COMPUTER AIDED ENGINEERING FOR REAL-TIME CONTROL SYSTEMS DESIGN AND RAPID PROTOTYPING

Abstract

A PC-based low cost intelligent Computer Aided Engineering (CAE) control system design software and hardware platform has been developed. This platform encompasses all the four phases for the design, namely, design, simulation, implementation and field-testing. It allows the whole design process to be cycled easily and efficiently, thus considerably speeding up the control system design and prototyping. The software of the platform is a Windows-based application and is targeted at a class of relatively slow process control problems. Three types of controller namely Proportional-Integral-Derivative (PID), Fuzzy Logic Control (FLC) and Iterative Learning Control (ILC) were incorporated in the platform. For each type, the design process is automated in such a way that it requires minimal operator's intervention. Besides, an auto-tuning method was included for the PID and ILC controllers. For the real-lime control, two hardware arrangements were made possible. One is through a plug-in data acquisition card where the PC controls the target plant. Another configuration is to use a stand-alone universal controller card to control the plant directly; the PC acts as a workstation for offline design and simulation. Previous work on the CAE platform has successfully implemented the PID and FLC components on the platform. The aim of this research is to explore the methods of designing and implementing the new ILC controller on the same platform. At the initial stage, the P-type ILC with Previous Cycle Feedback (PCF) and Current Cycle Feedback (CCF) were investigated. These schemes were implemented on the platform and tested on a first order plus dead-time system. It was shown that ILC with PCF may not be suitable for the above mentioned system due to the large initial tracking error, which in turn results in control signal saturation, and thus degrades the control performance. On the other hand, ILC with CCF performed much better but the appropriate gain had to be searched by trial and error. To overcome this difficulty, P and PI-type ILC which are based on well-established auto-tuning schemes were proposed to set the learning control gains. Experimental studies on the level control of a coupled tank system verified the effectiveness and implementability of the proposed ILC algorithms. A repetitive type control problem whereby the system is not allowed to restart from the same initial condition is studied with the aid of the CAE platform developed. In this case, Robust Learning Control (RLC), a variable structure learning control scheme (Xu et al.) was implemented to control an analog plant (Dual Process Simulator). Experimental works were conducted to verify the effectiveness of the RLC. The difficulties in realizing a customized hardware controller were also addressed in this thesis. The methodology of designing and construction of a micro controller based hardware system was investigated. The role of this CAE sub-system is to automate the entire digital design and implementation process of a prototyping control system. It is concluded that such a system requires a powerful database engine to handle the large pool of devices/components data. The timing issue in real-time control was studied. It is shown that using the timer message in Windows Operating System (WINOS), the best sampling time that can be achieved is only 55ms. Alternative methods such as bypassing the message queue in WINOS and using hardware interrupt are suggested to overcome the above limitation. InstallSheild®, an integrated installation system for software distribution to all Microsoft Windows platforms was used to prepare the software installation of the CAE package.