Process industry fault monitoring based on PSO-DualCFNN

Abstract To address challenges in industrial process production, such as poor data feature extraction, slow data processing speed, and low accuracy, a fault diagnosis model is proposed by integrating the advantages of the Particle Swarm Optimization (PSO) algorithm and the Dual-channel Cascade Forward Neural Network (DualCFNN). The PSO algorithm is employed to automatically optimize hyperparameters, while the DualCFNN leverages its efficient layer-by-layer data processing capabilities. This model enhances the extraction of nonlinear, high-dimensional spatial features across varying scales, even under conditions prone to model degradation, and significantly improves data processing speed. When tested on the Tennessee-Eastman (TE) dataset, the PSO-DualCFNN-based method demonstrated superior feature classification and extraction performance, achieving recognition accuracies of 99.44%. These results validate the method’s efficiency in data processing.