An intelligent controller for process automation

Abstract

This paper presents a novel supervisory controller that incorporates both a procedural and rule-based language and is capable of responding to asynchronous real-time sensor interrupts. This supervisory controller has been used for automating the manual operations in GaAs crystal growth in a liquid encapsulated Czochralski (LEC) process-based puller from Cambridge Instruments (CI-358). Although commerical crystal-growth controllers provide some degree of automation to help in relieving the burden on an operator, it is essentially the skill and experience of an operator that determines the quality of the crystals grown. Thus, reproducibility of the process is limited, thereby limiting the quality of crystals grown. This controller, referred to as theMeta-Controller (MC), executes ‘plans’ to sequence the operations of crystal growth. Plans are structured English-like representations of scripts that operators follow to grow crystals. Plans are then executed by the Meta-Controller, which responds to asynchronous sensor data interrupts in real time, and issues actuator commands to a real-time controller module. The notion of time is explicitly incorporated into the syntax of the language. Software structures referred to aslogical sensors and logical actuators perform the translation from numeric to symbolic values and vice-versa. The Meta-Controller has been successfully demonstrated on an LEC puller and has automated three phases of crystal growth, including the ‘seed-on’ phase which is currently the most manual-intensive operation in GaAs crystal growth. The other two phases are ‘synthesis’ and ‘meltback’. More plans are currently being written to automate the entire process of crystal growth.