Adaptive general regression neural network for modelling of dynamic plants

Abstract

This paper proposes an integrated general regression neural network (GRNN ) adaptation scheme for dynamic plant modelling. The scheme can be used in a noisy and dynamic environment for on-line process control. It possesses several distinguished features compared with the original GRNN proposed by Specht, such as a flexible pattern nodes add-in and delete-off mechanism, a dynamic initial sigma assignment using a non-statistical method, automatic target adjustment and sigma tuning. These adaptation strategies are formulated on the basis of the inherent advantageous features found in GRNN, such as highly localized pattern nodes, good interpolation capability and instantaneous learning. Good modelling performance was obtained when the GRNN is tested on a linear plant in a noisy environment. It performs better than the well-known extended recursive leastsquares identification algorithm. In this paper, analysis of the effects of some of the adaptation parameters involving a non-linear plant is also investigated. The results show that the proposed methodology is computationally efficient and exhibits several attractive features such as fast learning, flexible network sizing and good robustness, which are suitable for the construction of estimators or predictors for many model-based adaptive control strategies.

Evolving General Regression Neural Networks using Limited Incremental Evolution for Data-Driven Modeling of Non-linear Dynamic Systems

“…A benchmarking discrete non-linear plant from [3] is used for simulation. The plant can be written as:…”

“…GRNN can be used in approximation, prediction, modeling and control tasks. GRNN have been used to model dynamic plants [3], nonlinear system identification [4], batch processes modeling and monitoring [5]. Also, in fault detection and diagnosis in air-handling unit [6], control of microgrid hybrid power systems [7], dead-zone estimation and compensation in control of traveling wave ultrasonic motor [8].…”

Concerns in reporting of serum vascular endothelial growth factor levels: Comment on the article by Nakahara et al

“…Therefore, the size of pattern layer will be under control to avoid ever‐increasing computational complexity and benefits the simple implementation in the hardware. Different from the modification on GRNN algorithm (Seng, 1999) that can directly use the true signals of output Y all the time in brain control model of BMI application, the neural activities are the only available signals to predict the movement. Therefore, the output Y i stored in dynamic pattern layer uses the prediction of the updating network.…”

Combining Biometric Fractal Pattern and Particle Swarm Optimization‐Based Classifier for Fingerprint Recognition

“…However, it is difficult to derive the partial derivative of the nonlinear MSEF as 2.9 . To simplify the learning procedure, the first partial derivatives of MSEF are used in this study [24][25][26] . The best smoothing parameters and mean squared errors versus the number of iterations are also shown in Figure 5.…”