Robotic abrasive belt grinding with consistent quality under normal force variations

The grinding of blades for aeroengines under constant normal contact force is challenging. The surface roughness (Ra) and profile dimensional accuracy of the blades have notable influence on the overall performance and service life of aeroengines. This study aimed to minimize the influence of the changes in the normal contact force on Ra and the uniformity of material removal depth (MRD) in the cut-in/cut-off stage of robotic abrasive belt grinding and in the parts that undergo large changes in curvature. First, we established models for predicting Ra and MRD using orthogonal central composite design theory and a broad learning system algorithm. Second, by combining the established predictive model and applying the sensor-measured grinding force, we established a multiple learning backtracking search algorithm to derive the adaptive process parameters for the optimization objective function. Finally, the grinding quality was experimentally evaluated using the obtained process parameters, and the maximum errors between the test values and model-predicted values of Ra and MRD were 14.2% and 13.4%, respectively. The test results indicated that the method proposed in this article may be effective for achieving consistent Ra and MRD, which can considerably improve the quality of blade grinding.