XAI–RSM-based identification and optimization of key parameters in a parabolic trough solar collector drying system

Parabolic trough collector systems operate under fluctuating outdoor conditions, and their performance is strongly affected by environmental parameters, particularly solar radiation and ambient temperature. However, determining the dominant parameters and interpreting their combined effects remain challenging because solar thermal systems exhibit nonlinear and coupled behavior. In this study, an integrated explainable artificial intelligence–Response Surface Methodology framework was developed to evaluate the performance of a parabolic trough collector-assisted greenhouse drying system. In the first stage, an explainable AI model based on polynomial feature engineering, Ridge regression, cross-validation, and permutation importance analysis was used to identify the most influential parameters affecting drying, energy, and exergy efficiencies. The model considered outlet temperature, relative humidity, ambient temperature, and solar radiation as input variables. The XAI results indicated that solar radiation and ambient temperature had the highest model-based importance in the current dataset; however, these rankings partly reflect their mathematical involvement in the calculated efficiency expressions. In the second stage, Response Surface Methodology was applied to these two key parameters to examine their individual, quadratic, and interaction effects. Second-order RSM models were developed for drying, energy, and exergy efficiencies. ANOVA confirmed the statistical significance of the models, with p -values below 0.0001 and R 2 values of 0.9407, 0.8570, and 0.9017, respectively. The results revealed that the effects of solar radiation and ambient temperature were not purely linear but depended on their combined variation under real outdoor conditions. Multi-objective optimization identified a favorable environmental window of 32.6 °C ambient temperature and 794.6 W/m 2 solar radiation, corresponding to drying, energy, and exergy efficiencies of 0.021, 0.273, and 0.072, respectively.