Control chart pattern recognition using K-MICA clustering and neural networks

Automatic recognition of abnormal patterns in control charts has seen increasing demands nowadays in manufacturing processes. This paper presents a novel hybrid intelligent method (HIM) for recognition of the common types of control chart pattern (CCP). The proposed method includes two main modules: a clustering module and a classifier module. In the clustering module, the input data is first clustered by a new technique. This technique is a suitable combination of the modified imperialist competitive algorithm (MICA) and the K-means algorithm. Then the Euclidean distance of each pattern is computed from the determined clusters. The classifier module determines the membership of the patterns using the computed distance. In this module, several neural networks, such as the multilayer perceptron, probabilistic neural networks, and the radial basis function neural networks, are investigated. Using the experimental study, we choose the best classifier in order to recognize the CCPs. Simulation results show that a high recognition accuracy, about 99.65%, is achieved.     

Control chart pattern recognition using an optimized neural network and efficient features

Automatic recognition of abnormal patterns in control charts has seen increasing demands nowadays in manufacturing processes. This study investigates the design of an accurate system for control chart pattern (CCP) recognition from two aspects. First, an efficient system is introduced that includes two main modules: the feature extraction module and the classifier module. The feature extraction module uses the entropies of the wavelet packets. These are applied for the first time in this area. In the classifier module several neural networks, such as the multilayer perceptron and radial basis function, are investigated. Using an experimental study, we choose the best classifier in order to recognize the CCPs. Second, we propose a hybrid heuristic recognition system based on particle swarm optimization to improve the generalization performance of the classifier. The results obtained clearly confirm that further improvements in terms of recognition accuracy can be achieved by the proposed recognition system.     

基于MFCC参数和VQ的说话人识别系统

采用能够反映人对语音的感知特性的Mel频率倒谱系数(MFCC)作为特征参数,以及为避免时间规整问题采用矢量量化技术开发的说话人识别系统.MFCC主要的是模拟人耳的听觉过程,相对于其它参数它对语音波形的变化不敏感,更加稳定,系统取得很好的识别结果,实验表明系统训练和识别的计算量和存储量都比较低.     

应用动态时间规整与矢量量化的语音识别算法

提出了一种基于动态时间规整(DTW)的改进平均最小距离识别算法,改善了孤立词识别的鲁棒性并提高了识别率。同时对矢量量化(VQ)算法分析了不同码本大小下的识别率,并比较了各种算法的运算时间。通过在MatLab上实现特定人孤立词小词汇量语音识别,实验的结果表明:基于DTW算法的改进平均最小距离法识别率显著提高;码本较大时VQ算法的识别率最高;VQ算法的识别率一般高于DTW算法且运行时间短。     

Control chart pattern recognition using RBF neural network with new training algorithm and practical features

The control chart patterns are the most commonly used statistical process control (SPC) tools to monitor process changes. When a control chart produces an out-of-control signal, this means that the process has been changed. In this study, a new method based on optimized radial basis function neural network (RBFNN) is proposed for control chart patterns (CCPs) recognition. The proposed method consists of four main modules: feature extraction, feature selection, classification and learning algorithm. In the feature extraction module, shape and statistical features are used. Recently, various shape and statistical features have been presented for the CCPs recognition. In the feature selection module, the association rules (AR) method has been employed to select the best set of the shape and statistical features. In the classifier section, RBFNN is used and finally, in RBFNN, learning algorithm has a high impact on the network performance. Therefore, a new learning algorithm based on the bees algorithm has been used in the learning module. Most studies have considered only six patterns: Normal, Cyclic, Increasing Trend, Decreasing Trend, Upward Shift and Downward Shift. Since three patterns namely Normal, Stratification, and Systematic are very similar to each other and distinguishing them is very difficult, in most studies Stratification and Systematic have not been considered. Regarding to the continuous monitoring and control over the production process and the exact type detection of the problem encountered during the production process, eight patterns have been investigated in this study. The proposed method is tested on a dataset containing 1600 samples (200 samples from each pattern) and the results showed that the proposed method has a very good performance.     

Force control of hybrid actuator using learning vector quantization neural network

Hydraulic actuators are important in modern industry due to high power, fast response, and high stiffness. In recent years, hybrid actuation system, which combines electric and hydraulic technology in a compact unit, can be adapted to a wide variety of force, speed and torque requirements. Moreover, the hybrid actuation system has dealt with the energy consumption and noise problem existed in the conventional hydraulic system. Therefore, hybrid actuator has a wide range of application fields such as plastic injection-molding and metal forming technology, where force or pressure control is the most important technology. In this paper, the solution for force control of hybrid system is presented. However, some limitations still exist such as deterioration of the performance of transient response due to the variable environment stiffness. Therefore, intelligent switching control using Learning Vector Quantization Neural Network (LVQNN) is newly proposed in this paper in order to overcome these limitations. Experiments are carried out to evaluate the effectiveness of the proposed algorithm with large variation of stiffness of external environment. In addition, it is understood that the new system has energy saving effect even though it has almost the same response as that of valve controlled system.     

Bayesian spatial defect pattern recognition in semiconductor fabrication using support vector clustering

Defects generated during integrated circuit (IC) fabrication processes are classified into global defects and local defects according to their generation causes. Spatial patterns of locally clustered defects are likely to contain the information related to their defect generation mechanisms. In this paper, we propose a model-based clustering for spatial patterns of local defects to reflect real situations. A flexible two-step approach is proposed to classify the spatial defects patterns via support vector clustering and Bayesian method. Support vector clustering is employed to separate global defects from the local ones to improve both clustering accuracy and computational efficiency in further analysis. A new mixture model is proposed for modeling the distribution of local defects on the wafers. Local defect clusters with amorphous/linear, curvilinear, and ring patterns are modeled by multivariate normal distribution, principal curve, and spherical shell, respectively. A Bayesian inference procedure is then applied for parametric pattern recognition of the local defects. Results from both simulated data and real wafer map data demonstrate potential in applying our approach to analyze general defect patterns in IC manufacturing.     

Identification of ECG beats from cross-spectrum information aided learning vector quantization

This work describes the development of a computerized medical diagnostic tool for heart beat categorization. The main objective is to achieve an accurate, timely detection of cardiac arrhythmia for providing appropriate medical attention to a patient. The proposed scheme employs a feature extractor coupled with an Artificial Neural Network (ANN) classifier. The feature extractor is based on cross-correlation approach, utilizing the cross-spectral density information in frequency domain. The ANN classifier uses a Learning Vector Quantization (LVQ) scheme which classifies the ECG beats into three categories: normal beats, Premature Ventricular Contraction (PVC) beats and other beats. To demonstrate the generalization capability of the scheme, this classifier is developed utilizing a small training dataset and then tested with a large testing dataset. Our proposed scheme was employed for 40 benchmark ECG files of the MIT/BIH database. The system could produce classification accuracy as high as 95.24% and could outperform several competing algorithms.     

基于实例的学习矢量量化神经网络诊断方法及其应用

基于CBR良好的可扩充性、可移植性和神经网络极强的分类能力,提出了基于实例的学习矢量量化神经网络诊断方法。该方法应用于机械故障诊断系统中,可以减小实例搜索空间,提高实例检索效率。论述了系统的设计方法和应用步骤。     

A study on the various features for effective control chart pattern recognition

The identification of various unnatural patterns that are usually exhibited in quality control charts leads to more focussed diagnosis and, thereby, significantly minimises the effort towards effective troubleshooting. Feature-based control chart pattern (CCP) recognition systems have the advantage that the users can easily understand how a particular pattern is identified. Pham and Wani have presented a feature-based heuristic approach for CCP recognition which can differentiate six types of CCPs, based on the extraction of nine shape features. The extraction of some of these features requires users’ inputs and, thus, this CCP recognition system is not truly automated. Moreover, many real-life situations require detection of all of the eight basic CCPs, including stratification and systematic patterns. The extraction of the features after the scaling of pattern data into an (0, 1) interval can ensure that the magnitudes of the features are independent of the mean and standard deviation of the underlying process. But the distinction between normal and stratification patterns is lost when the pattern data are scaled. A CCP recogniser that will identify a stratification pattern, therefore, needs to be developed using unscaled pattern data. In this paper, potentially useful 32 features, the extraction of which do not require users’ inputs of any form, are proposed. The magnitudes of these features are independent of the process mean and are considerably insensitive to changes in the process standard deviation. An easy mechanism for the determination of the optimal set of features and a heuristic is also presented. Sensitivity studies on the performance of the heuristic show that it is robust against the estimation error of the process mean and is reasonably robust against the estimation error of the process standard deviation. Thus, it has enough potential for use in real-time process control applications.     

Pattern recognition for bivariate process mean shifts using feature-based artificial neural network

In multivariate quality control, the artificial neural networks (ANN)-based pattern recognition schemes generally performed better for monitoring bivariate process mean shifts and provided more efficient information for diagnosing the source variable(s) compared to the traditional multivariate statistical process control charting. However, these schemes revealed disadvantages in term of reference bivariate patterns in identifying the joint effect and excess false alarms in identifying stable process condition. In this study, feature-based ANN scheme was investigated for recognizing bivariate correlated patterns. Feature-based input representation was utilized into an ANN training and testing towards strengthening discrimination capability between bivariate normal and bivariate mean shift patterns. Besides indicating an effective diagnosis capability in dealing with low correlation bivariate patterns, the proposed scheme promotes a smaller network size and better monitoring capability as compared to the raw data-based ANN scheme.     

Intelligent switching control of pneumatic cylinders by learning vector quantization neural network

School of Mechanical and Automotive Engineering, University of Ulsan, San 29, Muger2-dong, Nam-gu, Uhan 680-749, Korea The development of a fast, accuiate, and inexpensive position-controlled pneumatic actuator that may be applied to various practical positioning applications with various external loads is described in this paper A novel modified pulse-width modulation (MPWM) valve pulsing algorithm allows on/off solenoid valves to be used in place of costly servo valves A comparison between the system response of the standardPWM technique and that of the modified PWM technique shows that the performance of the proposed technique was significantly increased A state-feedback controller with position, velocity and acceleration feedback was successfully implemented as a continuous controllei A switching algorithm foi control parameters using a learning vector quantization neural network (LVQNN) has newly proposed, which classifies the external load of the pneumatic actuator The effectiveness of this proposed control algorithm with smooth switching control has been demonstrated thiough experiments with various external loads     

Improved recognition of control chart patterns using artificial neural networks

Recognition of abnormal patterns in control charts provides clues to reveal potential quality problems in the manufacturing processes. One potentially popular approach for recognizing different control chart patterns (CCPs) is to develop heuristics based on various shape features of the patterns. The advantage of this approach is that the users can easily understand how a particular pattern is identified. However, consistency in the recognition performance is found to be considerably poor in the heuristics approach. Since shape features represent the main characteristics of the patterns in a condensed form, artificial neural network (ANN) with features extracted from the process data as input vector representation can facilitate efficient pattern recognition with a smaller network size. In this paper, a set of seven shape features is selected, whose magnitudes are independent of the process mean and standard deviation under a special representation of the sampling interval in the control chart plot. Based on these features, the CCPs are recognized using a multilayered perceptron neural network trained by back-propagation algorithm. The recognizer can recognize all the eight commonly observed CCPs. Extensive performance evaluation of this recognizer is carried out using simulated pattern data. Numerical results indicate that the developed ANN recognizer can perform well in real time process control applications with respect to both recognition accuracy and consistency.     

Confocal microscopic image sequence compression using vector quantization and three-dimensional pyramid

The three-dimensional (3-D) pyramid compressor project at the University of Glasgow has developed a compressor for images obtained from confocal laser scanning microscopy (CLSM) device. The proposed method using a combination of image pyramid coder and vector quantization techniques has good performance at compressing confocal volume image data. An experiment was conducted on several kinds of CLSM data using the presented compressor compared with other well-known volume data compressors, such as MPEG-1. The results showed that the 3-D pyramid compressor gave a higher subjective and objective image quality of reconstructed images at the same compression ratio, and presented more acceptable results when applying image processing filters on reconstructed images.     

A novel approach for recognition of control chart patterns: Type-2 fuzzy clustering optimized support vector machine

Unnatural patterns in the control charts can be associated with a specific set of assignable causes for process variation. Hence, pattern recognition is very useful in identifying the process problems. In this study, a multiclass SVM (SVM) based classifier is proposed because of the promising generalization capability of support vector machines. In the proposed method type-2 fuzzy c-means (T2FCM) clustering algorithm is used to make a SVM system more effective. The fuzzy support vector machine classifier suggested in this paper is composed of three main sub-networks: fuzzy classifier sub-network, SVM sub-network and optimization sub-network. In SVM training, the hyper-parameters plays a very important role in its recognition accuracy. Therefore, cuckoo optimization algorithm (COA) is proposed for selecting appropriate parameters of the classifier. Simulation results showed that the proposed system has very high recognition accuracy.     

Identifying change point of a non-random pattern on control chart using artificial neural networks

An important step in root cause analysis is the identification of the time when process first changed. The time when a disturbance first manifested itself into the process is referred to as change point. Identification of the change point could help process engineer to perform root cause analysis effectively. In this paper, an estimator for the change point of a normal process mean using artificial neural network (ANN) is proposed. Five patterns of change namely single step, linear trend, systematic, cyclic, and mixture are studied. Whenever possible, results are compared numerically to the results obtained by other methods proposed by different researchers. First the type of change to be recognized by an ANN-based pattern recognizer is identified and then the change point in the process mean is estimated. Results indicate satisfactory performance for the proposed method that could be used as an effective method for root cause analysis by process engineer.     

A novel sensor selection using pattern recognition in electronic nose

A novel sensor selection using pattern recognition technique in electronic nose (E-Nose) is proposed in this paper. This paper studies the portable E-Nose based on metal oxide semiconductor (MOS) gas sensors for detection of multiple kinds of indoor air contaminants. The characteristics of portability, low cost, multiple targets detection and high performance of E-Nose monitor are the main pursuit for home use. Formaldehyde, benzene, toluene, carbon monoxide, and ammonia are the primary targets of the proposed E-Nose which benefits from the characteristics of the broad spectrum, reproducibility, sensitivity and low-cost of MOS gas sensors. Therefore, a potential and full contribution analysis of the small sized sensor array, in detection of indoor air contaminants coupled with a kernel principal component analysis (KPCA) based linear discriminant analysis (LDA) pattern recognition technique, is presented in this paper. Some experimental findings on the roles of sensors in an E-Nose have also been concluded. The recognition results clearly demonstrate the contribution of each sensor to gas detection which helps the sensor selection in E-Nose design.     

Online qualitative nugget classification by using a linear vector quantization neural network for resistance spot welding

Real-time estimation of weld quality from process data is one of the key objectives in current weld control systems for resistance spot-welding processes. This task can be alleviated if the weld controller is equipped with a voltage sensor in the secondary circuit. Replacing the goal of quantifying the weld quality in terms of button size by the more modest objective of indirect estimation of the class of the weld, e.g., satisfactory (acceptable, “normal” button size), unsatisfactory (undersized, “cold” welds), and defects (“expulsion”), further improves the feasibility of the mission of indirect estimation of the weld quality. This paper proposes an algorithmic framework based on a linear vector quantization (LVQ) neural network for estimation of the button size class based on a small number of dynamic resistance patterns for cold, normal, and expulsion welds that are collected during the stabilization process. Nugget quality classification by using an LVQ network was tested on two types of controllers; medium-frequency direct current (MFDC) with constant current controller and alternating current (AC) with constant heat controller. In order to reduce the dimensionality of the input data vector, different sets of features are extracted from the dynamic resistance profile and are compared by using power of the test criteria. The results from all of these investigations are very promising and are reported here in detail.     

An inexpensive system for classifying tool wear states using pattern recognition

The purpose of this research work is to develop an inexpensive model tool wear sensing system using pattern recognition. Accordingly, the combined output of radial force, feed force and acoustic emission (r.m.s. value) is utilized to model the tool flank wear in a turning operation. The tool wear sensing system consists of two phases: training and classification. The training phase is done off-line and is used to determine the weight coefficients for the linear decision functions using the prototype patterns from the cutting tests. The classification phase is in real time. In the first stage of the classification phase, the minimum distance classifier selects a prototype (conditions already trained) cutting test that is closest to the cutting test to be performed. The linear decision functions of the prototype test selected are used for classifying the incoming signal of the actual cutting test into one of three wear classes.

The success rate of training for various tests varied between 39.57% and 100%. The success rate of classifying signals from actual tests was also encouraging, demonstrating that the proposed methodology can be successfully applied to predict the status of the cutting tool on-line using low budget equipment.