Membrane Fouling Prediction Based on Tent-SSA-BP

Area of Science:

• Environmental Engineering
• Water Treatment Technology
• Computational Intelligence

Background:

• Real-time membrane bioreactor (MBR) membrane flux monitoring is challenging.
• Back propagation (BP) networks face limitations like local minima and poor generalization for MBR flux prediction.

Purpose of the Study:

• To develop an accurate and robust soft sensing model for MBR membrane fouling prediction.
• To overcome the limitations of traditional BP networks using an optimized approach.

Main Methods:

• Introduced tent chaotic mapping into the sparrow search algorithm (SSA) to enhance population distribution and search capabilities.
• Optimized key parameters of the BP network using the enhanced SSA, creating the Tent-SSA-BP model.
• Compared the Tent-SSA-BP model against BP, GA-BP, PSO-BP, SSA-ELM, SSA-BP, and Tent-PSO-BP models.

Main Results:

• The Tent-SSA-BP model achieved a prediction accuracy of 97.4%, significantly outperforming the baseline BP model (48.52%).
• Compared to the un-improved BP model, the Tent-SSA-BP model reduced Mean Absolute Percentage Error (MAPE) by 96.76%, Root Mean Square Error (RMSE) by 99.78%, and Mean Absolute Error (MAE) by 95.61%.
• Demonstrated superior performance over other benchmark prediction models.

Conclusions:

• The proposed Tent-SSA-BP model offers a significant advancement in MBR membrane flux prediction accuracy.
• The integration of tent chaotic mapping with SSA provides a robust optimization strategy for neural network models.
• The developed model holds considerable engineering reference value for real-time MBR operation and fouling management.