Neural Network-based Control Research Articles

Artificial Neural Network-Based Control of Switched Reluctance Motor for Torque Ripple Reduction

Switched reluctance motor is acquiring major attention because of its simple design, economic development, and reduced dependability. These attributes make switched reluctance motors superior to other variable speed machines. The major challenge associated with the development of a switched reluctance motor is its high torque ripple. Torque ripple produces noise and vibration, resulting in degradation of its performance. Various techniques are developed to cope with torque ripples. Practically, there exists not a single mature technique for the minimization of torque ripples in switched reluctance motors. In this research, a switched reluctance motor is modelled and analysed. Its speed and current control are implemented through artificial neural networks. Artificial neural network is found to be a promising technique as compared with other techniques because of its accuracy, reduced complexity, stability, and generalization. The Levenberg–Marquardt algorithm is utilized in artificial neural networks due to its fast and stable convergence for training and testing. It is found from research that artificial neural network-based improved control shows better performance of the switched reluctance motor. Realization of this technique is further validated from its mean square error analysis. Operating parameters of the switched reluctance motor are improved significantly. Simulation environment is created in Matlab/Simulink.

Artificial neural network-based current control of field oriented controlled induction motor drive

A hysteresis current controller (HCC) is commonly used in high-performance AC motor drives to control the current directly. Recently, the predictive current controller (PCC) is also used as an alternative to the classical current controller for speed and torque regulation of induction motor (IM) drives. However, PCC has drawbacks of large flux and torque ripples, large total harmonic distortions (THDs) in current and voltage and dependency on parameters. This paper proposes current control with artificial neural network (ANN) for a field-oriented controlled induction motor (FOCIM) drive. The ANN has input current error between the reference and the measured stator currents. The output function of neuron is a hyperbolic tan (or tan-sigmoid) function to apply error Levenberg–Marquardt (L–M) back propagation as learning rule because of its fast convergence. The proposed method is based on a new approach in which hysteresis band is replaced by ANN comparator to improve the performance of the FOCIM drive. It minimizes torque ripples, flux ripples, voltage and current THDs over the existing HCC and PCC methods. The superiority of the proposed method compared to existing methods is established by simulation and experimental results.

Adaptive RBF neural network-based control of an underactuated control moment gyroscope

Radial basis function (RBF) neural networks have the advantages of excellent ability for the learning of the processes and certain immunity to disturbances when using in control systems. The robust trajectory tracking control of complex underactuated mechanical systems is a difficult problem that requires effective approaches. In particular, adaptive RBF neural networks are a good candidate to deal with that type of problems. In this document, a new method to solve the problem of trajectory tracking of an underactuated control moment gyroscope is addressed. This work is focused on the approximation of the unknown function by using an adaptive neural network with RBF fully tuned. The stability of the proposed method is studied by showing that the trajectory tracking error converges to zero while the solutions of the internal dynamics are bounded for all time. Comparisons between the model-based controller, a cascade PID scheme, the adaptive regressor-based controller, and an adaptive neural network-based controller previously studied are performed by experiments with and without two kinds of disturbances in order to validate the proposed method.

Accurate power-sharing, voltage regulation, and SOC regulation for LVDC microgrid with hybrid energy storage system using artificial neural network

ABSTRACT In this paper, an artificial neural network-based control strategy is proposed for low voltage DC microgrid (LVDC microgrid) with a hybrid energy storage system (HESS) to improve power-sharing between battery and supercapacitor (SC) to suit the demand-generation imbalance, maintain state-of-charge (SOC) within boundaries and thereby to regulate the dc bus voltage. The conventional controller cannot track the SCs current rapidly with the high-frequency component that will place dynamic stress on the battery, further resulting in shorter battery life. The significant advantage is that in the proposed control strategy, redirections of unwaged battery currents to SCs for fast compensations enhance battery life span. The proposed control strategy effectiveness was investigated by simulations, including a comparison of overshoot/undershoot and settling time in dc bus voltage with a conventional control strategy. The results have been experimentally verified by hardware-in-loop (HIL) on a field-programmable gate array (FPGA)-based real-time simulator.

Analysis on existence of compact set in neural network control for nonlinear systems

Neural network method is an effective tool for approximating the unknown function in controller design for nonlinear systems. To guarantee the validity of the approximation, state variables in approximated unknown functions need to stay in a compact set. However, in most existing results, the existence of the compact set has not been correctly proven; therefore, the proof is not actually complete in these existing works. In this paper, we analyze the existence of compact sets for two typical nonlinear systems with novel neural network-based controllers and show the strict proof for the semi-global uniform ultimate boundedness of the closed-loop system.

Intelligent Spectrum Controlled Supplemental Lighting for Daylight Harvesting

This article presents a neural network-based control method for daylight harvesting in a proof-of-concept greenhouse consisting of emulated sunlight and dimmable light emitting diode light fixtures. The objective of this multi-input-multi-output lighting system is to deliver desired levels of light, within a specific spectrum range, to locations of interest in a grow tent. To this end, a learning neural network controller with online adaptive weights is presented which can achieve stability with small errors in the presence of disturbances and modeling uncertainties. A stability analysis of the closed-loop system is presented along with a selection method for obtaining the control parameters. The neural controller is enhanced with an antiwindup mechanism to account for the nonlinear effect of actuator saturation. Experimental results are presented to verify the proposed daylighting control strategy which confirm analytic and simulation studies.

Multilayer neural networks-based control of underwater vehicles with uncertain dynamics and disturbances

In the presence of uncertain dynamic terms and external disturbances, the problem of trajectory tracking with application to an underactuated underwater vehicle is addressed in this paper. Based on Lyapunov theory and properties of neural networks, a nonlinear neural controller is designed, where multilayer neural networks are adopted to approximate the unmodeled dynamic terms and external disturbances. In order to confine the values of estimated weights within predefined bounds, smooth projection functions are employed. Moreover, measurement noises are considered so as to simulate realistic operation scenario, while filters are designed to get cleaner states. From the stability analysis, it is proven that the tracking errors are globally uniformly ultimately bounded. Numerical examples are provided to demonstrate the robustness of the controller in the presence of unmodeled terms, disturbances and measurement noises.

Neural network-based fault-tolerant control approach considering a submarine system

The stability and satisfactory performance of a fault-tolerant control approach is important for designers and therefore the corresponding closed-loop systems are essential to be dealt with. Considering one of the most important faults including the operator fault in robotic systems leads us to reach the better performance, usually. The radial basis function neural network-based fault-tolerant control method is designed in this research to estimate the model uncertainties, the noise affecting the system, and the operator fault, in order to compensate this one. The results of simulating the fault-tolerant control system propose that this technique successfully secured the tracking function of the control system besides ensuring closed-loop stability. Since the traditional control methods fail to effectively fix the aforementioned problem, a new method so-called neural-network-based fault-tolerant control is proposed. This study proposes the design of a nominal controller, its examinations under operator faults, and finally the expansion of the controller into a fault-tolerant controller. In a word, the contribution made in the research over the state-of-the-art materials focusing on fault-tolerant controls designed in the area of submarine systems is briefly taken into real consideration as its nonlinearity, its fixed structure, the lack of need for switching systems, resistance to noise, and the ability to compensate the effect of the performance drop fault on the collective fault of the operators. Subsequently, the simulation results verified the effectiveness of the proposed approach in coping with the operator’s performance faults, tangibly.

Design of neural network and PLC-based water flow controller

Flow rate is a fundamental physical quantity in the fluid transportation system from one place to another. To achieve this, a reliable controller that is able to produce a constant flowrate in industry is needed. The most used flow controllers in industries are PID-based controllers that are implemented using PLCs. However, there are still shortcomings, they can perform poorly in some applications, for example in the highly nonlinear system which cannot be overcome by conventional PID controllers. There are some other limitations of PID controller, such as PID has the overshoot and undershoots in the output of controlled system and PID gives late response in this study, a neural network-based flow controller is proposed to deal with that problems. The controller will be operated in a miniature plant which consists of a water tank, water pump, a control valve, and a flow transmitter. Due to PLC limitation that cannot be programmed with common programming languages such as MATLAB, a personal computer (PC) is used to run the proposed neural network controller. The PC communicates with the PLC using OPC (OLE for Process Control) server, while the PLC reads the flow transmitter and also controls the control valve directly based on the result output of the neural network controller. In order to evaluate the performance of the proposed controller, several experiments have been conducted. The performance of the proposed controller has been compared with the conventional PID controller. It shows that neural network-based controller outperformed the conventional PID controller, in terms of maximum overshoot and steady-state error, where the neural network controller has maximum overshoot = 5.36% and steady-state error = 0.85%, while the PID controller has 11.3% for overshoot and 1.10 % for steady-state error.

Fully Adaptive-Gain-Based Intelligent Failure-Tolerant Control for Spacecraft Attitude Stabilization Under Actuator Saturation

This article investigates the attitude stabilization problem of a rigid spacecraft with actuator saturation and failures. Two neural network-based control schemes are proposed using anti-saturation adaptive strategies. To satisfy the input constraint, we design two controllers in a saturation function structure. Taking into account the modeling uncertainties, external disturbances, and adverse effects from actuator faults and failures, the first anti-saturation adaptive controller is implemented based on radial basis function neural networks (RBFNNs) with a fixed-time terminal sliding mode (FTTSM) containing a tunable parameter. Then, we upgrade the proposed controller to a fully adaptive-gain anti-saturation version, in order to strengthen the robustness and adaptivity with respect to actuator faults and failures, unknown mass properties, and external disturbances. In the two schemes, all of the designed adaptive parameters are scalars, thus they only require light computational load and can avoid the redesign process of the controller during spacecraft operation. Finally, the feasibility of the proposed methods is illustrated via two numerical examples.

Self-adapting J-type air-based battery thermal management system via model predictive control

Battery thermal control plays an indispensable role in terms of the safety and performance for electric vehicles. For air-based cooling technologies, one of the most pressing challenges is to balance the temperature uniformity and constrain the maximum temperature simultaneously under varying driving conditions. This paper proposes a self-adaptive intelligent neural network-based model predictive control strategy for a J-type air-based battery thermal management system. The J-type structure is first optimized through surrogate-based optimization to improve the temperature uniformity before control. Based on the optimized J-type configuration, an operation mode switching module is developed to mitigate the temperature unbalance. The thermal control approach is tested using an integrated driving cycle, and its evaluations are threefold: (i) the neural network-based control without mode switching fails to meet the thermal requirements; (ii) the control with mode switching succeeds in constraining the maximum temperature and maintaining the temperature uniformity within 1.33 K; (iii) the added model predictive control approach slightly enhances the thermal performance but improves the energy efficiency significantly by 15.8%. The results show that the J-type structure with its appropriate control strategy is a promising solution for light-duty electric vehicles using an air-cooling technology.

Real-Time Optimal Approach and Capture of ENVISAT Based on Neural Networks

A neural network-based controller is developed to enable a chaser spacecraft to approach and capture a disabled Environmental Satellite (ENVISAT). This task is conventionally tackled by framing it as an optimal control problem. However, the optimization of such a problem is computationally expensive and not suitable for onboard implementation. In this work, a learning-based approach is used to rapidly generate the control outputs of the controller based on a series of training samples. These training samples are generated by solving multiple optimal control problems with successive iterations. Then, Radial Basis Function (RBF) neural networks are designed to mimic this optimal control strategy from the generated data. Compared with a traditional controller, the neural network controller is able to generate real-time high-quality control policies by simply passing the input through the feedforward neural network.

Modeling and Adaptive Control and Power Sharing in Islanded AC Microgrids

This paper deals with the problem of control and power sharing for distributed generators in AC islanded microgrids. A one-layer adaptive control strategy based on two fixed-time adaptive control methods is designed to ensure voltage and frequency regulation and reference tracking and accurate active and reactive power sharing. The main advantage of the proposed approach is accurate power sharing and voltage and frequency reference restoration before the desired pre-fixed time despite the unknown disturbances and without requiring a secondary control loop. Compared to the existing microgrid adaptive controllers such as adaptive neural network-based controllers and adaptive fuzzy logic controllers among others, the design and tuning of the proposed techniques is simple and straightforward and does not require any knowledge of the microgrid parameters, lines impedance or load power demand. The developed adaptive control law has also a simple mathematically expression and thus it’s easily implemented and does not require a heavy computational performance. A comparative simulation study with conventional secondary control and power sharing is provided to highlight the effectiveness of the control strategy in term of reference tracking, robustness to load power variations and to distributed generators plug-and-play.

Power enhancement in a grid connected solar PV System by using PLL-neural based converter

This work deals with neural network control algorithm-based grid connected to solar photo voltaic (PV) system consisting of DC-AC converter. The reference solar-grid current for three-leg VSC are estimated using neural network control algorithm. The neural network control algorithm based solar PV system is modeled in MATLAB R2018a along with SIMULINK.. This study presents an artificial neural network-based controller for regulating the level of active and reactive power output. First, the three phase currents from the VSI are measured and compared with the three reference currents. The neural network is trained to have minimum output error. It was concluded that the power output from the system was found to be190 KVA in case of system having no intelligent controller and 700 KVA in case of system with AAN based control. The voltage of output is maintained to be 20 kV in the grid system for analysis purpose. Thus the proposed control is expected to be implemented in the renewable energy resources for better output.

A robot learning framework based on adaptive admittance control and generalizable motion modeling with neural network controller

Robot learning from demonstration (LfD) enables robots to be fast programmed. This paper presents a novel LfD framework involving a teaching phase, a learning phase and a reproduction phase, and proposes methods in each of these phases to guarantee the overall system performance. An adaptive admittance controller is developed to take into account the unknown human dynamics so that the human tutor can smoothly move the robot around in the teaching phase. The task model in this controller is formulated by the Gaussian mixture regression to extract the human-related motion characteristics. In the learning and reproduction phases, the dynamic movement primitive is employed to model a robotic motion that is generalizable. A neural network-based controller is designed for the robot to track the trajectories generated from the motion model, and a radial basis function neural network is used to compensate for the effect caused by the dynamic environments. Experiments have been performed using a Baxter robot and the results have confirmed the validity of the proposed robot learning framework.

Neural network‐based quickprop control algorithm for grid connected solar PV‐DSTATCOM system

For the optimal operation of grid interfaced solar photovoltaic (PV) system, a neural network-based Quickprop control algorithm is presented in this study. The solar PV array supplies maximum power by utilising an incremental conductance-based maximum power point tracking technique to the grid and the load. When the solar power is not present, during cloudy days or at night, the distribution static compensator (DSTATCOM) operation is performed by harmonics mitigation and reactive power compensation of the loads connected at the point of intersection. The proposed system improves power quality when solar PV power is present, along with active power transfer from solar PV array to grid/load. Thus, a smooth transition is provided between these modes with neural network-based Quickprop control algorithm. Moreover, the neural network-based control technique offers enhanced accuracy due to the combinational neural structure in the estimation process. A laboratory prototype is developed for validation and experimental results corroborate reliable operation for modes of operation as DSTATCOM and grid interfaced PV system at varying load and solar insolation condition.

A Neural Network Based Landing Method for an Unmanned Aerial Vehicle with Soft Landing Gears

This paper presents the design, implementation, and testing of a soft landing gear together with a neural network-based control method for replicating avian landing behavior on non-flat surfaces. With full consideration of unmanned aerial vehicles and landing gear requirements, a quadrotor helicopter, comprised of one flying unit and one landing assistance unit, is employed. Considering the touchdown speed and posture, a novel design of a soft mechanism for non-flat surfaces is proposed, in order to absorb the remaining landing impact. The framework of the control strategy is designed based on a derived dynamic model. A neural network-based backstepping controller is applied to achieve the desired trajectory. The simulation and outdoor testing results attest to the effectiveness and reliability of the proposed control method.

Neural Network-Based Adaptive Motion Control for a Mobile Robot with Unknown Longitudinal Slipping

When the mobile robot performs certain motion tasks in complex environment, wheel slipping inevitably occurs due to the wet or icy road and other reasons, thus directly influences the motion control accuracy. To address unknown wheel longitudinal slipping problem for mobile robot, a RBF neural network approach based on whole model approximation is presented. The real-time data acquisition of inertial measure unit (IMU), encoders and other sensors is employed to get the mobile robot’s position and orientation in the movement, which is applied to compensate the unknown bounds of the longitudinal slipping using the adaptive technique. Both the simulation and experimental results prove that the control scheme possesses good practical performance and realize the motion control with unknown longitudinal slipping.

Biologically Inspired Motion Modeling and Neural Control for Robot Learning From Demonstrations

In this paper, we propose a biologically inspired framework for robot learning based on demonstrations. The dynamic movement primitive (DMP), which is motivated by neurobiology and human behavior, is employed to model a robotic motion that is generalizable. However, the DMP method can only be used to handle a single demonstration. To enable the robot to learn from multiple demonstrations, the DMP is combined with the Gaussian mixture model (GMM) to integrate the features of multiple demonstrations, where the conventional GMM is further replaced by the fuzzy GMM (FGMM) to improve the fitting performance. Also, a novel regression algorithm for FGMM is derived to retrieve the nonlinear term of the DMP. Additionally, a neural network-based controller is developed for the robot to track the generated motions. In this network, the cerebellar model articulation controller is employed to compensate for the unknown robot dynamics. The experiments have been performed on a Baxter robot to demonstrate the effectiveness of the proposed methods.

Resource and Energy Saving Neural Network-Based Control Approach for Continuous Carbon Steel Pickling Process

Resource and Energy Saving Neural Network-Based Control Approach for Continuous Carbon Steel Pickling Process