Yeou-Ren Shiue

On-line identification of control chart patterns using self-organizing approaches

Effective identification of unnatural control chart patterns (CCPs) is an important issue in statistical process control, as unnatural CCPs can be associated with specific assignable causes adversely affecting the process. The intention of this paper is to develop an automatic CCP identification system using self-organizing approaches—neural network and decision tree (DT) learning. Recently, back-propagation networks (BPNs) have been widely used in the research field of CCP identification. However, one of the major limitations of conventional BPN is in dealing with dynamic patterns that vary over time, such as CCPs. This limitation is one of the major reasons for the false classification problem commonly encountered in the BPN-based CCP identification schemes in the literature. A time-lagging input algorithm is proposed in this research to enhance the performances of the BPN-based CCP identifiers. Additionally, DT learning is employed as a novel approach to the CCP identification problem. The simulation experiments demonstrate that both the BPN-based system with time-lagging input and the DT-based system perform better than the conventional BPN-based system in terms of identification accuracy and speed. The proposed time-lagging input algorithm can greatly improve the identification speed and stability of the BPN-based CCP identifier. Besides, the empirical comparison indicates that the DT-based system outperforms the BPN-based system with respect to classification capability in an on-line CCP identification scheme. Moreover, the learning time of the DT-based system is much shorter than that of the BPN-based system.

An effective application of decision tree learning for on-line detection of mean shifts in multivariate control charts

With modern data acquisition system and on-line computer used during production, it is now common to monitor several correlated quality variables simultaneously. Various multivariate control charts (e.g., Hotellings T^2, multivariate cumulative sum, and multivariate exponentially weighted moving average charts) have been designed for detecting mean shifts. The main problem of such charts is that they can detect an out-of-control event but do not directly determine which variable or group of variables has caused the out-of-control signal. Using decision tree learning techniques, this work proposes a straightforward and effective model to detect the mean shifts in multivariate control charts. Experimental results using simulation show that the proposed model cannot only efficiently detect the mean shifts but also accurately identify the variables that have deviated from their original means. The shift direction of each of the deviated variables can also be determined in the meantime. The experimental results also indicate that the learning speed of the proposed decision tree learning-based model is much faster (25 times in this research) than that of a neural network-based model (another machine learning-based approach) for detecting mean shifts in multivariate control charts. The feature of fast learning makes the proposed DT learning-based model more adaptable to a dynamic process-monitoring scenario, in which constant model re-designing and re-learning is required. A bivariate case of the proposed multivariate model is presented. A demonstrative application is provided to illustrate the usage of the proposed decision tree learning-based approach to multivariate quality control.

Data-mining-based dynamic dispatching rule selection mechanism for shop floor control systems using a support vector machine approach

The purpose of this paper is to develop a data-mining-based dynamic dispatching rule selection mechanism for a shop floor control system to make real-time scheduling decisions. In data mining processes, data transformations (including data normalisation and feature selection) and data mining algorithms greatly influence the predictive accuracy of data mining tasks. Here, the z-scores data normalisation mechanism and genetic-algorithm-based feature selection mechanism are used for data transformation tasks, then support vector machines (SVMs) is applied for the dynamic dispatching rule selection classifier. The simulation experiments demonstrate that the proposed data-mining-based approach is more generalisable than approaches that do not employ a data-mining-based approach, in terms of accurately assigning the best dispatching strategy for the next scheduling period. Moreover, the proposed SVM classifier using the data-mining-based approach yields a better system performance than obtained with a classical SVM-based dynamic dispatching rule selection mechanism and heuristic individual dispatching rules under various performance criteria over a long period.

Effective recognition of control chart patterns in autocorrelated data using a support vector machine based approach

The effective recognition of unnatural control chart patterns (CCPs) is a critical issue in statistical process control, as unnatural CCPs can be associated with specific assignable causes adversely affecting the process. Machine learning techniques, such as artificial neural networks (ANNs), have been widely used in the research field of CCP recognition. However, ANN approaches can easily overfit the training data, producing models that can suffer from the difficulty of generalization. This causes a pattern misclassification problem when the training examples contain a high level of background noise (common cause variation). Support vector machines (SVMs) embody the structural risk minimization, which has been shown to be superior to the traditional empirical risk minimization principle employed by ANNs. This research presents a SVM-based CCP recognition model for the on-line real-time recognition of seven typical types of unnatural CCP, assuming that the process observations are AR(1) correlated over time. Empirical comparisons indicate that the proposed SVM-based model achieves better performance in both recognition accuracy and recognition speed than the model based on a learning vector quantization network. Furthermore, the proposed model is more robust toward background noise in the process data than the model based on a back propagation network. These results show the great potential of SVM methods for on-line CCP recognition.

GA-based learning bias selection mechanism for real-time scheduling systems

The use of machine learning technologies in order to develop knowledge bases (KBs) for real-time scheduling (RTS) problems has produced encouraging results in recent researches. However, few researches focus on the manner of selecting proper learning biases in the early developing stage of the RTS system to enhance the generalization ability of the resulting KBs. The selected learning bias usually assumes a set of proper system features that are known in advance. Moreover, the machine learning algorithm for developing scheduling KBs is predetermined. The purpose of this study is to develop a genetic algorithm (GA)-based learning bias selection mechanism to determine an appropriate learning bias that includes the machine learning algorithm, feature subset, and learning parameters. Three machine learning algorithms are considered: the back propagation neural network (BPNN), C4.5 decision tree (DT) learning, and support vector machines (SVMs). The proposed GA-based learning bias selection mechanism can search the best machine learning algorithm and simultaneously determine the optimal subset of features and the learning parameters used to build the RTS system KBs. In terms of the accuracy of prediction of unseen data under various performance criteria, it also offers better generalization ability as compared to the case where the learning bias selection mechanism is not used. Furthermore, the proposed approach to build RTS system KBs can improve the system performance as compared to other classifier KBs under various performance criteria over a long period.

The study of real time scheduling by an intelligent multi-controller approach

Earlier studies indicated that using multiple dispatching rules (MDRs) for the various zones in the system can enhance the production performance to a greater extent than using a single dispatching rule (SDR) over a given scheduling interval for all the machines in the system, since MDRs employ the multi-pass simulation approach for real-time scheduling (RTS). However, if a classical machine learning approach is used, an RTS knowledge base (KB) can be developed by using the appropriate MDRs strategy (this method is called an intelligent multi-controller in this paper) as obtained from training examples. The main disadvantage of using MDRs is that the classes (scheduling decision variables) to which training examples are assigned must be provided. Hence, developing an RTS KB using the intelligent multi-controller approach becomes an intolerably time-consuming task because MDRs for the next scheduling period must be determined. To address this issue, we proposed an intelligent multi-controller incorporating three main mechanisms: (1) simulation-based training example generation mechanism, (2) data pre-processing mechanism and (3) SOM-based real time MDRs selection mechanism. Under various performance criteria over a long period, the proposed approach yields better system performance than the machine learning-based RTS using the SDR approach and heuristic individual dispatching rules.

Study on shop floor control system in semiconductor fabrication by self-organizing map-based intelligent multi-controller

To confirm semiconductor wafer fabrication (FAB) operating characteristics, the scheduling decisions of shop floor control systems (SFCS) must develop a multiple scheduling rules (MSRs) approach in FABs. However, if a classical machine learning approach is used, an SFCS in FABs knowledge base (KB) can be developed by using the appropriate MSR strategy (this method is called an intelligent multi-controller in this study) as obtained from training examples. A classical machine learning approach main disadvantage is that the classes (scheduling decision variables) to which training examples are assigned must be pre-defined. This process becomes an intolerably time-consuming task. In addition, although the best decision rule can be determined for each scheduling decision variable, the combination of all the decision rules may not simultaneously satisfy the global objective function. To address these issues, this study proposes an intelligent multi-controller that incorporates three main mechanisms: (1) a simulation-based training example generation mechanism, (2) a data preprocessing mechanism, and (3) a self-organizing map (SOM)-based MSRs selection mechanism. These mechanisms can overcome the long training time problem of the classical machine learning approach in the training examples generation phase. Under various production performance criteria over a long period, the proposed intelligent multi-controller approach yields better system performance than fixed decision scheduling rules for each of the decision variables at the start of each production interval.

Development of machine learning‐based real time scheduling systems: using ensemble based on wrapper feature selection approach

There are two items that significantly enhance the generalisation ability (i.e. classification accuracy) of machine learning‐based classifiers: feature selection (including parameter optimisation) and an ensemble of the classifiers. Accordingly, the objective in this study is to develop an ensemble of classifiers based on a genetic algorithm (GA) wrapper feature selection approach for real time scheduling (RTS). The proposed approach can better enhance the generalisation ability of the RTS knowledge base (i.e. classifier) in comparison with three classical machine learning‐based classifier RTS systems, including the GA‐based wrapper feature selection mechanism, in terms of the prediction accuracy of 10‐fold cross validation as measured according to all the performance criteria. The proposed ensemble classifier RTS also provides better system performance than the three machine learning‐based RTS systems, including the GA‐based wrapper feature selection mechanism and heuristic dispatching rules, under all the performance criteria, over a long period in a flexible manufacturing system (FMS) case study.

Study of SOM-based intelligent multi-controller for real-time scheduling

Earlier studies have indicated that the use of multiple scheduling rules (MSRs) for various zones in a system can significantly enhance the production performance over the performance obtained with the use of a single scheduling rule (SSR) over a given scheduling interval for all machines in the system through a multi-pass simulation approach for a real time scheduling (RTS) problem. However, if a classical machine learning approach is used, an RTS knowledge base (KB) can be developed using the appropriate MSR strategy (this method is called an intelligent multi-controller in this study) as obtained from training examples. A classical machine learning approach main disadvantage is that the classes (scheduling decision variables) to which training examples are assigned must be pre-defined. Hence, developing an RTS KB by the classical machine learning approach to generate training examples becomes an intolerably time consuming task because the MSRs for the next scheduling period must be pre-determined. To address this issue, this study proposes an intelligent multi-controller that consists of three main mechanisms: (1) a simulation-based training example generation mechanism, (2) a data preprocessing mechanism, and (3) a self-organizing map (SOM)-based MSR selection mechanism. The results reveal that over a long period of time this approach provides better system performance based on various performance criteria than the system performance of the machine learning-based RTS based on the SSR approach for two different types of manufacturing systems (FMS and FAB). Hence, the proposed intelligent multi-controller approach is efficient enough to be incorporated into the operation of an RTS system.

Effective Pattern Recognition of Control Charts Using a Dynamically Trained Learning Vector Quantization Network

ABSTRACT Unnatural control chart patterns (CCPs) are associated with a particular set of assignable causes for process variation. Hence, effectively recognizing CCPs can substantially narrow down the set of possible causes to be examined, and accelerate the diagnostic search. Recently, machine-learning techniques, especially the artificial neural network (ANN), have been widely used as an effective tool for CCP recognition (CCPR) tasks. Most ANN applications in CCPR have been using static supervised ANNs, such as back propagation networks (BPNs) and learning vector quantization (LVQ) networks. The false recognition problem (i.e. the patterns are misclassified) commonly encountered for these ANN-based CCPR models is mainly due to the fact that the static ANNs cannot appropriately deal with dynamic patterns, such as CCPs. In this research, a dynamic training algorithm is designed to provide an LVQ network-based CCPR model the capability to on-line recognize the dynamic CCPs that vary over time. The numerical results using simulation show that the dynamically trained LVQ network-based model proposed in this research performs much better than other ANN-based models reported in literature with respective to recognition accuracy and speed. Although this research considers the specific application of a real-time CCPR model based on an LVQ network, the proposed dynamic training algorithm could be applied to CCPR systems based on other ANN architectures in general.