Physics-Guided Bioinspired Multi-Objective Optimization for Explainable Drying-Stage Identification in an R290/R600a Waste-Heat Drying System

Abstract

Integrating low-global-warming-potential refrigerants, waste-heat recovery, and intelligent process monitoring offers a promising route to energy-efficient drying systems. This study experimentally compares the natural refrigerants R290 and R600a in a vapor-compression refrigeration system and utilizes condenser waste heat for apple drying. Refrigeration tests were conducted at an evaporator temperature of −16 °C and condenser temperatures of 25, 30, 35, and 40 °C. R290 provided 73% higher cooling capacity and 59% greater condenser heat rejection than R600a, while reducing the drying time from 300 to 265 min. Conversely, R600a exhibited lower specific energy consumption and a higher specific moisture extraction rate. To identify drying progress without continuous product weighing, a physics-guided bioinspired multi-objective evolutionary algorithm, termed Physics-Guided Ordinal Chaotic Rule-Based SPEA2 (P-CRb-SPEA2), was developed. The proposed framework integrates Pareto-based evolutionary search, chaotic initialization, ordinal error, physical consistency, and rule complexity to extract interpretable IF–THEN rules from six thermal and environmental features. Ten-fold cross-validation showed physically consistent, interpretable drying-stage classification, with T10 and T8 emerging as dominant features for R290 and R600a, respectively. The proposed approach demonstrates the potential of bioinspired evolutionary intelligence for explainable monitoring of integrated refrigeration–drying energy systems.