DCS is short for Distributed Control System. It refers to a system where control is decentralized, management is centralized, and display is centralized. In the late 1960s, Programmable Logic Controllers (PLCs) were developed for logic operations, primarily used in the automotive industry. In the mid-1970s, DCS systems capable of analog control were introduced to the market, replacing analog instrument control based on PID calculations. The concept of DCS was first proposed by manufacturers of instruments, initially primarily used in the chemical industry. Later, the computer industry also began developing DCS systems. In the 1970s, microcomputer technology was still immature, and computer technology was not yet advanced. Operator stations, controllers, I/O boards, and network interface boards were all developed in-house by DCS manufacturers; all components were proprietary.
In the early 1970s, some people replaced the original centralized analog instrument control with minicomputers such as the PDP/1124. This resulted in numerous cables connecting to the central control room. If the minicomputer served as both the controller and the CRT connected to it as a display device (i.e., human-machine interface), a single minicomputer would need to receive signals from thousands of transmitters or other sensors and complete calculations for hundreds of loops. Clearly, the risks were concentrated. The number of cables connected to the analog instruments was as numerous as before, and if the minicomputer failed, control and display were lost. Digital control failed to achieve its intended purpose.
Later, some proposed separating control and display. One computer handles the control calculations, while another handles the display. Furthermore, a technological process, as the controlled object, may require many points for display and control, some of which require closed-loop control or logical operations. Since the various parts of the controlled process are relatively independent, they can be divided into several independent processes. The input and output points requiring display and control at these independent processes in the computer control system can then be distributed across several computers. The computational tasks originally performed by a single small computer can now be completed by several or dozens of computers (controllers). The failure of one machine will not affect the entire system. This is the so-called "wolf pack replacing tiger" tactic, which embodies the concept of risk dispersion. Display, operation, printing, and other management functions are centralized, and a network connects the control and display parts into a single system. At the time, some called this a distributed system.
How much risk should be distributed? This depends on the level of computer technology at the time. In the mid-1970s, complete distribution meant one controller completing the calculations for one loop. Because people were not very familiar with digital technology and were accustomed to analog instruments, loop controllers were popular in the late 1970s and 1980s. These digital controllers were made almost identical in appearance to the original analog instruments, without changing operating habits, and internally digitized PID calculations. One instrument (one computer) completed the control task for one loop. They were relatively expensive, but the risk was distributed. Then, communication networks were used to connect the various controllers and CRT-based human-machine interfaces into a system. At this time, the network structure was usually a star topology. The manufacturing cost of loop controllers was too high, and the price/performance ratio was poor. Later, to reduce costs, two-loop and four-loop controllers were developed, with a slightly better price/performance ratio. For a large or medium-sized system, the price/performance ratio of a DCS is better than that of a system composed of loop controllers. However, some special applications still require the use of loop controllers.
If there are too many loops to complete, such as a single controller collecting data from thousands of points and performing calculations on hundreds of loops, the risks become too concentrated. In such cases, the risks must be dispersed. With the development of computer technology, the computing power, storage capacity, and reliability of computers have continuously improved, allowing a single computer to perform more tasks. These tasks can also be concentrated in one area. Furthermore, redundancy technologies such as controllers and networks have also been developed, allowing control calculations to be more centralized.
From the current perspective of DCS, a single controller can handle the calculations of dozens of loops and the acquisition of hundreds of data points, plus a suitable amount of logical operations. Field use has shown that this performs quite well. This raises the issue of controller upgrades. Sometimes the distance between the controller and the sensing elements is still relatively far, which promotes the development of fieldbuses. Fieldbuses such as CAN, LOONWORKS, and FF, as well as HART protocol receiver boards, are all used in DCS systems.
DCS is divided into three main parts: controllers with I/O boards, communication networks, and human-machine interfaces (HMIs). The I/O boards are directly connected to the production process via terminal blocks to read signals from sensors. There are several different types of I/O boards, each with its corresponding terminal blocks.
Analog input, standard signal board of 4-20 mA and millivolt signal board for reading thermocouples; 4-16 channels;
Analog outputs are typically standard signals of 4-20 mA, and usually have a limited number of channels, ranging from 4 to 8.
Digital input; 16-32 channels:
The switch output and switch input/output are also available in different voltage levels, such as DC 24V and 125V; AC 220V or 115V, etc.; with 8-16 channels.
Pulse input, used to acquire the rate signal; 4-8 channels available.
Quickly interrupt input;
HART protocol input board;
Fieldbus I/O board;
Each I/O board is connected to the I/O bus. For signal safety and integrity, signals must be conditioned before entering the I/O board, such as checking upper and lower limits, temperature compensation, and filtering. These tasks can be done on the terminal block or separately. The board that performs signal conditioning is now sometimes called a signal conditioning board.
The I/O bus connects to the controller. In the 1980s, DCS controllers had limited processing power. To increase the number of I/O points, the controller's tasks were divided, resulting in three types of controllers: a controller performing closed-loop calculations, an analog data acquisition unit, and a logic unit. Each type had its own I/O bus, and the I/O buses varied across different DCS systems. For high speed requirements, a parallel bus was preferred. Serial buses were generally more common, especially RS485 buses, which allowed for a higher number of I/O points for the analog data acquisition unit and the logic unit.
Closed-loop controllers, analog data acquisition units, and logic units can be directly connected to the human-machine interface (HMI) on a communication network. Each controller on the network acts as an independent node, performing different functions. All nodes should have a network interface. Some DCS systems, to save network interfaces, pre-connect all process control equipment—closed-loop controllers, analog data acquisition units, and logic units—onto the control bus, forming a process control station. This increases the number of I/O points the process control station can receive while saving interfaces. Then, it connects to the network via the interface and to the HMI. With the development of computer technology, the computing power of controllers has continuously increased. For example, a PC-based controller can be very powerful, receiving both analog and digital logic inputs. A controller then becomes a node on the network, connecting to the HMI through the network.
The controller is the core component of a DCS (Distributed Control System), essentially functioning as a PC. Some DCS controllers are even PCs in themselves. It primarily consists of chips such as CPU, RAM, E2PROM, and ROM, along with two interfaces: one receiving signals from the I/O bus, and the other sending signals to the network for connection to the human-machine interface. The ROM stores control algorithms that perform various computational functions (in some DCS systems, this is called a function block library). The library stores function blocks, such as PID control algorithms (PID with dead zone, integral-separated PID), arithmetic operations (addition, subtraction, multiplication, division, squaring, square root), function operations (single-order filtering, sine, cosine, XY function generator, lead-lag), more advanced algorithms (Smith predictor, C language interface, matrix addition, matrix multiplication), and logical operations (AND, OR, NOT, NAND, etc.). Typically, function blocks not only combine analog and digital signals but also connect to the user interface. The more function blocks available, the easier it is for the user to write application programs (i.e., configure the system). Configuration involves connecting function blocks according to process requirements to form a control scheme. The control scheme is stored in an E2PROM. Because E2PROMs are erasable and rewritable, and the configuration needs to be changed as the process changes, the configuration is stored in the E2PROM. Different users have different configurations. During configuration, the user selects the required function blocks from the function block library, fills in the parameters, and connects the function blocks. The resulting control scheme is then stored in the E2PROM. At this point, the controller is in configuration mode; after being put into operation, it becomes the operating mode.
The controller is equipped with an operating system, function block configuration software, and communication software.
For safe system operation, the closed-loop controller must operate redundantly, with one in use and one on standby, and hot standby is required. To ensure successful redundancy, the following points should be noted: the hardware and software versions of the two controllers must be identical; check the functionality of the transmit-receive chips; check the functionality of the redundant chips; ensure the settings of the two modules are the same; and check if a manual operating station is included.
Communication networks connect process stations and human-machine interfaces into a single system. There are several different network topologies, such as bus, ring, and star. A bus topology is logically a ring topology. Star topologies are only suitable for small systems. Regardless of whether it's a ring or bus topology, broadcast topology is generally used. Other protocols are less commonly used. Communication network speeds are typically around 10 Mbps to 100 Mbps.
The human-machine interface has four different types of nodes: operator station, engineer workstation, historical trend station, and dynamic data server.
The operator station is equipped with the operating system, monitoring software, and controller driver software. It displays system labels, dynamic flowcharts, and alarm information.
The engineer workstation configures the controller (CAD) and also the operator station (creating dynamic flowcharts). If the monitoring software has strong drawing capabilities, the drawing work can be completed independently by the monitoring software. Another function of the engineer workstation is to read the controller configuration, which is used for controller upgrades and troubleshooting. We call this a reverse engineer workstation.
Historical trend stations are used to store historical data, and are generally equipped with disk arrays (known as RAID technology).
The dynamic data server serves as the interface between the DCS and MIS systems, and also as an isolation device between the DCS and the Web.
The design principles of DCS and PLC differ significantly. PLCs evolved from the original relay control principle. In the 1970s, PLCs only supported on/off logic control and were initially used in the automotive manufacturing industry. They store instructions for performing logic operations, sequential control, timing, counting, and arithmetic operations; and control various machines or production processes through digital input and output operations. The user-written control program expresses the technological requirements of the production process and is pre-stored in the PLC's user program memory. During runtime, the stored program is executed line by line to complete the required operations. The PLC's CPU contains a program counter that indicates the address of each program step. During program execution, this counter automatically increments by 1 for each step. The program executes sequentially from the starting step (step number zero) to the final step (usually the END instruction), and then returns to the starting step in a loop. The time required for the PLC to complete one loop operation is called a scan cycle. Different PLC models have scan cycles ranging from 1 microsecond to tens of microseconds. This cyclic operation of the program counter is something DCS lacks. This is also why PLCs have less redundancy than DCSs. DCS was developed based on operational amplifiers. All functions and relationships between process variables are made into function blocks (sometimes called expansion blocks in DCS systems). In the mid-1970s, DCS only supported analog control. For example, the TDC2000 system could complete the calculation of eight PID loops per second with a single controller. It was initially applied in the chemical industry. The main difference between DCS and PLC lies in the logic processing of digital signals and the computation of analog signals. Even though there has been some overlap between the two, differences remain. After the 1980s, PLCs, in addition to logic operations, also had algorithms for some control loops, but completing complex calculations was still relatively difficult. PLCs use ladder diagram programming, and analog signal computation is not very intuitive during programming, making programming more cumbersome. However, they exhibit a speed advantage in logic processing, processing 1K logic programs in less than 1 millisecond at the microsecond level. They treat all inputs as digital signals, with 16 bits (or sometimes 32 bits) as one analog signal. DCS, on the other hand, treats all inputs as analog signals, with 1 bit representing a digital signal. The computation time for a single logic logic is in the hundreds of microseconds to milliseconds range. A PLC calculates a PID controller in tens of milliseconds, comparable to the computation time of a DCS. Large PLCs use a separate CPU to perform analog calculations and send the results to the PLC controller. Different DCS models require different times to calculate PID controllers, but all are in the tens of milliseconds range. For example, the early TDC2000 system completed the control calculations for eight loops within one second. With the development of chip technology, the time required to calculate an algorithm is decreasing. The time required to calculate an algorithm is related to the arrangement and configuration of the function blocks.
Regarding grounding resistance, the requirements may not be high for PLCs, but for DCS, it must be below a few ohms (usually below 4 ohms). Analog isolation is also very important. Intrinsically safe barriers should be installed in areas with explosion hazards.
For systems with the same number of I/O points, using a PLC is less expensive than using a DCS (saving approximately 40%). PLCs don't have dedicated operator stations; their software and hardware are generic, resulting in significantly lower maintenance costs compared to DCS. A PLC controller can handle thousands of I/O points (up to 8000+), while a DCS controller can only handle a few hundred (no more than 500). If the controlled object mainly involves equipment interlocks and has few loops, a PLC is more suitable. If it primarily involves analog control and many function calculations, a DCS is preferable. DCS is far superior to PLCs in terms of redundancy in controllers, I/O boards, and communication networks, as well as in advanced computing and industry-specific requirements. Because PLCs use general-purpose monitoring software, designing enterprise management information systems is easier.
It is particularly important to point out that dedicated operator stations for DCS are not a given. Their existence is a product of historical reasons. If DCS manufacturers do not open up their operator stations and connect them to the factory's management information system, some DCS systems risk disappearing from the market.
With the emergence of new technologies, negative impacts also arise. The opening of new workstations makes the system more susceptible to intrusion by viruses and hackers. Therefore, during the design phase, passwords should be set on workstations, and the system should be equipped with multiple layers of isolation and firewalls to minimize these negative impacts.
