Development of a Machine Learning Method for Evaluating Tool Wear in the Gear Skiving Process Based on the Workpiece Topography

Abstract Manufacturing processes are influenced by external and internal disturbance variables, such as oscillations, clamping errors, and tool wear. At the end of tool life, the maximum wear mark width increases exponentially, leading to tool breakage or higher regrinding costs due to higher material removal. Progressive tool wear also negatively impacts workpiece quality. Gear skiving is a flexible and productive process which enables high quality requirements to be fulfilled, increasing its industrial and scientific relevance. To ensure the quality of manufactured gears, quality control is performed by measuring geometry and roughness of tooth flanks. In this paper a method to evaluate tool wear via gear measurements is presented. In this way, the gear measurement can be used to assess the wear condition of a tool in order to avoid the exponential increase in the maximum wear mark width. For these investigations, the potential of artificial intelligence-assisted analysis, which has not been sufficiently investigated in the field of gear manufacturing, is exploited. To this end a deep neural network is developed which assesses the wear condition based on the gear topography.