Neural networks and evolutionary computation for real-time quality control of complex processes

Quality control in general and automated quality control in particular are assuming major importance in modern society as technological systems are becoming increasingly complex and highly interconnected. Traditional statistical process control techniques are inadequate to address control problems in automated processes because of the high degree of data correlation characterized by such processes. Classical process control methods are unsuitable for the control of complex, nonlinear, multivariable processes. This paper discusses on-going research to design, develop, and implement a prototype that addresses product quality by monitoring and modifying, in real-time, dynamic, multivariable processes. The prototype is first described in terms of its components and then current results from testing the prototype are presented. It is anticipated that, when completed, the autonomous control system prototype will provide high level adaptation to changes in the plant, environment, and control objectives. It is also expected to enable real-time quality control of complex processes that do not lend themselves to control by traditional methods. For example, computer chip manufacturing involves a high degree of automation, characterized by significant data correlation. The proposed prototype is designed to handle this situation by using memory and learning techniques to incorporate the correlation into the process model.