Jixin Qian

Application of particle swarm optimization algorithm on robust PID controller tuning

The performance of the PID controller may deteriorate when the operating condition of a process fluctuates. A robust parameter tuning method to improve the PID controller performance under bounded model uncertainty is presented. First an enhanced performance criterion is proposed to reduce the overshoot and large control move. Then the robust tuning problem is formulated as a Min-Max optimization. Particle Swarm Optimization (PSO) is applied to solve the nonlinear, non-differentiable problem. Examples are given to demonstrate the effectiveness of the proposed method. Compared with other PID tuning methods, the result shows that better performance can be achieved with the model parameter fluctuation.

Multivariate Statistical Process Monitoring of an Industrial Polypropylene Catalyzer Reactor with Component Analysis and Kernel Density Estimation

Abstract Data-driven tools, such as principal component analysis (PCA) and independent component analysis (ICA) have been applied to different benchmarks as process monitoring methods. The difference between the two methods is that the components of PCA are still dependent while ICA has no orthogonality constraint and its latent variables are independent. Process monitoring with PCA often supposes that process data or principal components is Gaussian distribution. However, this kind of constraint cannot be satisfied by several practical processes. To extend the use of PCA, a nonparametric method is added to PCA to overcome the difficulty, and kernel density estimation (KDE) is rather a good choice. Though ICA is based on non-Gaussian distribution information, KDE can help in the close monitoring of the data. Methods, such as PCA, ICA, PCA with KDE (KPCA), and ICA with KDE (KICA), are demonstrated and compared by applying them to a practical industrial Spheripol craft polypropylene catalyzer reactor instead of a laboratory emulator.

A Novel Robust Tuning Strategy for Model Predictive Control

A novel and easy-to-use robust tuning strategy for model predictive control (MPC) is presented. The proposed strategy based on min-max algorithm can deal with model uncertainty explicitly; it could design an MPC controller with strong robustness and small overshooting, owing to the performance index employed in the strategy. Another contribution of the performance index is to avoid large prediction horizon and control horizon being selected to MPC controllers, which can reduce the MPC online computation. The superiority of the proposed robust tuning strategy has been demonstrated by simulation results

The Distributed Multicast Routing Scheme with Delay Constraint using Ant Colony Optimization

The delay-constrained minimum-cost multicast routing problem is known to be NP-complete. Currently many heuristic algorithms have been proposed, most of which are centralized, or centralized in nature. A completely distributed scheme based on ant colony optimization is proposed. In this scheme, no core node with the whole network topology information exists and the ants only use the local information of their current nodes to find the destinations. Combined the characteristics of multicast routing, the algorithm was improved, which accelerated the convergence speed and enhanced the quality of solution. The proposed scheme is easy to realize by three types of ants with simple structures. The scheme is loop-free and has small overhead compared with the flooding method. Simulation results show that the scheme is an effective distributed approach to multicast routing decision with delay constraint

An improved pseudo-random binary sequence design for multivariable system identification(A16-395)

An improved design method of PRBS signals for multivariable system identification is proposed. The approach sets different delay intervals to different channels instead of a unique delay interval. The minimal periodic length of PRBS is nearly the sum of the settling time of each channel. Consequently, a modified correlation analysis method is also put forward to identify the process model of multivariable system. The simulation shows that the proposed method can effectively shorten the periodic length of PRBS for multivariable system identification

An improvement of continuous tabu search for global optimization

An improved continuous tabu search algorithm is developed for global optimization problems. In this tabu search algorithm, the aspiration level is introduced and the neighbor space of current solution is partitioned by a set of concentric hyperrectangles. The neighborhood of the current solution is generated by randomly selecting one point inside each concentric hyperrectangle and certain other points inside the central hyperrectangle. Experiments prove that the extra selection inside the central hyperrectangle can improve the performance of the tabu search algorithm.

Multivariable optimal control with adaptation mechanism in rudder/fin stabilizing system

This paper presents the design of a rudder/fin stabilizing system. The intercoupled relationships between yaw, roll and sway make it necessary to introduce a MIMO control system to obtain optimal performance. After discussion of the mathematical model of ship motion, an LQG controller based on Kalman filter is introduced to minimize the quadratic performance index. Taking into account the changing circumstances and the nonlinear characteristics of actuators, we apply adaptation mechanisms to adjust the weighting parameters and to reconstruct the optimal controller automatically. As can be seen from the numerical results, the newly designed rudder/fin stabilizing system shows much better performance over conventional ones.<<ETX>>

Performance Monitoring of Chemical Process Based on Multivariable Statistical Technology

Principal component analysis (PCA) can be effectively used to eliminate system noise and correlation between process variables but to reserve enough original data information. Based on principal component model, performance monitoring and analysis was carried out on control system with multivariate statistical index, such as Q residuals, Hotelling T2 and principal scores. This method with multivariable statistical technologies gives explicit knowledge about control system: whether the system is in-control or out-control, how disturbance and system variables influence the performance and how the control strategy work. Control system then could be evaluated with these knowledges. Application to two typical chemical processes shows the example of the method practical function

FIR frequency sampling filters design based on adaptive particle swarm optimization algorithm

Based on the study of Particle Swarm Optimization (PSO) on the mechanism of information communion, a new adaptive method of PSO is presented in this paper. This new adaptive method is to avoid the particles getting into local best solution during the optimization. By applying Adaptive Particle Swarm Optimization (APSO) to optimize transition sample values in FIR filter, the maximum stop band attenuation is obtained. The simulations of designing low-pass FIR have been done and the simulation results show that APSO is better than PSO not only in the optimum ability but also in the convergence speed.

A method of process monitoring based on blind source separation with denoising information by wavelet transform and its application to chemical process

In this paper, a new process monitoring method based upon wavelet transform and blind source separation is presented. Wavelet transform is employed to de-noise measured signals to remove the process noise, and blind source separation based on information maximization is used to extract blind source signals. Later, control limits and monitoring plots are built by estimating the probability distribution of every blind signal by means of Parzen density estimator. For investigating the feasibility of this method, its fault-detection performance is evaluated and compared with other process monitoring method based on blind source analysis with direct process information to a simple AR(1) process and a continuous stirred-tank-reactor process. The results show the superiority of the method presented in this paper over other process monitoring method, which has high faulty warnings and missing warnings.