Positional, geometrical, and thermal errors compensation by tool path modification using three methods of regression, neural networks, and fuzzy logic

One of the main sources of inaccuracy in production is machine tool errors. In this study, a method is introduced to model and compensate positional, geometrical, and thermally induced errors of machine tools by an offline technique. Thermal errors are modeled by three ways of multiple linear regression, artificial neural networks, and neuro-fuzzy modeling. The required database is provided by measuring errors using a laser interferometer. Subsequently, the models are evaluated and the best one which is neuro-fuzzy with a mean square error of 0.375 μm is chosen to predict the errors using developed software. The experimental procedure is optimized by a preliminary broad assessment. Volumetric errors are calculated using rigid body kinematic and applied to modify the initial G-codes. The introduced method is validated by compensating errors on a free-form which shows significant average improvement of errors.