Condition Monitoring of Face Milling Tool Using Vibration and Sound Signals

Abstract

Fault diagnosis of the cutting tool is very essential for improving the quality and maintaining the accurate dimension of the products during machining process. The milling is a multi-toothed metal removing process. In face milling, because of dynamic variation of cutting forces, thermo-mechanical shocks and vibration, which results in catastrophic tool failure along with gradual wear of the tool inserts. Wear development during machining can reach up to unacceptable level, resulting in inaccurate dimension and poor surface finish of the components. Monitoring the condition of the cutting tool during face milling operation is a vital role before the tool causes any damage on the machined surface which becomes highly valuable in order to avoid loss of products, damage to the machine tool and associated loss in productivity. Keeping in view of the automation, it is necessary to choose an effective and efficient method for monitoring the cutting tool condition without affecting the machining setup and the work material. This study mainly deals with the fault diagnosis of the face milling tool using vibration and sound signals through signal processing techniques and machine learning approach. The face milling is a machining process with an intermittent cutting action. The milling tool will undergo different types of faults such as flank wear, breakage and chipping which occurs predominantly during milling. The vibration and sound signals under these faulty and healthy milling tool conditions are acquired and these signals are further analyzed. Current research work is mainly categorized into two phases. The first phase is to detect/diagnose the face milling tool conditions by analyzing the vibration and sound signals using signal processing techniques. The signal processing techniques such as time-domain analysis, spectrum analysis, cepstrum analysis and continuous wavelet transform (CWT) method are applied to recognize the face milling tool conditions. The cepstrum analysis has been applied for the first time in fault detection of the face milling tool and has provided the sufficient information about the face milling tool condition using both vibration and sound signals. Generally conventional data processing is computed in time or frequency domain which is not suitable for analyzing non-stationary signals. In order to overcome the lack of a global view on how to develop machining monitoring systems based on artificial intelligent models, machine learning approach is one of the best methods for developing an effective tool condition monitoring (TCM) system.In the second phase, fault diagnosis studies of the face milling tool using vibration and sound signals based on artificial intelligence techniques are conducted. Fault diagnosis of the different tool conditions based on machine learning technique is basically comprised of three steps; feature extraction, feature selection and feature classification. Different features such as, statistical features, histogram features, discrete wavelet transform (DWT) features and empirical mode decomposition (EMD) features are extracted from the acquired vibration and sound signals. For example, features such as skewness, mode, standard error, maximum, minimum, range, sum, mean, standard deviation, median, sample variance and kurtosis are computed from each acquired vibration and sound signals will serve as statistical features. The important features out of all extracted features are to be selected using induction based on decision tree technique (ID3 algorithm or J48 algorithm). The artificial intelligence techniques such as support vector machine (SVM), Naïve Bayes algorithm, artificial neural network (ANN), decision tree algorithm and K-star algorithm are used to classify the data using selected features. Fault diagnosis analysis with acquired vibration and sound signals are carried out by making use of different combinations of feature extraction methods and different classifiers with selected features based on decision tree technique. Overall results have shown that the vibration signal based fault diagnosis has given better classification accuracy than the sound signal based fault diagnosis. The current research work has demonstrated that the statistical features served as best features among all other features extracted such as, EMD features, Histogram features and DWT features. It is also found that the Naïve Bayes algorithm provides best classification accuracy in comparison with other classifiers used such as SVM, ANN, decision tree and K-star algorithm. Based on research work, it is proposed that the combination of statistical features and the Naïve Bayes algorithm as classifier is the best feature-classifier pair using vibration signals in tool condition monitoring system for the face milling process.