Multi-objective Optimization Design of Process Parameters for Organics Removal from Industrial Waste Salt by Pyrolysis-Oxidation Method

Efficient removal of organic impurities is essential for the resourceful utilization of industrial waste salt. Low-temperature pyrolysis-high-temperature oxidation is an effective approach; However, issues such as suboptimal parameter matching, low removal efficiency, and high energy consumption remain. This study employs machine learning and multi-objective optimization algorithms to enhance process efficiency and reduce energy consumption. Organic removal rate and energy consumption per gram are set as optimization objectives, while pyrolysis temperature, pyrolysis time, oxidation temperature, oxidation time, and air flow rate are considered optimization variables. A BP neural network combined with an improved Tuna Swarm Optimization algorithm (TSO-BPNN) is developed based on experimental data. NSGA-II is applied to obtain Pareto-optimal solutions, which are further refined using the CRITIC-TOPSIS evaluation model. Results indicate that the improved TSO-optimized BP neural network improves prediction accuracy with lower error metrics, achieving an average optimization improvement of $19.81 \%$ in organic removal rate and $17.21 \%$ in energy consumption per gram. The optimized process parameters are pyrolysis temperature of $279^{\circ} \text{C}$, pyrolysis time of 30 min, oxidation temperature of $552^{\circ} \text{C}$, oxidation time of 16 min, and air flow rate of $215 \text{ml} / \text{min}$, leading to an increase in organic removal rate to $91.2 \%$ and a reduction in energy consumption per gram to $1.865 \text{kW} / \text{g}$, representing improvements of $6.29 \%$ and $8.04 \%$, respectively.