During the drying of Fritillaria ussuriensis, complex nonlinear interactions occur between process parameters and quality attributes. Conventional approaches rely on empirical trial-and-error, limiting precise control and inverse optimization. This study proposes a hybrid optimization framework combining the whale optimization algorithm (WOA) and particle swarm optimization (PSO) to establish a bidirectional mapping between process variables and quality indicators. The WOA is applied for global optimization of the random forest (RF) hyperparameters, followed by PSO for local refinement. The resulting model enables both forward prediction (from temperature, heating air velocity, dehumidification air velocity, and infrared power to quality indicators) and inverse optimization (from target quality to process parameters). The model achieves high predictive performance, with mean R2 values of 0.9739 (forward) and 0.9736 (inverse), outperforming WOA-RF, PSO-RF, and conventional RF models in accuracy, stability, and generalization. Industrial validation shows prediction errors below 10%, meeting engineering requirements. These results provide an effective approach for drying optimization and support intelligent modeling of rhizome-based medicinal materials.
Wu et al. (Mon,) studied this question.