Search

This study focuses on the design of an observer-based adaptive fuzzy-neural-network control (OAFNNC) for real-time levitated balancing and propulsive positioning of a hybrid magnetic-levitation (maglev) transportation system with only position state feedback. The dynamic model of the hybrid maglev transportation system, including levitated hybrid electromagnets to reduce the suspension power loss and the friction force during linear movement and a propulsive linear induction motor (LIM) based on the concepts of mechanical geometry and motion dynamics, is firstly constructed. Then, an adaptive observer is designed to estimate velocity signals for the later control utilization, and an adaptive observer and control (AOC) scheme is formed by means of the stability analyses of the entire system. The ultimate goal is to design an on-line fuzzy-neural-network (FNN) velocity-sensorless control methodology to cope with the problem of the complicated control transformation and the requirement of detailed system parameters in the AOC scheme, and to directly ensure the stability of the entire system without the requirement of strict constraints, detailed system information and auxiliary compensated controllers despite the existence of uncertainties. In the proposed OAFNNC scheme, a FNN control is utilized to be the major control role by imitating the AOC strategy, and adaptation laws for network parameters are derived in the sense of projection algorithm and Lyapunov stability theorem to ensure the network convergence as well as stable control performance. The effectiveness of the proposed control strategy for the hybrid maglev transportation system is verified by experimental results, and the superiority of the OAFNNC scheme is indicated in comparison with the AOC strategy and the backstepping particle-swarm-optimization control (BSPSOC) system in previous research. [ABSTRACT FROM AUTHOR]

• A total-sliding-mode-control (TSMC)-based distributed secondary control (TSMC-DSC) scheme is designed. • Using an adaptive fuzzy-neural-network (FNN) to mimic the TSMC-DSC law. • A model-free control structure with favorable convergence speed, no chattering phenomenon, and strong robustness. • Training AFNN-DSC parameters online adaptively. • The network convergence as well as the Lyapunov stability. In this study, an adaptive fuzzy-neural-network (FNN) control scheme is proposed for an islanded micro-grid (MG) as a distributed secondary controller (DSC) to achieve the aims of voltage and frequency restoration and the optimal power sharing. Firstly, the dynamic model of an islanded MG is built, which consists of an inverter-interfaced distributed generation (DG) model and a MG architecture model. The DG model can be represented by considering the dynamics of a primary controller with an optimal active power sharing scheme. The MG architecture model is composed of power flow dynamics and loads. Then, a consensus-algorithm-based error function is defined, and a model-dependent total sliding-mode control (TSMC) technique is presented for dealing with synchronization and tracking problems. Moreover, an adaptive FNN (AFNN) scheme is designed to mimic the TSMC law to inherit its fast dynamic response with robust properties. Meanwhile, the requirement of precise information of the MG dynamic model in the TSMC law can be relaxed by the AFNN scheme. Adaptive tuning algorithms for FNN network parameters of the AFNN-based DSC (AFNN-DSC) strategy are derived by using the projection algorithm and the Lyapunov stability theorem, which can guarantee the stability of the AFNN-DSC-controlled system. The effectiveness of the proposed control method is verified by numerical simulations for real scenarios. © 2017 Elsevier Inc. All rights reserved. [ABSTRACT FROM AUTHOR]

The control performance of an adaptive and a fuzzy neural network (FNN) sliding-mode controlled quick-return mechanism, which is driven by a field-oriented control permanent magnet (PM) synchronous servo motor, is presented in this study. First, Hamilton’s principle and Lagrange multiplier method are applied to formulate the equation of motion. Then, based on the principle of the sliding-mode control, an adaptive sliding-mode controller is developed to control the slider position of the motor-mechanism coupled system. Moreover, an FNN sliding-mode controller is implemented to control the motor-quick-return servomechanism for comparison. Finally, the effectiveness of the proposed adaptive and FNN sliding-mode controllers are demonstrated by some simulated and experimental results. Compared with the adaptive sliding-mode controller, the FNN sliding-mode controller results much less tracking error with improved control performance. [Copyright &y& Elsevier]

This study presents a sliding-mode neural-network (SMNN) control system for the tracking control of an n rigid-link robot manipulator to achieve high-precision position control. The aim of this study is to overcome some of the shortcomings of conventional robust controllers such as a model-based adaptive controller requires the system dynamics in detail; the fuzzy rule learning scheme has a latent stability problem; an adaptive control scheme for robot manipulator via fuzzy compensator requires strict constrained conditions and prior system knowledge. In the SMNN control system, a neural network controller is developed to mimic an equivalent control law in the sliding-mode control, and a robust controller is designed to curb the system dynamics on the sliding surface for guaranteeing the asymptotic stability property. Moreover, an adaptive bound estimation algorithm is employed to estimate the upper bound of uncertainties. All adaptive learning algorithms in the SMNN control system are derived from the sense of Lyapunov stability analysis, so that system-tracking stability can be guaranteed in the closed-loop system whether the uncertainties occur or not. Computer simulations of a two-link robot manipulator verify the validity of the proposed control strategy in the possible presence of uncertainties and different trajectories. The proposed SMNN control scheme possesses two salient merits: (1) it guarantees the stability of the controlled system, and (2) no constrained conditions and prior knowledge of the controlled plant is required in the design process. This new intelligent methodology provides the designer with an alternative choice to control an n rigid-link robot manipulator. [Copyright &y& Elsevier]

A robust recurrent fuzzy neural network control (RFNNC) system is proposed to control the position of the mover of a permanent magnet linear synchronous motor drive system in this study. In the proposed RFNNC system, a RFNN controller is the main tracking controller, that is used to mimic an ideal feedback linearization control law, and a robust controller is proposed to confront the shortcoming of the RFNN controller. Moreover, to relax the requirement for the bound of lumped uncertainty, which comprises a minimum approximation error, optimal parameter vectors and higher order terms in Taylor series, a RFNNC system with adaptive bound estimation is investigated. In the control system a simple adaptive algorithm is utilized to estimate the bound of lumped uncertainty. In addition, simulated and experimental results due to periodic reference trajectories show that the dynamic behaviors of the proposed control systems are robust with regard to uncertainties. [Copyright &y& Elsevier]

The dynamic response of a hybrid computed torque controlled toggle mechanism, which is driven by a permanent magnet (PM) synchronous servo motor, is studied in this paper. First, based on the principle of computed torque control, a position controller is developed for the motor–toggle servomechanism. Moreover, to relax the requirement of the lumped uncertainty in the design of a computed torque controller, a fuzzy neural network (FNN) uncertainty observer is utilized to adapt the lumped uncertainty on line. Furthermore, based on the Lyapunov stability a hybrid control system, which combines the computed torque controller, the FNN uncertainty observer and a compensated controller, is proposed to control the position of a slider of the motor–toggle servomechanism. The computed torque controller with FNN uncertainty observer is the main tracking controller, and the compensated controller is designed to compensate the minimum approximation error of the uncertainty observer instead of increasing the rules of the FNN. Finally, simulated and experimental results due to a periodic sinusoidal command show that the dynamic behaviors of the proposed hybrid control system are robust with regard to parametric variations and external disturbances. [Copyright &y& Elsevier]

This study presents an intelligent control system for an induction servo motor drive to track periodic commands using a wavelet neural network (WNN). With the field orientation mechanism, the dynamic behavior of the induction servo motor drive system is rather similar to a linear system. However, the uncertainties, such as mechanical parametric variation, external disturbance, unstructured uncertainty due to nonideal field orientation in transient state, and unmodelled dynamics in practical applications influence the control performance. Therefore, an intelligent control system that is an on-line trained WNN controller with adaptive learning rates is proposed to control the rotor position of the induction servo motor drive. The adaptive learning rates are derived in the sense of discrete-type Lyapunov stability theorem, so that the convergence of the tracking error can be guaranteed in the closed-loop system. In the whole design process, the strict constrained conditions and prior knowledge of the controlled plant are not necessary according to the powerful learning ability of the intelligent control system. With the proposed intelligent control system, the controlled induction servo motor drive possesses the advantages of good tracking control performance and robustness to uncertainties under wide operating ranges. The effectiveness of the proposed control scheme is verified by both simulated and experimental results. [Copyright &y& Elsevier]

A mechanical cleavage (MC) approach has been demonstrated to synthesize graphene nanosheets (GNs) as heat sink materials from artificial graphite paper (GP). The facile MC method is composed of three main steps: GP isolation, GP exfoliation, and GN collection. The method is capable of preparing few layers of GNs repeatedly without using chemical oxidizing agents and costly deposition apparatus. The as-prepared GN powders are well characterized by X-ray diffraction and Raman spectroscopy. On the basis of the experimental results, the MC method shows a great feasibility to synthesize high-quality GN products with high repeatability and environmental friendliness. We also report that the addition of GN onto Cu foil induces an improved capability for heat dissipation, as compared with original GP and Cu heat foil. According to the calculations of Fourier's law, the thermal conductivity of the GN/Cu composite heat sink can reach as high as 2142 W/m K, leading to 26% increase of thermal conductivity compared to the GP heat sink. [ABSTRACT FROM AUTHOR]

[Display omitted] • A novel two-level classification strategy is designed to reduce the number ofmisjudged samples of arc fault diagnosis due to the similarity of waveforms between loads. • Aiming at the problem that the original waveform features are not prominent enough, a data random fusion strategy is proposed to construct a new one-dimensional data set to achieve the purpose of data feature enhancement. • A new neural network is designed, which integrates the multi-head attention mechanism and gated recurrent unit and constructs multiple data extraction channels. The network can effectively extract features from new data sets. The series arc faults of the electric lines in the low-voltage distribution network can lead to devastating fires, posing a significant threat to the lives and safety of residents. Aiming at the problems that arc faults are challenging to identify due to the variety of load types in the lines, a new diagnosis method of two-level classification for arc faults based on data random fusion and MC-MGCNN network is proposed. Firstly, the first level of load classification is carried out. Extreme weighted fuzzy entropy is calculated to separate the dimmer load that could easily lead to misjudgment from the total load pool, and the concepts of simple load pool and complex load pool are established. Secondly, the corresponding classifiers are designed for the two different load pools for the second level of diagnosis classification. Specifically, the time–frequency domain features are extracted for the load waveforms in a simple load pool, and extreme gradient boosting (XGBoost) is used for rapid diagnosis. On the contrary, for the nonlinear load waveforms and multi-load combination waveforms in the complex load pool, the random fusion mechanism of data is applied for feature enhancement, and a new one-dimensional data set is constructed. Finally, the latest data is input into the multi-channel convolution neural network combining multi-head attention and gated recurrent unit (MC-MGCNN) network for arc fault diagnosis. The experimental results show that the first level of load classification can effectively reduce the number of misjudgment samples in the complex load pool and improve the overall binary classification accuracy by 7.54 %. The new data set gains different degrees of improvement in the identification accuracy of other comparative networks. In addition, the new network has good feature learning ability and diagnostic accuracy of arc faults. The accuracy of the multi-classification is up to 99.5 %, and the binary classification accuracy is up to 99.94 %. The test is carried out in the Raspberry PI 4b environment, and the detection time is 52.3 ms, which verified the feasibility of hardware deployment. [ABSTRACT FROM AUTHOR]

The technology for identifying birds around power towers using cameras alone is still susceptible to environmental interference. This paper proposes a new bird damage recognition network, RCVNet, which addresses this issue by fusing radio-frequency (RF) images and visual images. The network employs a feature layer fusion approach that accurately identifies bird damages in the monitoring area. Initially, RCVNet takes a group of RF and visual images as input. Then, through a series of convolutional neural networks (CNNs), birds are identified and located. To overcome challenges in recognizing small targets, several improved modules such as cross-supervised fusion network (CSF-net), posture deformable convolution (PDF), small-target attention fusion mechanism (SAFM), and Tiny-YOLOHead are introduced throughout RCVNet, improving surface information utilization and small feature retention rates. Finally, a bird damage discrimination strategy is developed based on the recognition outcomes of birds. As there is currently no public dataset available for RCVNet training, a new bird dataset called CRB2022, which includes RF and visual images, was gathered. Through large-scale experiments utilizing these methods, RCVNet effectively identifies birds, achieving a mean average precision of 79.34% and a mean average recall of 83.29%. Additionally, the discrimination rate of the utilized strategy can reach up to 98%. [ABSTRACT FROM AUTHOR]