Improved Gradient Descent Research Articles

Variable structure fuzzy wavelet neural network controller for complex nonlinear systems

Controlling a complex nonlinear system has always been an important problem. A neuro fuzzy controller may be an appropriate controller for such systems. There have been always problems with large number of neurons which cause heavy complex computations in neuro fuzzy controllers. In this paper, a variable structure neuro fuzzy controller is introduced to prevent number of neurons from rising. The proposed variable structure controller, adapts itself on-line to the system. Therefore, because of the adaptive structure, there is no need to have a large number of neurons. As a result, calculations and controller complexity will decrease in this structure. The proposed controller also uses an improved gradient descent method to update different varying parameters. In this improved method, adaptive learning rates are used to prevent parameters getting stuck in local optima. Improved gradient descent method also takes the advantages of Lyapunov stability method in updating the parameters. The use of Lyapunov method in updating the network parameters will guarantee the stability of the controller, too. A complex nonlinear time delay system is introduced in this paper to evaluate the proposed variable structure fuzzy wavelet neural network controller. Simulation results illustrate the efficacy of this new controller.

A New PID Neural Network Controller Design for Nonlinear Processes

In this paper, a novel adaptive tuning method of PID neural network (PIDNN) controller for nonlinear process is proposed. The presented method utilizes an improved gradient descent method to adjust PIDNN parameters where the margin stability will be employed to get high tracking performance and robustness with regard to external load disturbance and parameter variation. Simulation results show the effectiveness of the proposed algorithm compared with other well-known learning methods.

Intelligent driving methods based on expert knowledge and online optimization for high-speed trains

Abstract In the engineering control practice of High-Speed Train (HST), the traditional automatic driving method increases the energy consumption and impairs the intelligence of train operation. Different from previous studies, we propose the intelligent driving methods (IDMs), including expert knowledge system and online optimization algorithms, to achieve the multi-objective (safety, punctuality, energy efficient, passengers’ riding comfort, and so on) control of HST. First, we establish the expert knowledge system based on the driving data and control rules of excellent drivers. Then, in order to enhance the adaptability and real-time performance of proposed IDMs, two online optimization algorithms, including exact online programming driving (EOPD) and inexact online programming driving (IOPD), are developed by improved gradient descent and stochastic meta-decent method to update the controller’s output online. Finally, using the field data collected from Beijing-Shanghai High-Speed Railway, the proposed IDMs are verified under the real speed-limit conditions. The simulation results show that EOPD and IOPD can achieve better performances than automatic driving method based on ATO, Fuzzy PID controller and traditional multi-objective optimization method, especially in passengers’ riding comfort and energy-consumption. Furthermore, as the step size is selected with wide randomness in the updating process, IOPD has more operating mode switching times than EOPD but its punctuality is better.

A Data-Driven Soft Sensor Modeling Method Based on Deep Learning and its Application

Soft sensors have been widely used in industrial processes. The core issue of data-driven soft sensors is building soft sensor models with excellent performance and robustness. This paper introduces deep learning to soft sensor modeling and proposes a novel soft sensor modeling method based on a deep learning network that integrates denoising autoencoders with a neural network (DAE-NN). An improved gradient descent is employed to update the model parameters. The proposed modeling method is able to capture the essential information of input data through deep architecture, building soft sensors with excellent performance. The DAE-NN-based soft sensor is applied in practical applications to estimate the oxygen content in flue gasses in 1000-MW ultrasuperficial units. Comparing conventional soft sensor modeling methods, i.e., shallow learning methods, DAE-NN-based soft sensor significantly improves the performance and generalization of data-driven soft sensors. Deep learning provides a very effective and promising method for soft sensor modeling.

Design of polynomial fuzzy observer–controller with membership functions using unmeasurable premise variables for nonlinear systems

In this paper, the stability of polynomial fuzzy-model-based (PFMB) observer-control system is investigated via Lyapunov stability theory. The polynomial fuzzy observer with unmeasurable premise variables is designed to estimate the system states. Then the estimated system states are used for the state-feedback control of nonlinear systems. Although the consideration of the polynomial fuzzy model and unmeasurable premise variables enhances the applicability of the fuzzy-model-based (FMB) control strategy, it leads to non-convex stability conditions. Therefore, the refined completing square approach is proposed to derive convex stability conditions in the form of sum of squares (SOS) with less manually designed parameters. In addition, the membership functions of the polynomial observer-controller are optimized by the improved gradient descent method, which outperforms the widely applied parallel distributed compensation (PDC) approach according to a general performance index. Simulation examples are provided to verify the proposed design and optimization scheme.

An adaptive PID neural network for complex nonlinear system control

Usually it is difficult to solve the control problem of a complex nonlinear system. In this paper, we present an effective control method based on adaptive PID neural network and particle swarm optimization (PSO) algorithm. PSO algorithm is introduced to initialize the neural network for improving the convergent speed and preventing weights trapping into local optima. To adapt the initially uncertain and varying parameters in the control system, we introduce an improved gradient descent method to adjust the network parameters. The stability of our controller is analyzed according to the Lyapunov method. The simulation of complex nonlinear multiple-input and multiple-output (MIMO) system is presented with strong coupling. Empirical results illustrate that the proposed controller can obtain good precision with shorter time compared with the other considered methods.

Power Transformers Fault Diagnosis Based on Fuzzy-RBF Neural Network

Dissolved Gas Analysis (DGA) is a popular method to detect and diagnose different types of faults occurring in power transformers. Improved three-ratio is an effective method for transformer fault diagnosis used in recent years. This paper applies appropriate Artificial Neural Networks (ANN) to resolve the online fault diagnosis problems for oil-filled power transformer based on improved three-ratio. Because of the characteristic of improved three-ratio boundary is too absolute, a method using fuzzy math theory to deal with the data of the neural network input is also proposed. A major kind of neural network, i.e. radial basis function neural network (RBFNN), is used to model the fault diagnosis structure. In addition, to improve the convergence speed, an improved gradient descent algorithm is used in training RBFNN. Through on-line monitoring the concentrations of the dissolved gases, the proposed diagnostic system can offer a way to interpret the incipient faults. The simulation diagnosis demonstrates the effectiveness and veracity of the proposed method.

Control of Multivariable Coupling System Based on Improved PID Neural Networks

The paper analyzes the characteristics and the situation of the control of multivariable coupling system, combining improved PID neural networks with PSO algorithm, and finally designs a suitable controller model. To achieve a controller model, an initial structure of PID neural networks is established in the first place. Then weights in the network is initialized by PSO algorithm and optimized by improved gradient descent algorithm. A simulation, MIMO coupling system, prove this controller has such characteristics with short time and high precision. The research of the paper provides a new idea and approach for control of complex coupling system.

Designing fuzzy inference system based on improved gradient descent method

The distribution of sampling data influences completeness of rule base so that extrapolating missing rules is very difficult. Based on data mining, a self-learning method is developed for identifying fuzzy model and extrapolating missing rules, by means of confidence measure and the improved gradient descent method. The proposed approach can not only identify fuzzy model, update its parameters and determine optimal output fuzzy sets simultaneously, but also resolve the uncontrollable problem led by the regions that data do not cover. The simulation results show the effectiveness and accuracy of the proposed approach with the classical truck backer-upper control problem verifying.