Katsuji Uosaki

System parameter estimation by evolutionary strategy

This paper discusses an application of Evolutionary Strategy (ES), a simple direct probabilistic search and optimization algorithm to system parameter estimation. We consider multiple estimates of the system parameters, which interactively or independently seeks for the better estimates to fit the environment, and construct an estimate by some algebraic operations on these estimates. Illustrating numerical examples show the applicability of the proposed idea from the viewpoints of adaptability and robustness.

Adaptive identification of non-stationary systems with multiple forgetting factors

Recently much attentions have been paid to the use of adaptive forgetting factor related to the level of system alertness to estimate the parameters in nonstationary stochastic systems. The approaches are, however, generally complex to apply. We propose, in this paper, an adaptive identification method called multiple forgetting factors (MMF) method for nonstationary linear stochastic systems with time-varying parameters. The parameter estimates are constructed as a weighted sum of the estimates obtained by multiple recursive least squares methods operating in parallel with constant but different forgetting factors and the weights for each weighted least squares method are adjusted to fit the time variation of the parameters. The identification method has a simple structure and is quite easy to implement. Simulation experiments show that the proposed method works well not only for systems with abruptly changing parameters but also for systems with gradually changing ones.

Optimal Auxiliary Input for On-Line Fault Detection and Fault Diagnosis

Abstract An optimal auxiliary input is introduced to detect the system fault quickly. The auxiliary input is designed to enlarge the distance measured by the Kullback discrimination information between the system models corresponding to the normal and the fault modes in single fault mode case. Furthermore the idea is extended to multiple fault mode case (fault diagnosis problem).Numerical simulation results indicate that theproposed auxiliary input in fault detection using the backward sequential probability ratio test reduces the mean detection time without making much effects on original system behavior and false alarm and diagnosis rates.

Evolutionary Approach to System Identification

Abstract We consider in this paper a novel approach to system identification using an evolution strategy(ES). In this approach, we consider the multiple estimates of the system parameters, which interactively or independently seek for the better estimates to fit the observed behavior of system and construct an overall estimate by some algebraic operations on these estimates. Numerical simulation results show the usefulness of the proposed idea from the viewpoints of adaptability, robustness and applicability to diverse systems including nonlinear and time—varying systems.

Optimal auxiliary input design for fault diagnosis

Abstract An optimal input design problem in frequency domain is discussed for quick fault diagnosis of dynamical systems without affecting the original system in normal mode. Since the optimal auxiliary input for detecting a certain fault mode is not always good for detection of other fault modes, i.e., it may reduce the the distance between the normal mode and the other fault modes and make harder to detect true fault mode, the max-min approach is proposed in this paper. The auxiliary input is designed to maximize the minimum distance measured by the Kullback discrimination information measure between system models corresponding to the normal and each fault mode in order to detect the true fault mode. Results of a numerical simulation result indicate the applicability of the proposed auxiliary input in fault diagnosis.

Optimal Auxiliary Input for Fault Detection - Frequency Domain Approach -

Abstract Optimal auxiliary input design problem for early fault detection is discussed in the frequency domain. Auto-covariance sequence of the optimal auxiliary- input signal is obtained by solving a mathematical programming problem. It provides the maximum of the time-average of the Kullback discrimination information between the system models corresponding to the normal and the fault modes. Numerical simulation results indicate that the proposed auxiliary input reduces the mean detection time without making much effects on the original system behavior and false alarm rate in fault detection.

Hammerstein model identification method based on genetic programming

We address a novel approach to identify a nonlinear dynamic system for a Hammerstein model. The Hammerstein model is composed of a nonlinear static block in series with a linear, dynamic system block. The aim of system identification is to provide the optimal mathematical model of both nonlinear static and linear dynamic system blocks in some appropriate sense. We use genetic programming to determine the functional structure for the nonlinear static block. Each individual in genetic programming represents a nonlinear function structure. The unknown parameters of the linear dynamic block and the nonlinear static block given by each individual are estimated with a least square method. The fitness is evaluated by AIC (Akaike information criterion) as representing the balance of model complexity and accuracy. It is calculated with the number of nodes in the genetic programming tree, the order of the linear dynamic model and the accuracy of model for the training data. The results of numerical studies indicate the usefulness of proposed approach to Hammerstein model identification.

Wiener model identification by evolutionary computation approach with piecewise linearization

We address a novel approach to identify a nonlinear dynamic system for Wiener models, which are composed of a linear dynamic system part followed by a nonlinear static part. The aim of the system identification is to provide an optimal mathematical model for both the linear dynamic and nonlinear static parts in an appropriate sense. Assuming the nonlinear static part is invertible, we approximate the inverse function by a piecewise linear function. We estimate the piecewise linear inverse function by using the evolutionary computation approach such as the genetic algorithm and evolution strategies, and estimate the linear dynamic system part by the least squares method. The results of numerical simulation studies indicate the usefulness of proposed approach to the Wiener model identification.

Optimal Auxiliary Input Design for Fault Detection of Systems with Model Uncertainty

Abstract Introduction of an auxiliary input is known to be useful to detect the system fault quickly without affecting the original system behavior in normal mode. Such an auxiliary input is designed to enlarge the distance measured by the Kullback discrimination information measure between the system models corresponding to the normal and the fault modes. However, in practice, the designer hardly knows the exact models. Hence, the optimal auxiliary input should be designed for the case of the system models with uncertainty. Here, the auxiliary input is designed to maximize the distance for the worst combination of system models. Numerical simulation results indicate the applicability of the proposed auxiliary input in fault detection

Optimal auxiliary input for fault detection of systems with model uncertainty

Introduction of an auxiliary input is known to be useful to detect the system fault quickly without affecting the original system behavior in normal mode. Such an auxiliary input is designed to enlarge the distance measured by the Kullback discrimination information measure between the system models corresponding to the normal and the fault modes. However, in practice, the designer hardly knows the exact models. Hence, the optimal auxiliary input should be designed for the case of the system models with uncertainty. Here, the auxiliary input is designed to maximize the distance for the worst combination of system models. Numerical simulation results indicate the applicability of the proposed auxiliary input in fault detection.