Search

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.ijpe.2005.03.012 Byline: Hongsheng Chu, Jingchun Wang, Yihui Jin, Hansheng Suo Abstract: We investigate a two-component assembly system facing stationary stochastic demand with constant component replenishment lead times. Holding cost are charged for each component inventory, which are managed by component suppliers; final product demand is backordered when cannot be fulfilled in time and a penalty cost is charged. We assume the long lead-time component supplier takes normal base-stock policy while the short lead-time component supplier takes a modified base-stock policy. Static game and Stackelberg games are considered. Equilibria are analyzed and the cost in equilibria are compared with the system optimal solution, difference shows the deterioration of system performance caused by decentralized inventory control. We try to design a mechanism to coordinate the decentralized system, that is to make decentralized system perform as well as system optimal solution though the decisions separated. We show that a liner transfer based on the backorder between the component suppliers qualifies this task. Author Affiliation: Department of Automation, Tsinghua University, 100084, China Article History: Received 23 March 2004; Accepted 24 March 2005 Article Note: (footnote) [star] Supported by the National Natural Science Foundation of China (No. 60174046).

A MIMO nonlinear adaptive predictive control strategy is presented in which the wavelet neural network based on a set of orthogonal wavelet functions is adopted. A nonlinear mapping from the network-input space to the wavelons output space in the hidden layer is performed firstly. Then, the output layer uses a linear structure. Its weight coefficients can be estimated by a linear least-squares estimation method. The excellent statistic properties of the weight parameter estimation can be obtained. Based on developed recursive algorithm, a MIMO nonlinear adaptive predictive control strategy is implemented. A simulated MIMO nonlinear process example shows that the control scheme is effective.

The proposed generalized minimum variance adaptive control system is based on an improved least squares estimator, a modified Kalman filter predictor and a state space predictive controller. The Kalman filter provides explicit estimates of the process states and predicted output y(k+d+l) and therefore simplifies controller design. The system is designed for MISO stochastic processes with time delays and non-zero mean disturbances. Examples show that the adaptive system is practical, easy to implement and that overall performance is equal to or better than that obtained using a PID controller or the self-tuning controller (STC) of Clarke and Gawthrop

This paper is a further discussion of Wang, et al (1995). First, the necessity for data reconstruction during the chemical process control has been discussed, and the basic concept of the wavelet transform is introduced. It is adopted to reconstruct the data with low sampling rate so that the data reconstructed can be used in process control or model identification. Then, the reasonableness of the method is proved. Simulation results are given to illustrate that the method is effective in model identification.

This paper describes a new multivariable bilinear self—tuning controller with a generalized minimum variance control technique. The controller can control a multivariable bilinear system whose linear part is not necessarily minimum phase. The theoretical proof of global convergence of the control algorithm is given.