Neural Network Based Control Algorithm Research Articles

A NEAT Based Two Stage Neural Network Approach to Generate a Control Algorithm for a Pultrusion System

Controlling complex systems by traditional control systems can sometimes lead to sub-optimal results since mathematical models do often not completely describe physical processes. An alternative approach is the use of a neural network based control algorithm. Neural Networks can approximate any function and as such are able to control even the most complex system. One challenge of this approach is the necessity of a high speed training loop to facilitate enough training rounds in a reasonable time frame to generate a viable control network. This paper overcomes this problem by employing a second neural network to approximate the output of a relatively slow 3D-FE-Pultrusion-Model. This approximation is by orders of magnitude faster than the original model with only minor deviations from the original models behaviour. This new model is then employed in a training loop to successfully train a NEAT based genetic control algorithm.

Active Compensation Method for Strong Magnetic Interference of MEMS Electronic Compass

In practical applications, an electronic compass composed of a three-axis magnetometer and a three-axis accelerometer is often due to the strong magnetic field interference of the carrier, which makes the magnetometer unable to accurately calculate the heading, and even causes the electronic compass to fail. This problem puts forward more stringent requirements on the installation and use of the electronic compass on the carrier. Generally, traditional magnetic field compensation methods such as ellipse/ellipsoid fitting method can effectively compensate electronic compasses. However, these methods cannot be used when the output of the magnetometer is saturated. In order to overcome the above influence, this paper proposes a three-axis external coil active compensation method. By analyzing the error characteristics of the magnetometer, a coil compensation model was established, and a neural network-based control algorithm was designed to realize the adaptive compensation of current. The use of genetic algorithm can improve the global search ability, while controlling the convergence of the BP neural network, which has better real-time performance and higher reliability. The simulation experiment results show that the method has better optimization ability and higher compensation accuracy, the strong magnetic interference of the carrier is effectively compensated, and the working environment adaptability of the MEMS electronic compass is significantly improved.

Development of a neural network based control algorithm for powered ankle prosthesis

Lower limb amputation is partial or complete removal of the limb due to disease, accident or trauma. Surface electromyograms (sEMG) of a large number of muscles and force sensors have been used to develop control algorithms for lower limb powered prostheses, but there are no commercial sEMG controlled prostheses available to date. Unlike ankle disarticulation, transtibial amputation yields less intact lower leg muscle mass. Therefore, minimizing the use of sEMG muscle sources utilized will make powered prosthesis controller economic, and limiting the use of specifically the lower leg muscles will make it flexible. Presently, we have used healthy population data to (1) test the feasibility of the neural network (NN) approach for developing a powered ankle prosthesis control algorithm that successfully predicts sagittal ankle angle and moment during walking using exclusively sEMG, and (2) rank all muscle combination variations according to their success to determine the economic and flexible NN’s. sEMG amplitudes of five lower extremity muscles were used as inputs: the tibialis anterior (TA), medial gastrocnemius (MG), rectus femoris (RF), biceps femoris (BF) and gluteus maximus (GM). A time-delay feed-forward-multilayer-architecture NN algorithm was developed. Muscle combination variations were ranked using Pearson's correlation coefficient (r > 0.95 indicates successful correlations) and root-mean-square error between actual vs. estimated ankle position and moment. The trained NN TA + MG was successful (rposition = 0.952, rmoment = 0.997) whereas, TA + MG + BF (rposition = 0.981, rmoment = 0.996) and MG + BF + GM (rposition = 0.955, rmoment = 0.995) were distinguished as the economic and flexible variations, respectively. The algorithms developed should be trained and tested for data acquired from amputees in new studies.

Zero-Sum Game Based Cooperative Control for Onboard Pulsed Power Load Accommodation

In this paper, the noncooperative control problem of onboard pulsed power load is formulated as a two-player zero-sum game. One player is the optimal controller which is designed to optimize a predefined cost function. The other player is the disturbance that represents the overall damping effect of the system, including that of unmodeled system dynamics. Neuro-dynamic programming based control design is developed to solve the nonlinear optimal control problem under disturbance. The neural network based control algorithm can achieve the near-optimal control without the acknowledgment of system dynamics. In addition, the control design can relax the requirements for initial admissible control conditions and predetermined control references. The effectiveness of the proposed algorithm is demonstrated through the real-time simulations on both simplified and detailed shipboard power system models.

Fast multilayer perceptron neural network‐based control algorithm for shunt compensator in distribution systems

In this study, a fast learning method of back-propagation (BP) multilayer perceptron neural network-based control algorithm for shunt compensator in three-phase distribution systems is presented. The proposed method comprises of quadratic linear and non-linear errors to determine optimisation criterion error function to train the BP algorithm while the existing methods have used only linear quadratic error term. The newly developed optimisation criterion error function accelerates the convergence efficiency of BP algorithm for performance improvement of shunt compensator at point of common coupling under non-linear loading conditions. With the help of the proposed algorithm, the weighted amplitude of fundamental active and reactive current components of the load current are extracted from which the reference source currents are estimated. The performance analysis of the proposed algorithm has been evaluated using two case studies for zero-voltage regulation and power factor correction. The total harmonic distortions are improved in comparison with standard BP algorithm which has been validated in above-mentioned two case studies. This is the quite important advantage of the proposed control algorithm to improve the power quality over existing control algorithms for shunt compensator.

Power Quality Improvement in Isolated Distributed Power Generating System Using DSTATCOM

This paper presents the control of a synchronous reluctance generator (SyRG) driven by a biogas/biomass diesel engine as a prime mover in a distributed power generating system. This generator is used as a source to feed linear and nonlinear loads. A three-leg voltage source converter (VSC)-based distribution static compensator (DSTATCOM) is implemented using an adaptive neural network-based control algorithm for harmonic suppression, load balancing, and voltage regulation in a three-phase SyRG system with a battery energy storage system. This control algorithm is used for extraction of active and reactive power components of distorted load currents. These components of load currents are used for estimation of reference source currents to generate the gating pulses of VSC used as DSTATCOM. The performance of DSTATCOM is observed satisfactory for this type of power generating system under balanced and unbalanced loads.

Neural Network-Based Control Algorithm for DSTATCOM Under Nonideal Source Voltage and Varying Load Conditions

Distribution static compensator (DSTATCOM) is the optimal choice of power quality (PQ) compensator in a three-phase four-wire distribution system for the mitigation of PQ problems. The performance of the PQ compensator under varying load and nonideal source conditions relies on the control strategy. A neural network-based p-q control algorithm is proposed in this paper for the DSTATCOM, which comprises of a four-leg voltage-source converter with a dc capacitor. The proposed control strategy implements five artificial neural network controllers for, the conversion of nonideal voltage source into ideal sinusoidal voltage, the extraction of dc component p of load real power supplied to the load, maintenance of the voltage across the capacitor, and mitigation of neutral current. The performance of the proposed neural network-based p-q control strategy for DSTATCOM is evaluated under various possible source and load conditions by simulating in MATLAB/Simulink environment, and the results obtained through the simulation are validated experimentally by a prototype developed in the laboratory. Both the experimental and simulation results prove that the performance of the proposed neural network-based control strategy is superior to the conventional method.

Implementation of four‐leg distribution static compensator

An implementation of a four-leg distribution static compensator (DSTATCOM) using an adaptive neural network-based control algorithm for compensation of linear/non-linear loads using voltage source converter is presented here. The proposed control algorithm, which is based on adaptive neural network, is used for extraction of fundamental active and reactive power components of load currents which are major components in the estimation of reference supply currents. This control algorithm is implemented on a developed four-leg DSTATCOM for reactive power compensation, harmonics elimination, load balancing and neutral current mitigation under linear and non-linear loads. The performance of DSTATCOM is observed quite satisfactory under various linear and non-linear loads.

Editorial to special issue: Biomedical engineering: information processing, modeling, and control

The Seventh International Symposium on Neural Networks (ISNN 2010) was held in Shanghai, China, from June 6 and 9, 2010. The ISNN 2010 was a great success and covered broad topics of neural network research and applications in diverse fields. As a continuation of this successful conference, we organized this special issue of Neural Computing and Applications on ‘‘Biomedical engineering: information processing, modeling, and control’’, with 10 selected highquality papers from ISNN 2010. The theme of this special issue is to present the state-ofthe-art developments in recent research focusing on neural networks for various biomedical engineering problems. With the increase of importance of the biomedical research, many interesting theories and applications in this challenging field have been studied over the past years. Meanwhile, neural networks have shown their powerful capacities in the learning and control field. It is reasonable to believe that the new theory, methodology, and tools for biomedical researches can be developed because of the advance of neural network study. To this end, the 10 selected papers in this special issue can be categorized into two major sections under this theme. The first section is focused on the neural network–based modeling and control for biomedical engineering. For instance, M. Amiri et al. studied the modeling problem of a severe neurological disease, epilepsy. They obtained a modified biologically inspired neural model for the epilepsy by integrating the functional role of astrocytes in the regulation of synaptic transmission. With this modified model, it is demonstrated that the healthy astrocytes are capable to compensate the variation of cortical excitatory input. The paper by P. C. Francisco and W. Yu proposes a neural network–based control algorithm for the magnetic levitation platform. This platform can be inserted into the patient’s abdominal cavity to perform diagnosis or surgery. The proposed controller combines the neural control and the sliding mode control together, which can reduce the chattering phenomenon and ensure the finite time convergence. The paper by G. Chen et al. considers the stabilization problem of a neutral stochastic hybrid systems with time-varying delays, which can describe some meaningful phenomena in biology. Based on the Lyapunov functional approach and linear matrix inequality (LMI) techniques, the delay-dependent sufficient conditions for the existence of the non-fragile observer–based H1 controller are given. The paper by C. Mu et al. suggests a novel least square support vector machine (LS-SVM)–based internal model controller for the MIMO nonlinear discrete-time systems. LS-SVM is employed to approximate the inverse system by using the input–output data. The proposed method can overcome the problem of the imperfect mathematical model and enhance the system robustness. The paper by Q. Cheng and J. Cao considers the global synchronization problem for a class of complex networks with both discrete time delays and stochastic disturbances. Recently, the complex network has been proved to be a useful modeling approach for many biological and social behaviors. The major tools for analyzing the network synchronization are the delay-fractioning approach, the properties of Kronecker product, and stochastic analysis technique. The obtained Z.-G. Hou (&) L. Cheng M. Tan Key Laboratory of Complex Systems and Intelligence Science, Institute of Automation, Chinese Academy of Sciences, P.O. Box 2728, 100190 Beijing, China e-mail: zengguang.hou@ia.ac.cn

Neural network model-based predictive control of liquid–liquid extraction contactors

The inherent complex nonlinear dynamic characteristics and time varying transients of the liquid–liquid extraction process draw the attention to the application of nonlinear control techniques. In this work, neural network-based control algorithms were applied to control the product compositions of a Scheibel agitated extractor of type I. Model predictive control algorithm was implemented to control the extractor. The extractor hydrodynamics and mass transfer behavior were modeled using the non-equilibrium backflow mixing cell model. It was found that model predictive control is capable of solving the servo control problem efficiently with minimum controller moves. This study will be followed by more work concentrated on using different neural network-based control algorithms for the control of extraction contactors.

Semiactive Neurocontrol for Seismic Response Reduction Using Smart Damping Strategy

A new semiactive control strategy that combines a neurocontrol system with a smart damper is proposed to reduce seismic responses of structures. In the proposed semiactive control system, the improved neurocontroller, which was developed by employing a training algorithm based on a cost function and a sensitivity evaluation algorithm to replace an emulator neural network, produces the desired active control force, and then a bang-bang-type controller clips the control forces that cannot be achieved by a smart damper (e.g., a variable orifice damper, controllable fluid damper, etc.). Therefore, the proposed semiactive control strategy is fail-safe in that the bounded-input, bounded-output stability of the controlled structure is guaranteed. Numerical simulation results show that the proposed semiactive control system that employs a neural network-based control algorithm is quite effective in reducing seismic responses.

Study of the control-relevant properties of backpropagation neural network models of nonlinear dynamical systems

Neural networks have been used for important applications in chemical engineering. Such applications include: fault detection in process control systems, modeling of nonlinear dynamics, control of nonlinear systems, solution of nonlinear programming problems and others. Most of the articles published on the subject consider neural networks to be “black boxes” and concentrate on the applications. After introducing the basic principles of neural networks, we will concentrate on studying control-relevant properties of neural network models of nonlinear systems. In particular, we study the stability of these models as well as the stability of the model's inverse. This provides a background with which to study the neural network-based control algorithms proposed in the literature.