Katsuhiro Inoue

Robust Fault Detection Using Index of Kullback Discrimination Information

Abstract This paper presents a robust fault detection system (FDS) for dynamic systems with unmodeled dynamics. In the FDS, umnodeled dynamics is first qualified as soft bound, which as well as model parameters are estimated using a robust identification algorithm. Then as a fault detection index, Kullback discrimination information (KDI) is derived into a feasible form and an index of umnodeled dynamics is also introduced. A decision making scheme is thus developed so that fault detection is carried out based on the KDI, the index of umnodeled dynamics and other prior information about the system.

An indirect approach to adaptive control of nonlinear systems using quasi-ARMAX model

Abstract This paper presents an adaptive control scheme for nonlinear systems based on a quasi-ARMAX prediction model that is a specially constructed associative memory network consisting of multiple neurofuzzy models and is distinctive to usual neural networks in that it is linear in both the parameters to be estimated and the input variables to be synthesized in a control system. This advantage is taken to develop a nonlinear adaptive control scheme similarly to linear one.

3-D surface recovery from range images by using multiresolution wavelet transform

In this paper, a surface recovery method using multiresolution wavelet transform is proposed. For representing 3-D surface shapes, 4th order B-spline functions are introduced as scaling functions of spline wavelets. A regularization problem is solved in order to estimate the surface function. The estimated surface function can be decomposed into an approximate surface function at specific lower resolution and the corresponding wavelet components by using the multiresolution wavelet transform. Consequently, by reducing the noise components which the wavelet components include, the surface recovery method can give a further accurate estimation of the surface function. Through several experiments, both the robustness to noises and the edge-preserving property in recovering the surface have been attained.

A Method of Robust Fault Detection for Dynamic Systems by Using Quasi-ARMAX Modeling

Abstract This paper presents a model-based approach to fault detection of dynamic systems, which is robust to unmodeled dynamics. A “Quasi-ARMAX model᾿ is first proposed for describing nonlinear systems by incorporating a group of certain nonlinear structures into a linear ARMAX structure. The model can be used for a best linear approximation of the system, as well as for the estimation of resulting unmodeled dynamics, by a hierarchical implementation of recursive identification. Then robust fault detection is performed based on thresholding approach using Kullback discrimination information as fault detection index, in which the estimated unmodeled dynamics is incorporated.

A Quick Identification Method of Continuous-Time Nonlinear Systems and Its Application to Power Plant Control

Abstract In this paper, a quick identification method based on the short time record of input-output data is introduced for joint state and parameter estimation of nonlinear continuous-time systems. The method can then be used for on-line monitoring the system with unknown parameters. An application way of the method to main-steam temperature control of a thermal power plant are developed in the framework of model reference adaptive control system (MRACS). Simulation studies on a super-heater model have been carried out to demonstrate the effectiveness of the proposed method.

A Neural Network Approach to Failure Decision of Adaptively Controlled Systems

Abstract In this paper, a combined method of change detection and failure decision is proposed for the system under the adaptive control based on the self-tuning regulator. The controlled system is assumed encounter unexpected parameter changes, which may be caused by a failure or a normal operation. Such a system change can effectively be detected by using Fullback Discrimination Information (KDI) as an index for model discrimination. In order to decide whether the detected system change is caused by a failure or not, a neural network approach to failure decision is introduced. Based on the knowledge about failure modes and system operations, the regulator parameter variations after the change detection are used as training data for the network learning. In this way an on-line monitoring scheme of adaptively controlled systems can be established. Simulation studies of a second-order damped oscillator have been earned out to demonstrate the effectiveness of the method.

Surface recovery by using regularization theory and its application to multiresolution analysis

In this paper, a surface recovery method using multiresolution wavelet transform is proposed. For representing 3D surface shapes, 4th order B-spline functions with uniform knots are introduced as scaling functions of spline wavelets. In order to estimate the surface function, a regularization problem is solved by an iterative algorithm. The estimated surface function can, be decomposed into an approximate surface function at the lowest resolution and the corresponding wavelet components. Consequently, by reducing the noise components which the wavelet components include, the surface recovery method can give an accurate estimation of the surface function. Through several experiments, both the robustness to noises and the edge-preserving property in recovering the surface have been confirmed.

Necessary Conditions for Multilayer Nets having Solutions and Convergent Superiority of Bipolar Nets

Abstract This paper formulates a necessary condition for multilayer nets to have solutions by a set of normal vectors orthogonal to separation hyperplanes. Comparing the necessary condition to the distributions of normal vectors with the weights and biases initialized ordinarily by random numbers with zero mean, it is derived that bipolar nets are superior to unipolar nets in convergence of the back propagation learning initialized in such an ordinary manner.

Feature Extraction Method for EEG Waves by Using MAP Detector of Input Signals with Multi-Discrete-Level Amplitude

Abstract In this paper, a dynamical model is introduced to develop a new automatic stage determination system of human sleep electroencephalogram (EEG) not using the wave pattern recognition step. The EEG generating mechanism is modeled by a damped system excited by impulse input process which subjects to Poisson process whose amplitude has a transition property. Furthermore, A Maximum A Posteriori (MAP) method is modified to be applicable to the model. We can then extract useful information which is directly related to sleep stages based on the model and MAP detector.