Machine Learning Based Random Forest Prediction for Solar Dryer under Thailand Climatic Conditions

In this study, selective and non-selective absorber-coated trays were employed to dry carrots and pears. Two trays with a selective absorber coating (1 mm thickness) were used, each loaded with 600 g of sliced carrots and pears. Similarly, two additional trays with a non-selective absorber coating were utilised. Furthermore, the performance of both selective and non-selective absorber-coated trays was compared with conventional open sun drying. The selective absorber-coated tray demonstrated higher thermal energy absorption and enabled the drying of carrots within 2 days, resulting in a weight loss of 529 g. In contrast, owing to the higher fructose content in pears, approximately 7 days were required to achieve a weight loss of 480 g. Overall, more effective drying was observed in the selective absorber-coated tray compared with the non-selective absorber-coated tray and open sun drying. A two-way ANOVA was performed to evaluate the effectiveness of the drying process across the examined methods. The higher F-statistic value of 63.566 indicates significant biological differences between carrots and pears, validating the comparison of their drying behavior. In addition, the F-statistic value of 5.099 for the drying method suggests a notable variation between selective and non-selective absorber-coated trays. It is concluded that the selective absorber-coated tray facilitates enhanced heat transfer to both carrots and pears. Following the ANOVA analysis, a Random Forest (RF) regression model was applied to the selective absorber-coated tray. Notably, the solar dome inner cover temperature exhibited a feature importance of 0.958, indicating its dominant role in increasing the tray temperature and improving drying performance. The average difference between actual and predicted tray temperatures was 0.16°C, demonstrating that the RF model is stable and reliable, with an R2 value of 0.975. The integration of machine learning approaches in solar dryers can support predictive maintenance and reduce operational costs.