| Abstract | This thesis is the result of work carried out during more than two years on a Teaching Company Scheme. Liaison took place between Rhone-Poulenc Agriculture Limited (the industrial partner), hereafter referred to as RPAL or the company, and Middlesex University (the academic partner). The aim of the Scheme was to realise the design, development, commissioning, testing and validation of an intelligent robotic system for sample analysis of trace
pesticides and metabolites in order to enable quicker product development. Due to the complexity of the project and the range of technical expertise and skills needed for its implementation, three associates participated in the
Programme. I joined as the second associate. With my degree in Industrial Engineering, I have been in overall charge of developing the computational aspects of the system, from control overview to implementation and validation.
Two distinct types of studies will be carried out with the robot based system:
• Routine extraction of pesticide from soil or plant material, which is compound as well as analyst dependant.
• Method development studies, to optimise those routine extraction processes.
Traditional strategies of control were not applicable for such system because we were dealing with the automation of a non repetitive process involving non-deterministic operations (evaporation, filtration, etc.). The resulting control system should provide a high degree of flexibility to allow workcell
reconfiguration without involving any reprogramming. Modularity is also a must if expansion and upgrading to new technologies and equipment is to involve relatively little cost and effort. In addition, all generated data has to be stored and reported following Good Laboratory Practice (GLP) standards.
As the system is both large and flexible in operation, it has proven a real challenge to develop. Software had to be written that can - among its many tasks - allow unrestricted analyst choice, optimise system performance, detect, prioritise and act upon error signals, dynamically schedule robot and
instrument operation in real time, trace samples as they pass through the system and generate results as reports stored in databases.
The system is now virtually complete, and is presently undergoing the last stages of the validation. Due to the success of this scheme, further cooperative ventures are being planned between Rhone-Poulenc and Middlesex University in both the UK and France. |
|---|