Research On Data Collection Analysis And Surface Quality Prediction In Ultra-precision Machining Process

In the ultra-precision machining process,the machine tool state,cutting process parameters,material performance and tool wear may all have a great impact on the surface roughness,and how to predict the machining surface roughness to guide industrial production is a hotspot and difficulty of advanced manufacturing technology.This paper investigates the domestic and foreign research on the prediction of the surface roughness of the machining,and detailed analysis of the factors affecting the surface roughness of the processing.Starting from the technological characteristics of ultraprecision machining and the factors affecting the roughness of the processed surface,an experimental plan for ultra-precision machining was designed,and a big data acquisition platform was established during the machining process.and an acceleration sensor was used to collect the vibration signal of the machine tool spindle,The force signal of the tool was collected with a dynamometer,and the surface roughness of the sample was measured with a white light interferometer.In the processing signal processing,the signal segmentation technology based on short-term energy was used to cut the signal,and the multi-dimensional characteristics of the signal were extracted through time domain analysis and frequency domain analysis,and the signal was decomposed into three layers using wavelet packet decomposition technology,the time-frequency characteristics of the signal are constructed with the energy proportion of each frequency band.In view of the characteristics of small signal sample size and high dimension,this paper introduces support vector machine and different feature dimensionality reduction algorithms,compares the classification prediction effect of support vector machine on surface roughness when using different kernel functions,and uses Grid search for parameter optimization.Finally,the classification accuracy of the support vector machine after using different dimensionality reduction algorithms to reduce the dimensionality of the sample data is compared,and the prediction effect of the multiclass prediction model is analyzed and evaluated.In this paper,by designing ultra-precision machining and signal acquisition experiments,to avoid considering the actual complex machining conditions,using the machining signals to extract high-dimensional features,the surface roughness of the machining is classified and predicted,and good prediction accuracy is effectively obtained.This study has important guiding significance for the accurate prediction of surface roughness in the ultra-precision processing and manufacturing of optical components.