Chapter 5 A Review of the Implementation of Tool Condition
5.4 Signal processing methods
The most common approaches regarding indirect methods of cutter tool monitoring are analysis of accelerations signal, dynamic forces and acoustic emissions. Fast Fourier Transformation (FFT) is widely used in order to present cutter tool wear or tool fault in the frequency domain.
If the FFT is taken into account, the second harmonic is an indicator of tool wear estimation. Another approach uses the increase of energy in the frequency domain as an indicator of cutter tool conditions. However, a question arises as to whether a
change of cutter tool geometry is as a result of wear or as a tool fault that can be observed in the frequency domain.
The limitation of FFT consists in the impossibility of processing non-linear and non- stationary data [101]. Since the Fourier transform approach has certain serious theoretical drawbacks in processing machining signals. It is the integration for all times. This fact makes it difficult to analyse any local property of the signal. Another shortcoming of the FFT is presentation of results only in frequency domain. However, the manufacturing process is described as a non-linear and non-stationary process. The signal processing methods used to analyse non-stationary signals are appropriate for cutting process monitoring. Therefore, reference [133] studied the relation between cutter tool wear and acceleration signal in frequency and time- frequency domain using a new method, Hilbert–Huang Transform (HHT) which presented data locally without harmonics. The idea is processed the data by short- time Fourier transform the cutter tool wear or tool fault is detected by increasing the power in the power spectral density. While by using HHT, the acceleration signals change the frequency in the marginal spectra as a result of geometric change of the cutter tool. Also, it is applied to the cutter tool wear and tool fault monitoring and compared to the FFT.
5.4.2 Wavelets Transformation (WT)
The reliability and applicability of tool breakage detection to assist in advancing high availability levels of sophisticated manufacturing systems, in conjunction with high quality levels of manufactured components, are considered in the resent research. In order to improve robustness of the tool from wear and breakage, the signal processing method of spectral analysis is the most commonly used technique in tool breakage detection. Yet, although it is resolution is good in the frequency domain, it has an inadequate time domain resolution. Also some signal information in time domain is lost in the spectral analysis process. The wavelet transforms (WT) which is localized in both time and frequency to detect a small change in the input signals. In addition, it requires less computation than FFT.
Continuous wavelet transformer is recognised as effective tools for both stationary and non-stationary signals. However, much of the information is superfluous and computationally very slow [134]. Discrete wavelet transform (DWT) uses an
frequency and time domains so that it can extract more information, which can be used to analyse tool breakage monitoring signals [135]. Reference [136] presented an effective algorithm for tool breakage monitoring system based on DWT of an acoustic emission (AE) and an electric feed current signal. The experiment results show overall 98.5% reliability and the good capability of real-time monitoring of the proposed for detecting tool breakage during machining process.
Among many machining condition monitoring systems, a spindle motor power monitoring system is considered as one of the most popular systems for plant floor applications. However, in practice, power signals are mixed with many signal sources relevant to cutting tool, which contaminate with each other in feature extraction processes and decrease the monitoring reliability. Reference [137] presented a new method based on the wavelet transform for the detection of tool damage. It is assumed that the vibration signal of the original structure of tool without any defect is already known. When the defects are presented, the vibration signals of the defected tool are then recorded. After comparing the DWTs of these two sets of vibration signals in the space domain, it could be used to detect the presence of defects; their number and location.
5.4.3 Principal Component Analysis (PCA)
Principal Component Analysis (PCA) is technique of identifying patterns in the correlated data, and expressing the data to highlight their similarities and differences. The main advantage of PCA is that once the patterns in data have been identified, the data can be compressed, i.e. by reducing the number of dimensions, without much loss of information. The methods involved in PCA are discussed below [138]:
1. Getting some data 2. Normalization of data
3. Calculation of covariance matrix. 4. Interpretation of covariance matrix.
Reference [139] proposed a signal processing method used on PCA and wavelet analysis, aiming to reduce the dimension of the data and obtain both frequency and time localisation information which could help to find abnormal phenomenon quickly and orient the position and the time of faults exactly in the complex textile
machinery systems. At first, the original signals are simplified by principal component transform, which was conducted by calculating the eigenvalue and eigenvector of correlation coefficient matrix, and by defining the first few Principal Components (PCs) containing most of the variables according to the contribution and cumulative contribution rates. Secondly, the restructured signals are decomposed into approximate and detailed ones for obtaining meaningful captures of instantaneous frequency by wavelet analysis. From practical application, this signal processing method was validated.
In addition, PCA is used for fault diagnosis based on different sensors. For example, the basic theory of principal component analysis and its basic procedures for fault detection are introduced the sound signal pre-processing is depicted, multi-domain feature vector is extracted from time, time-frequency and frequency domain, faults are diagnosed with principal component analysis method [140].
In the current research, the PCA is used to design an effective fusion model to detect the faults of tool and fixturing system.