In turning, surface integrity depends on numerous interrelated factors, including cutting parameters, tool wear, and process dynamics. Monitoring these manufacturing processes using sensors and intervening promptly in unstable conditions can help maintain process stability. However, conventional process monitoring quickly reaches its limits, especially when there are few or no options to identify potential disturbances and detect relevant features in sensor signals. For example, in single-part and small-batch production, there is often not enough time to collect and analyze the necessary data. Artificial intelligence offers the potential to overcome the current limitations in process monitoring by enabling reliable feature detection in machining processes in a time- and cost-efficient manner. Data-driven models provide an effective approach for analyzing the acquired data in-process and predicting parameters associated with surface integrity and process stability. In this research, a deep learning model is developed to organize the temporal data, like time step and sensor signals, along with the non-temporal data, like cutting depth, cutting speed, and feed rate, to predict the surface roughness of the final workpiece. Based on the results, the developed model with Long Short-Term Memory network (LSTMs) represents a promising approach to provide practical guidance for the surface roughness prediction in real industrial settings.
Lerez et al. (Thu,) studied this question.