Fuzzy neuron adaptive modeling to predict surface roughness under process variations in CNC turning

Abstract

Many surface roughness control models for turning operations are based on experimental data models that are often expensive and time consuming to construct. Empirical models are limited to domain-specific applications and are very sensitive to process variations that may occur during the implementation stages. Over time, a machining process output may drift due to changes in a machine tool structure and the environment. Empirical models have no means of compensating for changes in the process or responding to process variations. In this study, a fuzzy adaptive modeling technique, which adapts the membership functions in accordance with the magnitude of the process variations, is used to predict surface roughness. This approach can be viewed as a way to expand the domain of a cutting experiment. Test results show good agreement between the actual process output and the predicted surface roughness. Further development of the method may lead to an improved automated process control for machining operations.