Multipoint temperature control.

摘要

Millikelvin temperature control is critical to many applications. This dissertation is to demonstrate millikelvin temperature stability at 293 K of an aluminum plate by using thermoelectric devices (TED) and thermistors through the application of modern multivariable control algorithms despite the presence of uncertainties in the system model. The results and control methods can be extended to large structures, such as metrology frames in lithography steppers. As a first step, optimal sensor and actuator locations are found for two dimensional distributed parameter systems of parabolic type to achieve spatial uniformity of temperature on the plate. By the application of finite integral transforms, a reduced order, lumped parameter system is obtained. The optimal sensor locations are then determined by minimizing the Frobenius norm of the error covariance matrix of the Kalman filter. Further, using the notion of positive real systems, it is shown that the best locations for measurement are also the best locations for actuation.; In the second step, an experimental test bed was built, and several control algorithms, P1, PID both using Ziegler-Nichols and Chien-Hrones-Reswick methods, LQR and mu controllers, were implemented through Labview(TM) on the test bed. Comparisons were then made between the performances of control algorithms, and then mu analysis was done on the closed loop system to determine the robustness properties of different controllers. It was determined that multivariable implementation of Ziegler-Nichols PID controller was the most effective of all the controllers, as well as having one of the best robustness properties. Thus, the contributions of this dissertation are: (1) Experimental demonstration of millikelvin temperature stability of an aluminum structure by active control and comparison of performance and robustness of the different control algorithms for precision temperature control. (2) Extension of an existing methodology for optimal placements of sensors and actuators on 2D structure and study the optimal locations by varying the order and sampling time of the system.