How to survive the oncoming train of technology

However, control engineers want to know the time response of the process variables to changes in controller outputs and disturbances. The first order plus dead time response for the loop in manual (open loop) is the simplest way of representing the dynamic response and can be used for the tuning of PID controllers or the set up of model predictive control. The overall open loop gain is the product of the valve, process and measurement gains;  the total loop dead time is the sum of the pure dead times (time delay) and small time constants (time lags); and the open loop time constant is the largest time constant in the loop, wherever it exists.

Process simulators typically don’t address mixing and transportation in delays in the response of the process and the sensor. They also don’t include lags or delays for different types of sensors or actuators, or the installed characteristic, deadband (backlash), and resolution (stick-slip) for various control valve and positioner designs. The thermal lags of coils and jackets may not even be simulated and non-equilibrium conditions are ignored.  In other words, the parameters most “high fidelity” process simulators focus on are the gain and primary time constant of a volume when the continuous process is up to speed.

Process simulators may not even get the process gain right because the manipulated variable used by the process engineer might be a heating or cooling rate, which neglects the dynamics associated with utility system design and interactions. The vapor is also assumed to be instantaneously in equilibrium with the liquid by means of a “flash” calculation. A dynamic process simulator often shows the temperature response of a perfectly mixed volume to an instantaneous change in heat input or removal rate for equilibrium conditions, which for columns and evaporators is after the mixture reaches its boiling point. Wouldn’t it be grand if real plants started up or responded this rapidly and ideally? Wouldn’t it be super if the severe discontinuities of split range valves and the lags and upsets from fighting utility streams were just a bad dream?

Taking Control
Control engineers who understand the importance of dead time and loop dynamics need to take the lead in the development of the models in dynamic simulations. A new definition of model fidelity is needed that emphasizes the dynamic response of a change in process variables to changes in manipulated or disturbance variables. It should cover unusual and non steady operating conditions during startup and batch sequencing, even though this may mean a sacrifice in the match between a given process input and output at design conditions. If we define and seek fidelity in terms of what a control system and operator sees and has to deal with, the promise of the virtual plant can be fulfilled.

(Click the Download Now button below for a .pdf version of the chart referred to in this article.)