Digital machining database for machine learning

Abstract

The purpose of the open-source Digital Machining Database is to support machine learning algorithm development for machining process modeling. The database is made available to the research community and provides organized data for algorithm training, validation, and benchmarking. The data are generated using deterministic, physics-based numerical (time domain) and analytical (frequency domain) simulations for many milling conditions. For each dataset corresponding to a selected set of milling conditions, hundreds of simulation results are provided for individual spindle speed-axial depth of cut combinations. The time domain signals are: 1) tool displacement, velocity, and acceleration; 2) workpiece displacement, velocity, and acceleration; and 3) cutting force. The spindle speed-dependent stability boundary is also included for each dataset.