June 26, 2005
Process Control
A while ago I wrote of process monitoring, and now the time has come for an article about how to control chemical/industrial processes.
Process control implies to monitor the process paramenters, compare them with the setpoints, decide if action is required and take the appropriate action. This cycle can be performed in different ways: with simple electrical circuits (or even mechanical devices); with more complex electronics and finally with complex computerized systems that can take into account several parameters and their interactions. Sometimes humans are included in the loop, but there are drawbacks: the reponse time of a human is slow, and watching indicators waiting for an alarm is an incredibily tedious job, that can easily cause attention drops. Electronic circuits never get bored, instead.
The simplest system, conceptually, is an electric heater, so let's deal with it first.
An electric heater is a resistor, a coil of high resistivity wire or a rod of SiC, that will produce a great amount of heat when electric current passes through it.
The simplest control system is an on/off thermostat, such as the ones used for central heating: in this case, the heating elements are either on at full power when the temperature is below the setpoint, and off from the setpoint upward. This system, when used with electric heaters, is too crude: temperature initially overshoots sensibly the setpoint. So, better electronic controllers have been built; these will turn off the power before reaching the setpoint in order to reduce the overshoot.
But the most advanced controllers are called PID: proportional, integral, derivative. These can adjust the outpust power from 0 to 100% on the basis of the variation with time of the error - defined as the difference between setpoint and measurement (this is accomplished using electronics circuits for data processing and particular transormers and solid-state relays for power regulation). For example, my furnace has a PID controller, and when I turn it on I can notice that the heating power is 100% initially, then it decreases when the temperature gets closer to the setpoint and when the setpoint is reached only short bursts of low power are required to mantain the correct temperature. The internal parameters of the controller can be adjust manually or automatically for that specific furnace, reactor and process conditions (nature and flow of the reactants) and when this tuning process is completed, the furnace will work optimally.
If the system to control is a heat exchanger (another device widely used in industry), the controller(s) will open and close metering valves in order to adjust the flow of heating or cooling fluid. Similarly, the amount of heat produced by a gas burner can be regulated varying the flow of gas feed.
In mass flow controllers, the signal from the flow measuring element is processed by electronic circuits, or software for certain models, and compared with the setpoint (that can be entered by turning a potentiometer or typing a value in the software): an output signal is thus generated, and this signal pilotes a metering valve that will provide the correct flow.
Instead, pressure regulators are mostly mechanical: semplifying, the pressure of the gas (or liquid) acts against a regulable spring, and the valve opens only when the forces are balanced. Pressure regulators, back pressure regulators and relief valves all use variations of the basic design. There are also electronic pressure regulators, in which a valve is opened electrically according to the signal from a pressure transducer.
All control systems have a response time that is never zero: the electrical signals need time to travel along the wires, then to be processed and to travel back to the actuator or device in the plant. This is a minor problem for research applications where cables are no more than a few meters long, but in industry, where cables (called fieldbuses) are easily longer than 100 m the transmission time needs to be taken into account - not to mention the problem of electromagnetic interference, that is bigger for longer cables*.
All this and other factors cause the measurement to oscillate around the setpoint; good controllers will rapidly damp the oscillations to a minimal level. Less good ones cannot damp the oscillations completely, and in unfurtunate cases a positive feedback loop can occur and the oscillations will actually be reinforced with potentially damaging consequences.
Designing control systems is quite a difficult job, and while small research reactors can succesfully be controlled with off-the-shelf instruments (like this), petrochemical and other plants need custom-made ones, which also comprehend dedicated computers and fail-safe software; not your basic Windows (or Linux, for that matter) prone to crash too often. Just to name a few names, the big companies in this business are the likes of Honeywell and Siemens.
*Indeed, modern refinery control systems use fiber optics for data transmission, because they are resistant to EMI and do not constitute a fire/explosion hazard. Instead, traditional copper conductors must run inside nitrogen-filled steel pipes (to keep flammable gases and vapours out), junction boxes and switchboxes in order to reduce the explosion hazard - one can only imagine the cost and complexity of that.
Process control implies to monitor the process paramenters, compare them with the setpoints, decide if action is required and take the appropriate action. This cycle can be performed in different ways: with simple electrical circuits (or even mechanical devices); with more complex electronics and finally with complex computerized systems that can take into account several parameters and their interactions. Sometimes humans are included in the loop, but there are drawbacks: the reponse time of a human is slow, and watching indicators waiting for an alarm is an incredibily tedious job, that can easily cause attention drops. Electronic circuits never get bored, instead.
The simplest system, conceptually, is an electric heater, so let's deal with it first.
An electric heater is a resistor, a coil of high resistivity wire or a rod of SiC, that will produce a great amount of heat when electric current passes through it.
The simplest control system is an on/off thermostat, such as the ones used for central heating: in this case, the heating elements are either on at full power when the temperature is below the setpoint, and off from the setpoint upward. This system, when used with electric heaters, is too crude: temperature initially overshoots sensibly the setpoint. So, better electronic controllers have been built; these will turn off the power before reaching the setpoint in order to reduce the overshoot.
But the most advanced controllers are called PID: proportional, integral, derivative. These can adjust the outpust power from 0 to 100% on the basis of the variation with time of the error - defined as the difference between setpoint and measurement (this is accomplished using electronics circuits for data processing and particular transormers and solid-state relays for power regulation). For example, my furnace has a PID controller, and when I turn it on I can notice that the heating power is 100% initially, then it decreases when the temperature gets closer to the setpoint and when the setpoint is reached only short bursts of low power are required to mantain the correct temperature. The internal parameters of the controller can be adjust manually or automatically for that specific furnace, reactor and process conditions (nature and flow of the reactants) and when this tuning process is completed, the furnace will work optimally.
If the system to control is a heat exchanger (another device widely used in industry), the controller(s) will open and close metering valves in order to adjust the flow of heating or cooling fluid. Similarly, the amount of heat produced by a gas burner can be regulated varying the flow of gas feed.
In mass flow controllers, the signal from the flow measuring element is processed by electronic circuits, or software for certain models, and compared with the setpoint (that can be entered by turning a potentiometer or typing a value in the software): an output signal is thus generated, and this signal pilotes a metering valve that will provide the correct flow.
Instead, pressure regulators are mostly mechanical: semplifying, the pressure of the gas (or liquid) acts against a regulable spring, and the valve opens only when the forces are balanced. Pressure regulators, back pressure regulators and relief valves all use variations of the basic design. There are also electronic pressure regulators, in which a valve is opened electrically according to the signal from a pressure transducer.
All control systems have a response time that is never zero: the electrical signals need time to travel along the wires, then to be processed and to travel back to the actuator or device in the plant. This is a minor problem for research applications where cables are no more than a few meters long, but in industry, where cables (called fieldbuses) are easily longer than 100 m the transmission time needs to be taken into account - not to mention the problem of electromagnetic interference, that is bigger for longer cables*.
All this and other factors cause the measurement to oscillate around the setpoint; good controllers will rapidly damp the oscillations to a minimal level. Less good ones cannot damp the oscillations completely, and in unfurtunate cases a positive feedback loop can occur and the oscillations will actually be reinforced with potentially damaging consequences.
Designing control systems is quite a difficult job, and while small research reactors can succesfully be controlled with off-the-shelf instruments (like this), petrochemical and other plants need custom-made ones, which also comprehend dedicated computers and fail-safe software; not your basic Windows (or Linux, for that matter) prone to crash too often. Just to name a few names, the big companies in this business are the likes of Honeywell and Siemens.
*Indeed, modern refinery control systems use fiber optics for data transmission, because they are resistant to EMI and do not constitute a fire/explosion hazard. Instead, traditional copper conductors must run inside nitrogen-filled steel pipes (to keep flammable gases and vapours out), junction boxes and switchboxes in order to reduce the explosion hazard - one can only imagine the cost and complexity of that.
Comments:
Post a Comment
