Moisture prediction of municipal sludge drying process using BP neural network modeling and genetic algorithm optimization

Abstract

The detection of internal moisture during the municipal sludge drying process is challenging. To explore the dynamic changes in internal moisture during the hot air drying process of municipal sludge and achieve accurate predictions of moisture content(MC) throughout the drying process, experiments were conducted to monitor the moisture variations of municipal sludge under different drying temperatures (50, 60, 70℃), sludge layer thicknesses (5, 10, 15mm), and flow pressure differentials (0.61, 0.85, 1Kpa). Drying time, drying temperature, sludge layer thickness, and flow pressure differential were employed as input variables, while moisture ratio served as the output variable. Predictive models for MC during the hot air drying of municipal sludge were established using both Back Propagation (BP) neural network and Genetic Algorithm-optimized BP (GA-BP) neural network algorithms. Additionally, the drying characteristic curves obtained from experiments were nonlinearly fitted to seven different mathematical models. The results revealed that temperature exerted the most significant influence on the drying process of municipal sludge. Reducing the sludge layer thickness enhanced the drying rate of municipal sludge. The moisture reduction rate of municipal sludge exhibited a trend of initially rapid reduction followed by a slower decline throughout the entire drying process. The Page model demonstrated greater accuracy in describing the hot air drying process of municipal sludge compared to other models. Validation and error analysis of the BP and GA-BP moisture prediction models were performed, with coefficient of determination for both models' test sets reaching 0.99978 and 0.99989, and root mean square errors at 0.49148 and 0.29468, respectively. This indicates that the predictive model established by the GA-BP algorithm exhibits superior performance in accurately predicting the dynamic changes in moisture during the drying process of municipal sludge. The research findings provide a theoretical basis for optimizing municipal sludge drying processes and guiding the efficient utilization of municipal sludge resources. Keywords: Municipal sludge, Drying characteristics, Genetic algorithm, Neural network, Moisture prediction, Sensitivity analysis