The development of an expert system and adaptive process models for hot mill setup
thesisposted on 15.12.2014, 10:37 by Ian. Robinson
A study was performed to develop new techniques for rolling mill setup and supervisory control. The study was based around three main components, namely a mathematical model of the process, its associated adaptation and an expert system. A novel architecture was developed to integrate the three components into a setup control system, along with some additional functions. The objective of the mill setup system is to determine the optimum mill actuator set points and control targets prior to the rolling of the slab. It is the function of the mill setup and supervisory control system to ensure that the material produced is of primary quality and that a high productivity level is achieved.;The novel control architecture incorporates three main components. Firstly, process models are used to predict the states of the rolling process prior to rolling. These models predict the rolling load, motor power and strip temperature, thermal camber of the work rolls, deflection of the mill stack and the profile and shape of the strip. Adaptation ensures that there is a good agreement between measurements and the model predictions. The adaptation is split into two main levels. A Kalman filter is used to predict short term errors in the process model from one pass to the next. Long term variations in the process are tracked using the recursive least squares algorithm. Finally the expert system is used to schedule the mill, diagnose possible faults occurring within the process and to supervise the activities of the other components in the control system.;The system is demonstrated in simulation and comparisons are made with and without the expert system control. The results show that there are distinct improvements to be gained with the application of artificial intelligence to an industrial control problem, in this case a hot aluminium rolling mill.