Chapter Highlights
1.5 Control System Architectures
Sensors and actuators are important components in the instrumentation of a control system. The control-ler, which is an essential part of any control system, makes the plant (i.e., the process that is being con-trolled) behave in a desired manner, according to some specifications. The overall system that includes at least the plant and the controller is called the control system. The system can be quite complex and may be subjected to known or unknown excitations (i.e., inputs), as in the case of an aircraft.
Some useful terminology related to a control system is listed as follows:
• Plant or process: System to be controlled
• Inputs: Commands, driving signals, or excitations (known, unknown)
• Outputs: Responses of the system
• Sensors: Devices that measure system variables (excitations, responses, etc.)
• Actuators: Devices that drive various parts of the system
• Controller: Device that generates control signal
• Control law: Relation or scheme according to which the control signal is generated
• Control system: At least the plant and the controller (may include sensors, signal conditioning, and other components as well)
• Feedback control: Control signal is determined according to plant response
• Open-loop control: Plant response is not used to determine the control action
• Control: Control signal is determined according to plant excitation or a model of the plant In Figure 1.6, we have identified key components of a feedback control system. Several discrete blocks are shown, depending on the various functions that take place in a typical control system. In a practical control system, this type of clear demarcation of components might be difficult; one piece of hardware might perform several functions, or more than one distinct unit of equipment might be associated with one function. Embedded systems in particular may have distributed multifunctional components where demarcation of the functional blocks will be difficult. Nevertheless, Figure 1.6 is useful in understand-ing the architecture of a general feedback control system. In an analog control system, control signals are continuous-time variables generated by analog hardware; no signal sampling or data encoding is involved in a digital control system.
The control problem can become challenging due to such reasons as
• Complex system (many inputs and many outputs, dynamic coupling, nonlinear, time-varying parameters, etc.)
• Rigorous performance specifications
• Unknown or unmeasurable excitations (unknown inputs/disturbances/noise)
• Unknown or unmeasurable responses (unmeasurable state variables and outputs, measurement errors and noise)
• Unknown dynamics (incompletely known plant)
Since the operation of a control system is based on a set of performance specifications, it is important to identify key performance characteristics that a good control system should possess. In particular, the following performance requirements are important:
1. Sufficiently stable response (stability): Specifically, the response of the system to an initial-condition excitation should decay back to the initial steady state (asymptotic stability). The response to a bounded input should be bounded (bounded-input–bounded-output, BIBO, stability).
2. Sufficiently fast response (speed of response or bandwidth): The system should react quickly to a control input or excitation.
3. Low sensitivity to noise, external disturbances, modeling errors and parameter variations (sensi-tivity and robustness).
4. High sensitivity to control inputs (input sensitivity).
5. Low error: For example, tracking error and steady-state error (accuracy).
6. Reduced coupling among system variables (cross sensitivity or dynamic coupling).
As listed here, some of these specifications are rather general. Table 1.2 summarizes typical performance requirements for a control system. Some requirements might be conflicting. For example, fast response
Sensor/
FIGURE 1.6 Key components of a feedback control system.
is often achieved by increasing the system gain, and increased gain increases the actuation signal, which has a tendency to destabilize a control system. Note further that what is given here are primarily quali-tative descriptions for good performance. In designing a control system, however, these descriptions have to be specified in a quantitative manner. The nature of the used quantitative design specifications depends considerably on the particular design technique that is employed. Some of the design specifica-tions are time-domain parameters and the others are frequency-domain parameters.
1.5.1 Feedback and Feedforward Control
As seen earlier, in a feedback control system, the control loop has to be closed, where the system response is sensed and employed to generate the control signals. Hence, feedback control is also known as closed-loop control.
If the plant is stable and is completely and accurately known, and if the inputs to the plant can be precisely generated (by the controller) and applied, then accurate control might be possible even with-out feedback control. Under these circumstances, a measurement system is not needed (or at least not needed for feedback) and thus, we have an open-loop control system. In open-loop control, we do not use current information on system response to determine the control signals. In other words, there is no feedback. Even though a sensor is not explicitly needed in an open-loop architecture, sensors may be employed within an open-loop system to monitor the applied input, the resulting response, and possible disturbance inputs.
The significance and importance of sensors and actuators hold regardless of the specific control sys-tem architecture that is implemented in a given application. We now outline several architectures of control system implementation while indicating the presence of sensors and actuators in them.
Even in a feedback control system, there may be inputs that are not sensed and used in feedback control. Some of these inputs might be important variables for the plant, but more commonly they are undesirable inputs, such as external disturbances, which are unwanted yet unavoidable. Generally, the performance of a control system can be improved by measuring these (unknown) inputs and somehow using the information to generate control signals. In feedforward control, unknown inputs are mea-sured and that information, along with desired inputs, is used to generate control signals that can reduce errors due to these unknown inputs or variations in them. The reason for calling this method feedfor-ward control stems from the fact that the associated measurement and control (and compensation) both take place in the forward path of the control system. Note: In some types of feedforward control, the input signal is generated by using a model of the plant (and may not involve sensing).
TABLE 1.2 Performance Specifications for a Control System
Attribute Desired
Value Objective Specifications
Stability level High Response does not grow without limit
and decays to the desired value Percentage overshoot, settling time, pole (eigenvalue) locations, time constants, phase and gain margins, damping ratios Speed of response Fast Plant responds quickly to inputs/
excitations Rise time, peak time, delay time, natural frequencies, resonant frequencies, bandwidth
Steady-state error Low Offset from the desired response is
negligible Error tolerance for a step input Robustness High Accurate response under uncertain
conditions (input disturbances, noise, model error, etc.) and under parameter variation
Input disturbance/noise tolerance, measurement error tolerance, model error tolerance
Dynamic interaction Low One input affects only one output Cross-sensitivity, cross-transfer functions
As a practical example, consider the natural gas home heating system shown in Figure 1.7a. A simpli-fied block diagram of the system is shown in Figure 1.7b. In conventional feedback control, the room temperature is measured and its deviation from the desired temperature (set point) is used to adjust the natural gas flow into the furnace. On/off control through a thermostat is used in most such applications.
Even if proportional or three-mode (proportional-integral-derivative or PID) control is employed, it is not easy to steadily maintain the room temperature at the desired value if there are large changes in other (unknown) inputs to the system, such as water flow rate through the furnace, temperature of water entering the furnace, and outdoor temperature.
(a)
Pilot flame detector
Valve actuator Natural
gas Valve
Cold water in Thermal insulation
Vent
Exhaust gases
Mainflame
Thermostat
Thermocouple
Temperature controller (on/off) and transmitter
Hot water
out Room
radiator Burner
chamber Water
Pilot flame
Room temperature (output) Temperature
set point
(input) Controller Furnace
Sensor-transducer Unknown
inputs w1 w2 w3
(b)
w1= Water flow rate
w2= Temperature of cold water into furnace w3= Temperature outside the room
–
FIGURE 1.7 (a) Natural gas home heating system and (b) a block diagram representation of the system.
Better results can be obtained by measuring these disturbance inputs and using that information in generating the control action. This is feedforward control. Note that in the absence of feedforward con-trol, any changes in the inputs w1, w2, and w3 in Figure 1.7 would be detected only through their effect on the feedback signal (i.e., room temperature). Hence, the subsequent corrective action can consider-ably lag behind the cause (i.e., changes in wi). This delay will lead to large errors and possible instability problems. With feedforward control, information on the disturbance input wi will be available to the controller immediately, and its effect on the system response can be anticipated, thereby speeding up the control action and also improving the response accuracy. Faster action and improved accuracy are two very desirable effects of feedforward control.
1.5.2 Digital Control
In digital control, a digital computer serves as the controller. Virtually any control law may be pro-grammed into the control computer. Control computers have to be fast and dedicated machines for real-time operations where processing has to be synchronized with plant operation and actuation requirements. This requires a real-time operating system. Apart from these requirements, control com-puters are basically not different from general-purpose digital comcom-puters. They consist of a processor to perform computations and to oversee data transfer; memory for program and data storage during pro-cessing; mass-storage devices to store information that is not immediately needed; and input or output devices to read in and send out information.
Digital control systems might use digital instruments and additional processors for actuating, signal-conditioning, or measuring functions. For example, a stepper motor that responds with incre-mental motion steps when driven by pulse signals can be considered as a digital actuator. Furthermore, it usually contains digital logic circuitry in its drive system. Similarly, a two-position solenoid is a digi-tal (binary) actuator. Digidigi-tal flow control may be accomplished using a digidigi-tal control valve. A typical digital valve consists of a bank of orifices, each sized in proportion to a place value of a binary word (2i, i = 0, 1, 2, …, n). Each orifice is actuated by a separate rapid-acting on/off solenoid. In this manner, many digital combinations of flow values can be obtained. Direct digital measurement of displace-ments and velocities can be made using shaft encoders. These are digital transducers that generate coded outputs (e.g., in binary or gray-scale representation) or pulse signals that can be coded using counting circuitry. Such outputs can be read in by the control computer with relative ease. Frequency counters also generate digital signals that can be fed directly into a digital controller. When measured signals are in the analog form, an analog front-end is necessary to interface the transducer and the digital controller. Input/output interface cards that can take both analog and digital signals are avail-able with digital controllers.
Analog measurements and reference signals have to be sampled and encoded before digital process-ing within the controller. Digital processprocess-ing can be effectively used for signal conditionprocess-ing as well.
Alternatively, digital signal processing (DSP) chips can function as digital controllers. However, ana-log signals have to be preconditioned using anaana-log circuitry before digitizing in order to eliminate or minimize problems due to aliasing distortion (high-frequency components above half the sampling frequency appearing as low-frequency components) and leakage (error due to signal truncation) as well as to improve the signal level and filter out extraneous noise. The drive system of a plant typically takes in analog signals. Often, the digital output from the controller has to be converted into ana-log form for this reason. Both anaana-log-to-digital conversion (ADC) and digital-to-anaana-log conversion (DAC) can be interpreted as signal-conditioning (modification) procedures. If more than one output signal is measured, each signal will have to be conditioned and processed separately. Ideally, this will require separate conditioning and processing hardware for each signal channel. A less expensive (but slower) alternative would be to time-share this expensive equipment by using a multiplexer. This device will pick one channel of data from a bank of data channels in a sequential manner and connect it to a common input device.
The current practice of using dedicated, microprocessor-based, and often decentralized (i.e., distrib-uted) digital control systems in industrial applications can be rationalized in terms of the major advan-tages of digital control. The following are some of the important considerations.
1. Digital control is less susceptible to noise or parameter variation in instrumentation because data can be represented, generated, transmitted, and processed as binary words, with bits possessing two identifiable states.
2. Very high accuracy and speed are possible through digital processing. Hardware implementation is usually faster than software implementation.
3. Digital control systems can handle repetitive tasks extremely well, through programming.
4. Complex control laws and signal-conditioning algorithms that might be impractical to imple-ment using analog devices can be programmed.
5. High reliability in operation can be achieved by minimizing analog hardware components and through decentralization using dedicated microprocessors for various control tasks.
6. Large amounts of data can be stored using compact, high-density data-storage methods.
7. Data can be stored or maintained for very long periods of time without drift and without getting affected by adverse environmental conditions.
8. Fast data transmission is possible over long distances without introducing excessive dynamic delays and attenuation, as in analog systems.
9. Digital control has easy and fast data retrieval capabilities.
10. Digital processing uses low operational voltages (e.g., 0–12 V dc).
11. Digital control is cost-effective.
1.5.3 Programmable Logic Controllers
There are many control systems and industrial tasks that involve the execution of a sequence of steps, depending on the state of some elements in the system and on some external input states. A program-mable logic controller (PLC) is essentially a digital-computer-like system that can properly sequence a complex task, consisting of many discrete operations and involving several devices, that needs to be carried out in a particular order. The process operation might consist of a set of two-state (on–off) actions, which the PLC can sequence in the proper order and at correct times. PLCs are typically used in factories and process plants, to connect input devices such as switches to output devices such as valves, at high speed at appropriate times in a task, as governed by a program (ladder logic). Examples of such tasks include sequencing the production line operations, starting a complex process plant, and activat-ing the local controllers in a distributed control environment.
In the early days of industrial control solenoid-operated electromechanical relays, mechanical timers, and drum controllers were used to sequence such operations. Today’s PLCs are rugged computers. An advantage of using a PLC is that the devices in a plant can be permanently wired, and the plant opera-tion can be modified or restructured by software means (by properly programming the PLC) without requiring hardware modifications and reconnection.
Internally, a PLC performs basic computer functions such as logic, sequencing, timing, and counting.
It can carry out simpler computations and control tasks such as PID control. Such control operations are called continuous-state control, where process variables are continuously monitored and made to stay very close to desired values. There is another important class of controls, known as discrete-state control (or, discrete-event control), where the control objective is for the process to follow a required sequence of states (or steps). In each state, however, some form of continuous-state control might be operated, but it is not quite relevant to the task of discrete-state control. PLCs are particularly intended for accomplish-ing discrete-state control tasks.
As an example for PLC application, consider an operation of turbine blade manufacture. The discrete steps in this operation might be as follows:
1. Move the cylindrical steel billets into furnace.
2. Heat the billets.
3. When a billet is properly heated, move it to the forging machine and fixture it.
4. Forge the billet into shape.
5. Perform surface finishing operations to get the required aerofoil shape.
6. When the surface finish is satisfactory, machine the blade root.
Note that the entire task involves a sequence of events where each event depends on the completion of the previous event. In addition, it may be necessary for each event to start and end at specified time instants. Such time sequencing would be important for coordinating the current operation with other activities, and perhaps for proper execution of each operational step. For example, activities of the parts handling robot have to be coordinated with the schedules of the forging machine and milling machine.
Furthermore, the billets have to be heated for a specified time, and machining operation cannot be rushed without compromising the product quality, tool failure rate, safety, and so on. The task of each step in the discrete sequence might be carried out under continuous-state control. For example, the milling machine would operate using several direct digital control (DDC) loops (say, PID control loops), but discrete-state control is not concerned with this except for the starting point and the end point of each task.
A schematic representation of a PLC is shown in Figure 1.8. A PLC operates according to some logic sequence programmed into it. Connected to a PLC are a set of input devices (e.g., pushbuttons, limit switches, and analog sensors such as RTD temperature sensors, diaphragm-type pressure sensors, piezo-electric accelerometers, and strain-gauge load sensors) and a set of output devices (e.g., actuators such as dc motors, solenoids, and hydraulic rams, warning signal indicators such as lights, alphanumeric
Push button
Display screen
Power supply
Keyboard Memory (ROM, RAM)
Processor
Interface hardware Interface hardware
Input devices Output devices
Limit switch
Temperature sensor (RTD)
Load sensor
Motor
Solenoid
Valve actuator
Indicator light
Alarm
PLC
FIGURE 1.8 Schematic representation of a PLC.
LED displays and bells, valves, and continuous control elements such as PID controllers). Each device is assumed to be a two-state device (taking the logical value 0 or 1). Now, depending on the condition of each input device and according to the programmed-in logic, the PLC will activate the proper state (e.g., on or off) of each output device. Hence, the PLC performs a switching function. Unlike the older generation of sequencing controllers, in the case of PLC, the logic that determines the state of each out-put device is processed using software, and not by hardware elements such as hardware relays. Hardware switching takes place at the output port, however, for turning on or off the output devices controlled by the PLC.
1.5.3.1 PLC Hardware
As noted earlier, a PLC is a digital computer that is dedicated to perform discrete-state control tasks. A typical PLC consists of a microprocessor, RAM and ROM memory units, and interface hardware, all interconnected through a suitable bus structure. In addition, there will be a keyboard, a display screen, and other common peripherals. A basic PLC system can be expanded by adding expansion modules (memory, I/O modules, etc.) into the system rack.
A PLC can be programmed using a keyboard or touch-screen. An already developed program could be transferred into the PLC memory from another computer or a peripheral mass-storage medium such as hard disk. The primary function of a PLC is to switch (energize or de-energize) the output devices connected to it, in a proper sequence, depending on the states of the input devices and according to the
A PLC can be programmed using a keyboard or touch-screen. An already developed program could be transferred into the PLC memory from another computer or a peripheral mass-storage medium such as hard disk. The primary function of a PLC is to switch (energize or de-energize) the output devices connected to it, in a proper sequence, depending on the states of the input devices and according to the