Study on blockage monitoring of differential-speed roller potato-soil separation device

Soil blockage in potato-soil separation devices during operation significantly compromises harvesting efficiency, necessitating real-time monitoring solutions. This study developed a multi-sensor data acquisition system to capture strain signals from the comb teeth, vibration signals from the differential-speed roller and rod screen bearings, speed signals from the differential-speed roller and rod screen under different working conditions. Then, a genetic algorithm was used to optimize the Gauss kernel parameters, and a support vector machine model for identifying soil blockage was established based on the extracted features. The results show that the device is a risk of blockage if one or more of the following conditions occur: (1) the filtered peak strain of comb teeth 5 and 6 exceeds 0.3×${10}^{-4}$; (2) the amplitude of meshing frequency between rod screen and rubber wheel is reduced to 0.01; (3) the peak-to-peak value of soil skateboard vibration signal is lower than 70 % of the normal value; (4) the rod screen and the differential-speed roller speed are lower than 80 % of the normal value. The model with optimal kernel parameters exhibited high accuracy, with 96.7 % precision, 94.2 % recall rate and 95.4 % F1-score for the test set. This study establishes a theoretical framework for intelligent blockage monitoring in potato harvesters, with practical implications for improving harvesting efficiency.