A precision control approach for variable-speed motors in smart kitchen range hoods
Traditional kitchen ventilation systems commonly suffer from issues such as response lag, imprecise airflow matching, high energy consumption, and low smoke control efficiency. This paper proposes an intelligent adaptive airflow control method based on a hybrid physical-AI modeling approach. The system uses a multi-source sensing network to collect real-time data on fume concentration, ambient temperature and humidity, cookware temperature, and acoustic features. Based on this, a lightweight one-dimensional convolutional neural network, CookNet-1D, is designed to achieve highprecision identification of six typical cooking scenarios. To address the mismatch issues in pure physical models caused by disturbances such as filter clogging and duct backpressure, a hybrid airflow model combining physical and neural residual correction is constructed, utilizing a three-layer MLP network to compensate for speed-airflow mapping errors in real time. Furthermore, an adaptive model predictive control strategy is proposed to dynamically configure the optimization objective weights and baseline speed based on the recognition results. Experimental results on the constructed tested demonstrate that, compared to fixed-frequency control, open-loop variable-frequency control, and purely data-driven methods, the proposed method is capable of improving grease removal efficiency while reducing energy consumption and noise. These findings validate the feasibility of the approach in the tested environments and show its potential for enhancing user experience in smart kitchen applications.