Jui-Yiao Su

A study of cluster validity criteria for the fuzzy c-regression models clustering algorithm

The fuzzy c-regression models (FCRM) clustering algorithm can fit data to locally regression models which are linear in their parameters and be used as a tool to the identification of complex nonlinear systems. To date, only a few cluster validity criteria have been proposed for the FCRM clustering algorithm to validate the partitions produced by the FCRM clustering algorithm. In this article, we examine the role of a subtle but important parameter - the weighting exponent m - plays in determining the validity of FCRM partitions. The criteria considered are the partition coefficient and two cluster validity criteria we have proposed before. The limit analysis is applied to study the behavior of these cluster validity criteria as mrarr1 and mrarrinfin . It is shown that the proposed cluster validity criteria provide well responses over a wide range of m to choose the correct cluster number.

Quasi-sliding mode control for a class of multivariable discrete time bilinear systems

In this paper, we adopt the two quasi-sliding mode control for a class of multivariable discrete time bilinear systems. By applying the proposed control strategies, the system states would move into a boundary layer of the sliding surface and remain in the boundary layer in finite time. Thus, the stability of the overall system could be guaranteed. Finally, the computer simulation is applied to illustrate the control performance of the proposed control strategies.

A novel cluster validity criterion for fuzzy C-regression models

This paper proposed a novel cluster validity criterion for fuzzy c-regression models (FCRM) clustering algorithm with hyper-plane-shaped clusters. We combined the concept of fuzzy hypervolume with the compactness validity function in the cluster validity criterion. The proposed cluster validity criterion determined the appropriate number of clusters by calculating the overall compactness and separateness of the FCRM partition. The simulation results demonstrated the validness and effectiveness of the proposed method.

T-S fuzzy model identification and the fuzzy model based controller design

This paper presents an algorithm to identify T-S fuzzy models and design fuzzy model based controllers (FMBC) for a class of nonlinear plant. First, the algorithm using fuzzy c-regression models (FCRM) clustering to find the functional relationships in the product space of the input-output data. A new cluster validity criterion is proposed to calculate overall compactness and separateness of the FCRM partition and then determine the appropriate number of regression models. Besides, the fine-tuning of the antecedent fuzzy set and the consequent parameters are considered. Thus, an efficient T-S fuzzy model with compact if-then rules can be generated systematically. Finally, an FMBC is proposed to make the nonlinear plant track the reference trajectory signal. A simulation example is provided to demonstrate the accuracy and effectiveness of the proposed algorithm.

Possibilistic c-regression models clustering algorithm

The purpose of this paper is to apply the possibilistic c-means (PCM) clustering algorithm to the fuzzy c-regression models (FCRM) clustering algorithm and propose a new clustering algorithm named possibilistic c-regression models (PCRM). The PCRM clustering algorithms relaxes the column sum constrain result in each cluster, it will alleviate the noisy data effectively. Finally, the simulation examples are provided to demonstrate the effectiveness of the PCRM clustering algorithm.

Fuzzy dynamic model identification by fuzzy c-regressoin models clustering

This paper presents an algorithm to identify fuzzy dynamic (FD) model for a class of nonlinear plant. Firstly, the fuzzy c-regression models (FCRM) clustering technique is applied to partition the product space of the given input-output data into regression functional clusters. A novel cluster validity criterion with fuzzy hypervolume is set up to determine the appropriate number of clusters which has hyper-plane-shaped representatives. Furthermore, the fine-tuning procedures are included to adjust the antecedent and consequent parameters precisely. Finally, a FD model with compact number of IF-THEN rules could be generated systematically. A simulation example is provided to demonstrate the accuracy and effectiveness of the proposed algorithm.

T-S fuzzy modeling by FCRM clustering

This paper presents an algorithm to establish the T-S fuzzy model. The algorithm using fuzzy C-regression models (FCRM) clustering to find the functional relationships in the product space of the input-output data. We propose a novel cluster validity criterion to calculate overall compactness and separateness of the FCRM results and then determine an appropriate number of regression Junctions. Besides, the repartition of overlapped antecedent fuzzy set is considered. Thus, an efficient T-S fuzzy model with fewer IF-THEN rules can be generated systematically. A simulation example is provided to demonstrate the accuracy and effectiveness of our algorithm.

Fuzzy sliding mode controller design with fuzzy sliding mode observer

This paper addresses the design strategy of a decoupled fuzzy sliding mode controller and a fuzzy sliding mode observer on the basis of the Takagi-Sugenos (T-S) fuzzy model. A class of fourth-order systems such as a cart-pole system could be well controlled and observed in both the pole hyperplane and the cart-hyperplane simultaneously by the proposed fuzzy controller and observer without increasing the number of fuzzy rules. The numerical simulation of a cart-pole system confirms the validity of the proposed approaches.

The quasi-sliding mode control for a class of discrete-time bilinear systems

In this paper, we propose the two quasi-sliding mode control for bilinear systems. By proper choice of the parameters in the control law, the system states will move within the quasi-sliding mode control band. Finally, the simulations are given to illustrate the effectiveness of the proposed controller.

Chemical vapor deposition quality prediction system based on support vector regression and fuzzy learning mechanism

In advanced semiconductor manufacturing, the in-process wafers need to be monitored periodically in order to obtain high stability and high yield rate. However, the actual measurement is usually obtained after all the work-pieces of the same lot have been processed. The parameter drift or shift of the production equipment could not be detected in real-time thereby increasing the production cost. We proposed a quality prediction system (QPS) based on support vector regression (SVR) and fuzzy learning mechanism (FLM) to overcome this problem. The SVR provided good generalization performance for prediction, and the embedded FLM implied a continuous improvement or at least non-degradation of the system performance in an ever changing environment. The effectiveness of the proposed QPS was validated by test on chemical vapor deposition (CVD) process in practical 12-inch wafer fabrication. The results show that the proposed QPS not only fulfills real-time quality measurement of each wafer, but also detects the performance degradation of the corresponding machines from the information of manufacturing process.