Prediction of surface roughness in single-point diamond turning by combining machine tool internal signals and deep learning method

Abstract

This paper presents a novel surface roughness prediction model for the single-point diamond turning (SPDT) process which takes into account the relative vibration between the cutting tool and workpiece as well as the material swelling effect. In previous prediction models, the relative vibration between the cutting tool and the workpiece (or the relative tool-work vibration) was commonly measured prior to the machining operation and simplified into a superposition of multiple steady simple harmonic motions, neglecting its variations throughout the machining process and resulting in the loss of significant details. Moreover, few models have simultaneously considered the impact of the relative tool-work vibration and material properties, both of which substantially influence surface roughness. In this study, a signal acquisition system is developed to collect the machine tool internal signals that can accurately reflect the varying vibration states during the cutting process. A dynamic volumetric topography simulation model is established to realize in-process surface generation simulation using the collected internal signals. To consider the influence of material properties, a deep Long Short-Term Memory (LSTM) network is built to estimate the changes in surface topography caused by the material swelling effect. Experiments are conducted to validate the proposed surface roughness prediction model, and the results demonstrate that it can stably and accurately predict surface roughness with relative prediction errors lower than 10 % for all test experiments. Moreover, the dynamic characteristic of the model also provides the capability for conducting online quality and anomaly monitoring tasks.