Search

This paper proposes Runge-Kutta neural disturbance observer to enhance the robustness of PID control of a system with general multicompartment lung mechanics. It is designed to observe the states of a particular type continous time, single-input single-output system where the states cannot be measured but can be observed through the single output and there exists parametric uncertainity or disturbance affecting the underlying system. It utilizes artificial neural network to estimate the disturbance online. Once an accurate disturbance estimation is obtained, it is incorporated in the system state equation and passed through the well-known Runge-Kutta integrator to predict the state values. Hence, the predicted states are obtained considering the disturbance and more robust state observation is achieved. The proposed observer is simple and easy to implement. Adaptation of the neural network is performed using gradient descent with an adaptive learning rate which guarantees convergence. The simulation results demonstrate that the proposed observer gains a significant success in enhancing the robustness of PID control at even high level of disturbance. Note that, multicompartment lung mechanics system is a stand-in model that can mimic the behavior of human lung. Thus, it is appropriate for hardware-in-the-loop simulation which opens a path to the real-patient-tests of mechanical respiratory systems in the future. Copyright (C) 2020 The Authors.

In this paper, least-squares support vector machine (LS-SVM), whose parameters are updated by unscented Kalman filter (UKF), is adopted in the generalized predictive control (GPC) of a system with general multicompartment lung mechanics. Gaussian kernel function is employed since it presents a good approximation to the inner product of nonlinear mapping possessed in the SVM formulation. In the SVM literature, it is well known that the width parameter a of the Gaussian kernel function has an important effect on the performance. However, it is not possible to train that parameter together with the other parameters of SVM when using linear least squares. This is why we use UKF for parameter adaptation in the SVM formulation. At each time instant of the control task, all parameters of the LS-SVM model, including a, are tuned simultaneously. Another reason to employ UKF is; it avoids the suboptimal solutions caused by linearization based filters, e.g., extended Kalman filter. Due to these facts, we train the SVM model using UKF and it will be referred to as the UKF-SVM model. Simulation results concerning the application of UKF-SVM based GPC to a multicompartment lung mechanics model yields plausible performance using small amount of support vectors even when there are time-varying lung parameters and disturbance of high level affecting the system. The adopted approach can also be useful when there is not any knowledge of the system dynamics, i.e., black box. Note that, multicompartment lung mechanics system is a stand-in model that can mimic the behavior of human lung. Thus, it is appropriate for hardware-in-the-loop simulation which opens a path to the real-patient-tests of mechanical respiratory systems in the future. Copyright (C) 2020 The Authors.

In this paper, two novel approaches are proposed to improve the performance of online least squares support vector machine for classification problem. First, the parameters of support vector classifier model including kernel width parameter are simultaneously updated when a new sample arrives. In that model, kernel width parameter is a nonlinear term which cannot be estimated via least squares solution. Therefore, unscented Kalman filter is adopted to train all the parameters where Karush–Kuhn–Tucker conditions are satisfied. Second, a variable-size moving window, which is updated by an intelligent strategy, is proposed to construct the support vector set. Thus, the proposed model captures the dynamics of data quickly while precluding itself to become clumsy due to big amount of useless data. In addition, adaptive support vector set provides a lower computational load especially for the large data sets. Simultaneous training of the model parameters by unscented Kalman filter and intelligent update of support vector set provides a superior classification performance compared to the online support vector classification approaches in the literature. © 2017, Springer-Verlag GmbH Germany.

This paper proposes a novel state space least squares support vector machine (SS LS-SVM) for polynomial nonlinear state space (PNLSS) model based recursive system identification. SS LS-SVM, which also possesses an adaptive kernel function, provides an optimum formulation of the monomials (ζ) of the states and input in the PNLSS model. Hence, the PNLSS model encompasses the proposed SS LS-SVM. Recursive nonlinear state space identification is developed in the output error prediction context. The input-output observations are processed sequentially, hence leading to recursive update of the parameters using conventional Gauss-Newton optimization. System states do not need to be measured. However, to to yield a conformal representation of the actual system, number of states need to be known via some physical insight. This characterizes the identification procedure as a grey box one. The PNLSS model is employed in the generalized predictive control (GPC) of a nonlinear continuously stirred tank reactor (CSTR) system. The case which includes additive white noise on the output measurements and a time-varying parameter in the nonlinear system is considered. Numerical applications give the results of a high closed loop identification performance addition to the smooth control input and closely tracking the reference in the GPC scheme. © 2018 IEEE.

The function approximation capability of a regressor model in generalized predictive control (GPC) directly affects the tracking performance of unknown nonlinear systems. In this paper, a novel deep recurrent support vector regressor (DRSVR) is proposed as a function approximator to be adopted in the GPC scheme. This study is an extension of the authors' work [1] to the control task. The DRSVR model has a recurrent state-space structure based on the least-squares support vector regressor (LS-SVR), infinite-impulse response filter (IIR) and adaptive kernel function. The model parameters, including the Gaussian kernel width parameter σ, are updated simultaneously, providing the model to capture the time-varying system dynamics quickly. Parameters are tuned online using error-square minimization via conventional Gauss-Newton optimization while keeping the poles of the IIR filter constrained in the unit circle to maintain stability. The proposed DRSVR based GPC is applied to control nonlinear HIV dynamics. The numerical applications indicate that the proposed regressor model provides high closed loop identification performance in the GPC scheme. Hence, it provides the controller with a significant tracking capability. © 2018 IEEE.

This paper introduces a novel deep recurrent support vector regressor (DRSVR) model for online regression. DRSVR model is constructed by a state equation followed by an output construction. The inner layer is actually a least squares support vector regressor (LS-SVR) of the states with an adaptive kernel function. In addition, an infinite impulse response (UR) filter is adopted in the model. LS-SVR and UR filter together constitute an intermediate layer which performs the recursive state update. Each internal state has a recurrency which is a function of the observed input-output data and the previous states. Hence, internal states track the temporal dependencies in the feature space. The outer layer is a linear combination of the states. The model parameters, including the Gaussian kernel width parameter, are updated simultaneously, that provides the model to capture the time-varying dynamics of the data quickly. Parameters are adaptively tuned using error-square minimization via conventional Gauss-Newton optimization while keeping the poles of the IIR filter constrained to maintain stability. The proposed DRSVR model is applied for real-time nonlinear system identification. The identification results indicate the accurate regression performance of the proposed model. © 2017 IEEE.

Artificial Bee Colony (ABC) and Levenberg-Marquardt (LM) optimization algorithms are applied efficiently for nonlinear constrained and unconstrained optimization problems in literature. In this paper, an intelligent hybridization method of the ABC and LM algorithms is proposed such that their global and local exploitation superiorities are unified to reduce the computational time and escape from local minima of optimization problem. In order to prove the capability of proposed hybrid algorithm, twofold experiment is conducted. In the first phase, the hybrid algorithm is applied to optimize several nonlinear unimodal, multimodal and shifted benchmark functions. Secondly, it is applied to the constrained engineering problems and compared to literature works in several performance criteria. © 2017 Elsevier Inc. All rights reserved.

In this paper, a recently introduced nonlinear gradient-based observer [1] has been adopted for Takagi-Sugeno (TS) fuzzy systems. The designed observer is especially aimed to estimate the unmeasurable states of the TS fuzzy systems where the LMI solution is not feasible to find the observer gains. The estimation of gradient observer is evaluated based on the Levenberg-Marquardt direction where the local convergence property is guaranteed using Lyapunov function approach. The numerical simulations present accurate estimation results for TS fuzzy nonlinear systems including a comparison with the conventional Extended Kalman Filter (EKF) yielding acceptable results. © 2015 IEEE.

In this paper, the well-known heuristic Artificial Bee Colony algorithm (ABC) and deterministic Levenberg-Marquardt (LM) optimization method are unified to get better performance of nonlinear optimization. In the proposed cascaded ABC-LM algorithm, the power of the ABC and LM algorithms are synergized to reduce computational-time and get rid of the problem 'stucking at local minima' of some nonlinear functions. Then, the proved power of the cascaded optimization is also tested on the training of Artificial Neural Network (ANN) for classification of XOR data and nonlinear system identification of real-time inverted pendulum set-up. The comparisons in function optimization and system identification using ABC, LM and ABC-LM showed that ABC-LM optimized nonlinear functions and ABC-LM trained ANN has resulted smaller cost functions and mean-squared-error (MSE) values, respectively. © 2013 IEEE.

In this paper, Autoregressive with exogenous input (ARX) and dynamic neural network (DNN) based generalized predictive control (GPC) methods are designed to control of nonlinear systems. ARX and DNN models adaptively approximate the plant dynamics and predict the future behavior of the nonlinear system. While control process goes on, the poles of the ARX and DNN models are constrained in a stable region using a projection operator for structural stability. Simulation results are given to compare the tracking performances of the methods. ARX-GPC and DNN-GPC both yield good tracking performances while keeping the changes in control signal as low as possible. The simulation results show that even though ARX is a linear model, it provides acceptable tracking results as well as DNN model.