Constrained identification of virtual CNC drives using a genetic algorithm

This paper presents a genetic algorithm (GA) for identifying virtual models of machine tool drives with minimal intervention to the production machine. The proposed solution builds on the "rapid identification" concept reported earlier in literature, in which a short series of motion data is captured from the Computer Numerical Control (CNC) and used for closed-loop transfer function identification subject to stability constraints. Compared to the Lagrange multipliers solution that was used in the original implementation of rapid identification, the proposed GA is both significantly faster and better suited for industrial use. The novel contributions in the paper include the reduction of the search space from eight unknown variables down to three, which also enables stability constraints to be incorporated in a natural manner, and the re-formulation of the objective function, which streamlines the convergence of the GA by two to three orders of magnitude. Following initial verification in simulations, the GA has been used to identify the closed-loop dynamics of a stand-alone ball screw drive and the x- and y-axes of a five-axis machining center. In experimental results, it is shown that drive models constructed using the GA can be successfully used to predict the tracking and contouring errors for different part programs in a virtual process planning environment.