Skip to content
Open access

METHOD FOR IDENTIFYING DEVIATIONS OF A MINING MACHINE BASED ON FILTERING OF DATA FROM ONBOARD DISTANCE SENSORS

2026 · Bulletin of the South Ural State University. Ser. Computer Technologies, Automatic Control & Radioelectronics · 0 citations

Abstract

Potash ore mining is associated with high risks, the key of which is the violation of the integrity of the water-resistant stratum due to the deviation of mining machines from the design course. Aim. The aim of the study is to develop a method for accurately identifying deviations of a mining machine in real time to ensure efficient and safe underground mining. Materials and Methods. The initial data used were readings from four onboard distance sensors installed perpendicular to the axis of the combine's movement. The distance between the axes of the front and rear sensors is 8 m, and the excavation width is 5.1 m. To suppress noise caused by vibration and dust, three data filtering algorithms were compared: a robust Kalman filter, a two-stage filter (median Hampel + Kalman), and a model-oriented extended Kalman filter (EKF). Filtering efficiency was assessed using the smoothing coefficient. Based on the filtered data, a geometric model was constructed to calculate the heading angle and reconstruct the combine's movement trajectory. The starting point of deviation from the straight-line course was determined by the criterion of stable parameter exit beyond the 3σ limits. Results. It was found that the two-stage filtering method “Hampel + Kalman” showed the best results (average smoothing coefficient of 0.8528). Based on the filtered data, the movement trajectory of the mining machine was reconstructed, and the moment of deviation from the straight-line course was determined at a distance of 8.925 m from the starting point. Conclusion. The proposed method of filtering and geometric modeling allows identifying deviations of the mining machine from the specified trajectory but is characterized by error accumulation (drift) during sequential trajectory reconstruction and assumes a constant speed. To eliminate systematic errors, future work involves combining the filter with a neural network correcting module trained on a simulation model of movement, as well as using a correction mechanism based on odometric data for periodic reset of accumulated error.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.