Fouling Prediction of a Heat Exchanger Based on Wavelet Neural Network Optimized by Improved Particle Swarm Optimization Algorithm
Yandong Liang,Lipeng Zhu,Yang Wang,Hao Wu,Junwei Zhang,Jing Guan,Jianguo Wang
DOI: https://doi.org/10.3390/pr12112412
IF: 3.5
2024-11-02
Processes
Abstract:The relevant experimental data of the fouling formation process of a heat exchanger were obtained through the fouling monitoring experimental platform. Whereafter, with regard to the conventional particle swarm optimization (PSO) algorithm, this study commenced from the iteration formula and innovatively presented an optimization approach for improving the inertia weight, thereby obtaining the improved particle swarm optimization (IPSO) algorithm. The wavelet neural network (WNN) was optimized through the application of the IPSO–WNN algorithm, resulting in the development of the IPSO–WNN model. Utilizing this model, a predictive model for fouling thermal resistance was constructed, incorporating input variables such as conductivity, pH, dissolved oxygen, average wall temperature, and bulk temperature, while the output variable represented fouling thermal resistance. Comparative analyses demonstrated that the IPSO–WNN model exhibited superior prediction accuracy and robust generalization capabilities to that of the conventional WNN and PSO–WNN models, as evidenced by significantly lower values across all indicators, including MAPE, MAE, and RMSE. The IPSO algorithm effectively optimized the initial parameters of the WNN, addressing the challenge of local minimum and enhancing the model's overall capacity to identify optimal solutions. This model effectively captures the dynamic trends of fouling thermal resistance during its growth stage and approaches the asymptotic value in the stable stage. Precise prediction models for heat exchanger fouling contribute valuable insights for its prediction in practical industrial applications.
engineering, chemical