm exhibits notable strengths in optimizing parameters for intricate engineering challenges, which serves as an efficient approach to enhancing the performance of various fault diagnosis models. This study investigates the application of fault diagnosis technology in the industrial process field. We propose dung beetle optimization algorithm (DBO) to improve the diagnostic performance of long short-term memory (LSTM) in detecting Continuous Stirred Tank Reactor (CSTR). Collaborative optimization model as the optimization objective, DBO's global search capability is leveraged to achieve automatic parameter optimization. For fault detection in a CSTR system, the enhanced model achieves a 0.6% higher accuracy compared to the standard LSTM, which validates both the efficacy and real-world applicability of the suggested approach.