The Problem
A leading manufacturer of Analytical instruments was looking for high precision temperature control for their gas chromatography based analysis instrument. To ensure high repeatability of test results, it was essential to maintain the oven temperature to within 0.05% of the set point.
The current control algorithm was helping them maintain it to within a band of 0.1%. Besides halving this tolerance band, it was also required to reduce the time taken for the oven temperature to stabilize.
The Solution
The oven had a typical asymmetric temperature control structure. The heating was active, by a heating element connected to controlled mains power, while the cooling was passive. The temperature sensing was done through a four wire RTD sensor kept at the far end of the heater within the oven chamber. The asymmetric structure gave a rapid rise in temperature while heating but a slow drooping tail as the temperature dropped. Multiple gas columns within the oven had varying gas flows at different temperatures and these would cause process disturbances in the oven.
A quick software model was constructed based on the oven characteristics. The impulse response of the actual oven was constructed from its step response. This impulse response was analyzed and an approximate second order system was modeled in the software. The software model helped to fine tune the control algorithm for the various timing parameters. These included a faster rising time, reducing the overshoot to zero, reduce the settling time and reduce the temperature band at steady state to within the tolerance band of 0.05%. The process disturbance was modeled too, to study the response of the control system to the actual gas flows in the oven.
It was quickly realized that the system cannot afford overshoot due to the time taken to bring the system back to set point. This meant heating up to the set point had to be fast but tightly controlled to avoid an overshot. The large thermal lag in the system made it a difficult system to control. This lag was partly compensated in a software filter. A feed forward mechanism was also constructed. This helped forecast the impact to the oven temperature due to varying gas flows through the columns. The compensation to this impact would commence well before the temperature would start to droop.
The algorithm was tested with a large set of instruments to understand how it would behave due to component variations, it was tested under multiple environment conditions where the external temperatures were varied through the supported temperature range of the instrument and under different initial conditions. Testability was built in to the software, where it tracked and logged the temperature deviations, the time taken to stabilize, the temperature band under which it is operating and also had built in test cases that could simulate difficult to create external conditions.
A combination of feed forward, software filters, PI software algorithm and a systematic verification helped achieve the objectives set out for this project.