Study On Tool Wear Classification Using Support Vector Machines In A Milling Process

A milling cutter is an important tool that use in milling process to remove material from the surface of the workpiece.There are many different types of milling cutting tools are used in milling process,so cutting tools are subjected to frequent failures during machining,the failures of cutting tools are due to chip formation,wear and chattering.Tool wear can be measured by different methods such as grooving and indentation method,optical microscope fitted with micrometer,Scanning electron microscope(SEM).A situation or conditions of a machine is very important for maintenance engineers,that can be used to estimate the chance of recovery a breakdown or to prevent possible failure of a machine.Once sudden failure occurs the maintenance cost and the required equipment by unscheduled downtime are always huge.The machine running with tool wear or tool breakage will lead the problem,machine running in vibration condition and can result in damage to the workpiece.Therefore,in milling process requires effective and reliable cutting tools performance degradation identification to detect faults at the early stage.By collecting data of vibration and noise as well from sensors mounted on the spindle of milling machine,these data would be important information to analyze for tools degradation identification.The main objective of this research is to identify the condition of the tools to avoid machine failure caused by the breakage or wear out of the cutting tools.The condition of the milling machine cutting tool(performance degradation identification)is used to prevent damages on the cutting tool.They help in finding faults at an early stage so that diagnosis and correction of those abnormalities can be initiated before they become more pronounced and lead to a loss in productivity.In most case,the collected raw data cannot use directly.Therefore,we have to find a way to get the most useful data,this process we usually called data processing.The acquired vibration data will plot in time and frequency domain to see the characteristic of the signal.Then we obtained 16 features by using wavelet packet transform in the purpose of tool wear classification.Some of the received features are too noisy leading poor classification and some of them are better to use.If we use all of them it will increase the computational time,every researcher will not choose that way.Hence,every feature needs to pass through the step of features selection.By using Fisher's discriminant ratio(FDR)analysis to get significant detail of the features with less affected by noise and other variations.Finally,the prominent features were selected,Support Vector Machine approach would be used to classify tool wear degradation into 5 states(I)initial wear,(II)normal wear,(III)breakage wear,(IV)semi-wear and(V)tool failure state.Support Vector Machine is a prestige methodology for solving problems.it can be used in nonlinear classification function estimation and density estimation that has also led to many other recent developments in kernel-based methods.Hence,by using Support Vector Machine(SVM)tool wear information of the machine would be provided.Worn tools can be changed in time to reduce waste product and tools cost noticeably.It is even possible to guarantee a certain surface quality.The result shows that the classification accuracy between Energy sub-band 3 with Energy sub-band 8 using SVM with the polynomial kernel is 94.00% better than others.The accuracy obtained is acceptable for industrial applications.