Prediction of Surface Integrity Characteristics of Alloy 718 when Machining

Abstract

High Efficiency Milling (HEM) is widely used in aerospace industry for machining difficult-to-cut materials like Alloy 718. HEM increases the productivity and efficiency of the machining process. Generally, the high cutting speeds and feed rates as well as large axial depth of cuts are used in HEM to improve productivity while the radial depth of cut is reduced to decrease the thermo-mechanical loads on the cutting edges and to provide chip-thinning effects. However, the changes in these machining parameters have direct impacts on the surface quality of the machined parts. In addition, the tool wear during the milling process changes the distribution of stress and temperature on the machined surfaces. Monitoring and assessment of the surface integrity in terms of residual stress, roughness, and sub-layer deformation when machining the aerospace components is of vital importance, as it significantly affects the fatigue performance of machined parts. This investigation aims to predict and analyze the surface integrity of Alloy 718 workpieces fabricated by HEM process using Machine Learning (ML). The milling tests were performed at various cutting conditions using a ceramic tool for roughing, followed by a finishing process using the cemented carbide tool. Surface roughness was measured using an optical measurement system and the tool wear was examined by stereo optical microscopy. The subsurface of the machined surface was examined using optical and scanning electron microscopy. The residual stresses and intensity of surface deformation were also measured using X-ray diffraction techniques. Additionally, a sensory tool was used to monitor the cutting stability during the milling process. Tool wear and cutting speeds from the roughing stage were found to have a substantial effect on the surface integrity of the finished part. However, this investigation showed that the desirable surface integrity parameters can be attained if the parameters for the rough machining and finishing process are optimized and carefully selected. Principal Component Analysis (PCA) was used for the multi-dimensional reduction. The k-means clustering method was used for classification of the surfaces when the unworn and worn ceramic tools were used for machining. The next stage consisted of creating a model capable of predicting the cluster to which the new surface would belong using the machining parameters and the stability criterion signal as predictors. For this, a k-nearest neighbor (kNN) model was used which classifies a new surface as the same cluster as the nearest known examples. A commonly-used metric to evaluate the model accuracy is the misclassification rate. Depending on the observations were partitioned into training and testing sets, the prediction accuracy ranged from 54.7% to 74.9%. This approach can be considered as an effective way to predict the surface integrity characteristics at the current stage of research.