Wang Fuli

A model for parameter estimation of multistage centrifugal compressor and compressor performance analysis using genetic algorithm

A model for performance prediction of multistage centrifugal compressor is proposed. The model allows the users to predict the compressor performance, e.g. pressure ratio, efficiency and losses using the compressor geometric information and speed by a stage stacking calculation based on the characteristics of each stage. To develop the compressor elemental stage characteristics, the compressor losses, such as incidence losses and friction losses, are mathematically modeled. For a composite systems, for instance a gas turbine power plant, the performance of the multistage centrifugal compressor can be evaluated. Since some important parameters of the compressor model, e.g., the slip factor σ, shock loss coefficient ζ and reference diameter D1, are hard to be determined by empirical laws, a genetic algorithm (GA) is used to solve the parameter estimation problem of the proposed model, and in turn the compressor performance analysis and parameters study are performed. The surge line for the multistage centrifugal compressor can also be determined from the simulation results. Furthermore, the model presented here provides a valuable tool for evaluating the multistage centrifugal compressor performance as a function of various operation parameters.

PID-like controller using a modified neural network

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.Abstract This paper presents an adaptive PID-like controller (PIDLC) using a modified Neural network (MNN) for learning the characteristics of a dynamic system. A PID-like cost function is proposed, and it can be shown that the Proportional-Integral-Dijferential (PID) algorithm is the gradient descent method if a specific cost function is selected. The PIDLC can cope with parametric variation and uncertainty in the controlled plant through online learning. The MNNs learning algorithm is considerably faster through the introduction of a recursive least squares (RLS) algorithm. A simulation shows that this kind of control algorithm is very effective, especially when there are variations in the plant dynamics.

A knowledge-based controller with fuzzy reasoning used in process control

A knowledge-based controller (KBC) used in process control is presented. It has three features: first, it does not need a mathematical model; secondly, the adjustable parameters of a knowledge-based controller have practical interpretation, so they can be determined easily using human experience; and thirdly, the contribution of a KBC to a controlled plant is separated into two parts: the steady-state contribution and the transient contribution. A simple fuzzy reasoning method is employed to tune the KBC parameters. The experiment and simulation results show that the KBC is very effective, especially when there are variations in the process dynamics.A knowledge-based controller with fuzzy reasoning used in process control WANG JIN , GAO WENZHONG & WANG FULI To cite this article: WANG JIN , GAO WENZHONG & WANG FULI (1997) A knowledge-based controller with fuzzy reasoning used in process control, International Journal of Systems Science, 28:6, 579-585, DOI: 10.1080/00207729708929419 To link to this article: http://dx.doi.org/10.1080/00207729708929419

Globally convergent direct adaptive control algorithm for multivariable systems with general time-delay structure

A simple adaptive control algorithm is presented which is applicable to multi-variable systems with a general time-delay structure. The proposed algorithm has the advantage that it only requires a priori knowledge of integer-valued parameters (i.e. time delays and order parameters). It is shown that, with probability one, the algorithm will ensure the system inputs and outputs are sample mean-square bounded and the conditional mean-square generalized output tracking error achieves its global minimum possible value for linear feedback control. Some simulation results are presented to demonstrate the performance of the algorithm.

A new fault detection and diagnosis method based on principal component analysis in multivariate continuous processes

The fault detection and diagnosis methods based on principal component analysis (PCA) have been developed widely because they need no detailed information about the process mechanism model and really can detect faults promptly. However the existing diagnosis algorithms such as expert systems or contribution plots, etc. still have some trouble when they are applied in real industrial processes, which leads to more extensive research on this topic. In this paper, the proposed diagnosis method utilizes the on-line loading plot and cluster analysis to give accurate cause for abnormal process conditions, which is grounded on the fact that faults normally change the correlation of process variables which may indicate more direct information about the failure cause. Thus, the principal components score plot and square predicted error (SPE) plot are used to detect the abnormal process operation condition, the loading plot and cluster analysis are used to diagnose the faults. The result shows that accurate conclusion could be obtained easily by this method.

Adaptive control of black-box nonlinear systems using recurrent neural networks

An adaptive control method of black-box nonlinear systems is presented. The control law is derived based on minimizing a suitably chosen performance index, and its implementation requires only the calculation of two key quantities, i.e., the sensitivity between the controlled system input and output and the quasi-one-step-ahead predictive output of the controlled system. In the paper, the sensitivity of the plant is estimated using the recursive rectangular window least square algorithm, and the predictive output is obtained by a recurrent neural network. The simulation results show that the proposed adaptive control method can effectively control a class of unknown nonlinear systems.

An improved genetic algorithm for a type of nonlinear programming problems

Based on the study on how to apply penalty strategy for solving a type of nonlinear programming problems by genetic algorithm, such conclusion can be drawn that only applying penalty strategy is inadequate to deal with nonlinear programming problems well. It is important to lead infeasible individuals into the feasible set during the evolution process. Penalty and repair strategy are associated to improve the performance of the algorithm. Based on such thought that the constraint which has the highest degree of violation can be satisfied first by enlarging the penalty on the individuals and repair, repair operator is proposed to perform repair operation of infeasible individuals. At the same time, based on optimization design theory, a method has been proposed to establish initial population by using uniform design. Thus, repair genetic algorithm (RGA) is proposed.

Feed-forward and inferential control and its application

Inferential control system has many excellent performances such as disturbance resisting and set-point tracking, however, the application is restricted when strong load disturbance exists or stable control accuracy and response speed are highly required in the system. Feed-forward control system responds quickly to the system with measurable disturbance, but the control accuracy is easily affected by disturbance, so there will be a big error if the disturbance model is not accurate enough. Compound control algorithm which combines the feed-forward and inferential control algorithm is proposed in this paper. Feed-forward control algorithm is used to cancel the load- torque disturbance. Inferential control algorithm is used to cancel the error caused by soft sensing method. It is used to control the speed control of guide disc. The accuracy, robustness and fast response of the control system are proved by simulation.

Adaptive generalized predictive control for nonlinear systems using neural networks

A modified neural network is presented, and an adaptive generalized predictive controller for nonlinear dynamic systems based on this modified neural network is then designed. Through extensive simulations, the proposed approach is shown to be effective for adaptive control of nonlinear systems.

Support Vector Machines and Genetic Algorithms for Soft-Sensing Modeling

Soft sensors have been widely used in industrial process control to improve the quality of product and assure safety in production. This paper introduces support vector machines (SVM) into soft-sensing modeling. Building the models, on one hand we want to have the best set of input variables, on the other hand we want to get the best possible performance of the SVM model. So the Genetic Algorithms is used to choose the input variables and select the parameters of SVM. Moreover, training the model on data coming a real experiment process—Nosiheptide fermentation process and evaluating the model performance on the same process. Results show that SVM model optimized by Genetic Algorithms provides a new and effective method for softsensing modeling and has promising application in industrial process applications.Soft sensors have been widely used in industrial process control to improve the quality of product and assure safety in production. This paper introduces support vector machines (SVM) into soft-sensing modeling. Building the models, on one hand we want to have the best set of input variables, on the other hand we want to get the best possible performance of the SVM model. So the Genetic Algorithms is used to choose the input variables and select the parameters of SVM. Moreover, training the model on data coming a real experiment process--Nosiheptidefermentation process and evaluating the model performance on the same process. Results show that SVM model optimized by Genetic Algorithms provides a new and effective method for soft- sensing modeling and has promising application in industrial process applications.