Optimization of multi stage Co-pyrolysis process using municipal solid waste and sawdust blends: A hybrid approach using iso-conversional modeling and machine learning

Abstract

An innovative approach combining iso-conversional modeling with machine learning (ML), was used to study the co-pyrolysis of Municipal Solid Waste (MSW) and sawdust (SD) blends to optimize product yield along with their improved properties. Blends with MSW-to-SD ratios of 100:0, 90:10, 75:25, 60:40, and 0:100 were utilized and are represented as MSW, SM-I, SM-II, SM-III, and SD, respectively. The optimal conditions of the process were predicted and validated through state-of-the-art ML algorithms. Four machine learning models, Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Networks (ANN) and Extra Trees (ET), were employed to evaluate the activation energy of the co-pyrolysis process. RF showed the highest accuracy (R² = 0.92), followed by ET (0.90), SVM (0.76), and ANN (0.71). The optimal process parameters for co-pyrolysis included a conversion rate of 0.55, heating rate of 10–40 °C min⁻¹, temperature range of 500–600 °C and blending ratio of 0.12–0.3. The study recorded an optimal activation energy range between 68 and 124 kJ mol⁻¹ for an efficient co-pyrolysis of the blends.