Neural and MTS Algorithms for Feature Selection

Su, Chao-Ton,Li, Te-Sheng 한국품질경영학회 2002 The Asian Journal on Quality Vol.3 No.2

The relationships among multi-dimensional data (such as medical examination data) with ambiguity and variation are difficult to explore. The traditional approach to building a data classification system requires the formulation of rules by which the input data can be analyzed. The formulation of such rules is very difficult with large sets of input data. This paper first describes two classification approaches using back-propagation (BP) neural network and Mahalanobis distance (MD) classifier, and then proposes two classification approaches for multi-dimensional feature selection. The first one proposed is a feature selection procedure from the trained back-propagation (BP) neural network. The basic idea of this procedure is to compare the multiplication weights between input and hidden layer and hidden and output layer. In order to simplify the structure, only the multiplication weights of large absolute values are used. The second approach is Mahalanobis-Taguchi system (MTS) originally suggested by Dr. Taguchi. The MTS performs Taguchi's fractional factorial design based on the Mahalanobis distance as a performance metric. We combine the automatic thresholding with MD; it can deal with a reduced model, which is the focus of this paper. In this work, two case studies will be used as examples to compare and discuss the complete and reduced models employing BP neural network and MD classifier. The implementation results show that proposed approaches are effective and powerful for the classification.

Optimization of Multiple Quality Characteristics for Polyether Ether Ketone Injection Molding Process

Kuo Chung-Feng Jeffrey,Su Te-Li The Korean Fiber Society 2006 Fibers and polymers Vol.7 No.4

This study examines multiple quality optimization of the injection molding for Polyether Ether Ketone (PEEK). It also looks into the dimensional deviation and strength of screws that are reduced and improved for the molding quality, respectively. This study applies the Taguchi method to cut down on the number of experiments and combines grey relational analysis to determine the optimal processing parameters for multiple quality characteristics. The quality characteristics of this experiment are the screws' outer diameter, tensile strength and twisting strength. First, one should determine the processing parameters that may affect the injection molding with the $L_{18}(2^1{\times}3^7)$ orthogonal, including mold temperature, pre-plasticity amount, injection pressure, injection speed, screw speed, packing pressure, packing time and cooling time. Then, the grey relational analysis, whose response table and response graph indicate the optimum processing parameters for multiple quality characteristics, is applied to resolve this drawback. The Taguchi method only takes a single quality characteristic into consideration. Finally, a processing parameter prediction system is established by using the back-propagation neural network. The percentage errors all fall within 2%, between the predicted values and the target values. This reveals that the prediction system established in this study produces excellent results.

Analysis and Construction of a Quality Prediction System for Needle-Punched Non-woven Fabrics

Kuo Chung-Feng Jeffrey,Su Te-Li,Chiu Chin-Hsun,Tsai Cheng-Ping The Korean Fiber Society 2007 Fibers and polymers Vol.8 No.1

In this study, polyester and polypropylene staple fibers were selected as the raw material, and then processed through roller-carder, cross-lapper and needle-punching machine to produce needle-punched non-woven fabrics. First, the experiment was planned using the Taguchi method to select processing parameters that affect the quality of the needle-punched non-woven fabric to act as the control factors for this experiment. The quality characteristics were the longitudinal and transverse tensile strength of the non-woven fabric as well as longitudinal and transverse tear strength. The $L_{18}(2^1{\times}3^7)$ orthogonal array was selected for the experiment as it offered an improvement on the traditional method that wastes a lot of time, effort and cost. By using the analysis of variance(ANOVA) technique at the same time, the effect of significant factors on the production process of needle-punched non-woven fabrics could be determined. Finally, the processing parameters were set as the input parameters of a back-propagation neural network(BPNN). The BPNN consists of an input layer, a hidden layer and an output layer where the longitudinal/transverse tensile and tear strength of the non-woven fabric were set as the output parameters. This was used to construct a quality prediction system for needle-punched non-woven fabrics. The experimental results indicated that the prediction system implemented in this study provided accurate predictions.

Computerized Color Separation System for Printed Fabrics by Using Backward-Propagation Neural Network

Kuo, Chung-Feng Jeffrey,Su, Te-Li,Huang, Yi-Jen The Korean Fiber Society 2007 Fibers and polymers Vol.8 No.5

Textile production must be coupled with hi-tech assistant system to save cost of labor, material, time. Therefore color quality control is one very important step in any textiles, however excellent the fabric material itself is, if it lacks good color, then it may still result in dull sale. Therefore, this paper proposes a printed fabrics computerized color separation system based on backward-propagation neural network, whose primary function is to separate rich color of printed fabrics pattern so as to reduce time-consuming manual color separation color matching of current players. What it adopted was RGB color space, expressed in red, green, and blue. Analyze color features of printed fabrics, use gene algorithm to find sub-image with same color distribution as original image of printed fabrics yet smaller area, for later color separation algorithm use. In terms of color separation algorithm, this paper relied on supervised backward-propagation neural network to conduct color separation of printed fabrics RGB sub-image, and utilized $PANTONE^{(R)}$ standard color ticket to do color matching, so as to realize accurate color separation.

SOPC Based Weight Lifting Control Design for Small-Sized Humanoid Robot

Tzuu-Hseng S. Li,Chia-Ling Hsu,Chun-Yang Hu,Yu-Te Su,Ming-Feng Lu,Shao-Hsien Liu 제어로봇시스템학회 2008 제어로봇시스템학회 국제학술대회 논문집 Vol.2008 No.10

An SOPC based small-sized humanoid robot for weight lifting is introduced in this paper. All performances will be implemented using the SOPC chip, Altera EP1C12F324C8. The contribution of this paper is mainly the design of the self-balanced control of a small sized humanoid robot. According to the concept of sensory reflex, we combine the signals of the accelerometers and force sensors with a fuzzy controller to design a dynamic balanced controller for the humanoid robot. Through the implementation of the controller, we can strengthen the activity stability and the robustness of adapting to lift the weight. Finally, the experiment results show that the robot can successfully execute weight lifting function.

Optimization of the Needle Punching Process for the Nonwoven Fabrics with Multiple Quality Characteristics by Grey-Based Taguchi Method

Kuo, Chung-Feng Jeffrey,Su, Te-Li,Tsai, Cheng-Ping The Korean Fiber Society 2007 Fibers and polymers Vol.8 No.6

This study is intended for finding out the optimal processing parameters for needle punching nonwoven fabrics in order to work out its maximal strength. Taguchi method together with grey relational analysis is employed to resolve the problem as regards multiple-quality optimization, and further discover the optimal combination of processing parameters for needle punching nonwoven fabrics. Firstly, orthogonal array $L_{18}(2^1{\times}3^7)$ is used to deal with the processing parameters that may exert influence over the manufacturing of needle punching nonwoven fabrics. Then grey relational analysis is applied to resolve the deficiency of Taguchi method that focus on single quality characteristic. Next, the response table of grey relational analysis is used to obtain the optimal combination of processing parameters for multiple quality characteristics. In the current experiment quality characteristic refers to the tensile strength and tear strength of the nonwoven fabrics. Additionally, signal-to-noise ratio (SN ratio) calculation and analysis of variance (ANOVA) can be adopted to explore the experimental results. Through ANOVA, the significant factors that exert comparatively significant influence over the quality characteristic of the needle punching nonwoven fabrics, that is, the control factors are determined so that the quality characteristic of the needle punching nonwoven fabrics can be effectively controlled. Finally, confirmation experiment is conducted within 95 % confidence interval to verify the experimental reliability and reproducibility.