2015 AIChE Spring Meeting and 11th Global Congress on Process Safety
(212a) Control Strategies and Parameters to Improve Reactor Performance
There is an incredible difference in process dynamics, control strategies and tuning rules based on the type of reactor. Considerable disagreement has consequently arisen as to the tuning settings and the importance of the dynamics of automation system components. Each control strategy is considered special for the particular production unit and improvements by simple configuration changes are often not recognized.
The type of dynamic response and the source of the dynamics drastically changes with the type of equipment. Reactors can be characterized to having a dead time dominant, moderate self-regulating, near-integrating, true integrating, and runaway response. The dead time can range from less than a second to over an hour. These differences have enormous implications as to the importance of automation system dynamics and the best tuning rules and control strategies.
Reactors can be categorized as the type of dynamic response and relative dead time expected based on the reactor type. Dead time dominant responses are rare if you include the dynamics associated the manipulated variable (MV) and the sensor for the process variable (PV) being controlled. Dead time dominance is largely relegated to plug flow reactors with slow at-line analyzers. Normally, plug flow liquid and gas reactors are found to be have a moderate self-regulating response due to heat transfer surface and thermowell lags. Liquid back mixed reactors have a near-integrating, true integrating, and runaway response depending upon degree and sign of internal feedback within the process. Endothermic or moderately exothermic liquid reactors tend to have a near-integrating response (nearly zero process feedback) for continuous operation and a true integrating response (truly zero process feedback) for batch operation. Highly exothermic reactors can have a runaway response (positive process feedback).
The set of tuning rules and performance objectives change as a reactor moves from having a dead time dominant to a moderate self-regulating response. The changes correspond to a decrease in negative feedback within the process and possible amplification of noise and interactions.
The objective for dead time dominant and moderate self-regulating process may well be to minimize MV movement and MV overshoot of the final resting value. This is accomplished by tuning rules that minimize proportional and derivative action and maximize integral action.
The objectives for near-integrating, true integrating, and runaway processes are to minimize the peak and integrated error for load disturbances and to minimize overshoot and possibly rise time for setpoint changes. By definition this necessitates significant MV movement and MV overshoot. The tuning rules maximize the use of proportional and derivative action and minimize the use of integral action. The 2 degrees of freedom (2DOF) PID structure or a setpoint lead-lag is used to prevent overshoot.
The effect of a secondary lag (second largest time constant) increases as negative feedback is decreased and positive feedback is increased within the process. There is a window of allowable controller gains where a proportional mode gain too small as well as too large can cause severe oscillations. Equations are detailed that quantify the maximum and minimum controller gain based on the rate of change of the process variable and the relative size of a secondary lag.
Inferential measurements of conversion rate and composition are also given that are based on first principles and common sensors for batch and continuous reactor control and performance analysis.
Control strategies are presented based on reactor type that will inherently maximize production rate or maintain reaction stoichiometry. The production rate of plug flow gas reactors can be inherently maximized by the temperature controller manipulating reactant feed rates. Reaction stoichiometry for two phase continuous reactors can be inherently maintained by pressure and level control when the product and reactants are in different phases (e.g., gas reactant flow manipulated by pressure control for liquid product flow and gas purge flow). When reactant is recycled, various control strategies are offered to prevent the snowballing effect depending upon the source and path of the recycle.
When maximizing production rate and maintaining reaction stoichiometry is not inherent, control strategies are presented along with PID tuning and the key PID feature required to provide the feedback control needed. The PID feature is found to provide the output response needed to deal with discontinuous signals from analyzers and valves. The feature also enables directional move suppression needed to provide a slow gradual optimization and a fast recovery for large fast disturbances and abnormal operation.
Where possible the control strategies are generalized for fed-batch and continuous operations. In cases where the batch response is unidirectional, a PID structure without integral action is used. Alternately, the controlled variable may be translated to the rate of change of a key process variable to enable the use of integral action and the control of the batch profile and end point. A future process variable value is estimated to make decisions as to whether the batch should be terminated or extended.
MYNAH Tecnologies
CDI Process & Industrial