Optimization of dewatering process of concentrate pressure filtering by support vector regression

Area of Science:

• Chemical Engineering
• Process Optimization
• Machine Learning Applications

Background:

• Filter press dehydration is crucial for concentrate processing.
• Improving operational efficiency and reducing costs are key challenges.
• Existing methods lack advanced optimization capabilities.

Purpose of the Study:

• To investigate filter press dehydration mechanisms.
• To develop and apply machine learning models for process optimization.
• To improve the efficiency and reduce the cost of concentrate dewatering.

Main Methods:

• Construction of machine learning models: Radial Basis Function-Ordinary Least Squares (RBF-OLS), RBF-generalized regression neural network, and Support Vector Regression (SVR).
• Laboratory and industrial simulations of the filtration dewatering process.
• Design and application of control variable-based optimization methods.

Main Results:

• Support Vector Regression (SVR) showed the best simulation effect in laboratory settings.
• In industrial simulations, SVR achieved high accuracy with a mean relative error of 1.57% for moisture and 3.81% for processing capacity.
• The SVR model's credibility was verified with newly collected industrial data.

Conclusions:

• The integration of SVR and control variable optimization significantly improves filter press operation efficiency.
• This approach offers substantial energy savings and cost reductions.
• The findings provide a foundation for intelligent dewatering processes and broader industrial production optimization.