Abstract: This paper analyzes the EPA communication structure and scheduling strategy, discusses the possible factors affecting the determinism of EPA industrial Ethernet , introduces the processing methods of various transmitted data by the MAC medium access layer in the EPA network, and draws conclusions on the determinism of EPA industrial Ethernet, providing a reference for further design and implementation of EPA industrial Ethernet.
Keywords: Industrial Ethernet; EPA; Determinism
1 Overview
"EPA (Ethernet for Plant Automation) System Architecture and Communication Standard for Industrial Measurement and Control Systems" is an industrial control network communication standard based on information network communication technologies such as Ethernet, wireless LAN, and Bluetooth, suitable for communication between devices and instruments in industrial automation control systems.
Industrial control networks require the transmission of control information to drive events; therefore, high real-time performance is paramount, generally allowing for delays on the order of seconds, and in special cases, even millisecond-level latency. Real-time performance encompasses two aspects: high transmission speed (i.e., high network communication rate) and short response time. Response time is determined by factors such as the ability of field devices to control interruptions, the transmission time of information within the communication system, the time spent waiting for network idle time, and the time required to avoid information collisions on the network. Determinism in time for control networks means that tasks (such as the execution of function blocks) are predictable in time, requiring maximum values to be predictable and less than a certain threshold. The real-time and deterministic aspects of industrial Ethernet are interconnected yet each represents distinct performance requirements.
Because Ethernet uses CSMA/CD media access, traditional Ethernet is fundamentally determinate, which was the main obstacle initially preventing Ethernet from entering the industrial control field. With the advent of full-duplex switched Ethernet, the deterministic drawbacks of traditional Ethernet have disappeared, and network communication speed and efficiency now depend on the switch. To reduce the negative impact of switches on Ethernet communication speed and efficiency, the selection of switching technology becomes a key consideration when architectureling Ethernet. The adoption of advanced switching technologies, VLAN technology, and priority processing technology has become crucial for improving the real-time performance of industrial Ethernet. Meanwhile, major developers have adopted methods to improve Ethernet real-time performance by developing real-time communication protocols at the application layer. This approach maintains the original Ethernet structure, preserves Ethernet's simplicity and cost-effectiveness, and simultaneously meets real-time requirements, becoming the primary means of improving the real-time performance of industrial Ethernet. All currently released industrial Ethernet standards adopt this method. This paper analyzes the deterministic problem of EPA industrial Ethernet through an analysis of its communication structure and communication scheduling strategies, providing a reference for further realizing EPA industrial Ethernet with deterministic control performance.
2. EPA Industrial Ethernet Architecture
From a system hierarchy perspective, the EPA (Effective Device Area) is divided into two layers: the process monitoring layer and the field device layer. Based on the communication coupling relationships and physical installation locations between EPA field devices, the field device layer can be further divided into several sub-segments or control areas. EPA devices within a sub-segment are connected via Ethernet switches and communicate using the EPA communication protocol. Sub-segments are logically isolated from other sub-segments and field device layer sub-segments through EPA bridges, ensuring that communication data within a sub-segment does not flow through other sub-segments, reducing sub-segment load, and improving real-time performance and security.
The EPA communication model is based on the ISO/OSI model and consists of a physical layer, a data link layer, a network layer, a transport layer, and an application layer. A user layer has also been added to form a six-layer communication model. See Figure 1 [1].
The EPA physical layer and data link layer adopt the IEEE 802 series, namely the wired Ethernet standard IEEE 802.3, the wireless LAN standard IEEE 802.11, and the Bluetooth wireless communication standard IEEE 802.15. In order to improve real-time performance, a communication scheduling management entity is added to the MAC layer in the data link layer, which is responsible for managing the parallel operation of real-time EPA communication and non-real-time network communication. The communication scheduling management entity divides the communication messages in the network into periodic messages and non-periodic messages. Periodic messages include data related to process control and have the highest priority. Non-periodic messages include event notifications, application data, etc., which are assigned different priorities according to data categories and then sent in order of priority. All EPA devices communicate periodically. The EPA application layer provides channels and service interfaces for periodic and non-periodic data communication between EPA devices, and is divided into EPA real-time communication specifications and non-real-time communication specifications. The real-time communication specifications include two layers: EPA application layer services and EPA socket mapping interfaces [2].
3. EPA Industrial Ethernet Deterministic Analysis
3.1 Factors Affecting the Determinism of Industrial Ethernet
The determinism of industrial control networks is mainly used to describe the predictable response time and latency of the system, that is, the maximum time of total delay between information transmission and information reception when any two nodes in the network communicate.
The information delay from one EPA field device to another depends on the system implementation. According to the information transmission process, it is divided into [3]: (1) upper layer processing and queuing delay T up-send (2) MAC layer queuing delay T MAC-send (3) information transmission delay T send (4) information propagation delay T trans (5) upper layer processing and queuing delay T up-rec (6) MAC layer queuing delay T MAC-rec . If the total delay time is T delay , then:
Figure 2. Composition of information delay between intelligent devices in the field
Therefore, the latency is determined by both the network and the host. Higher-layer processing and queuing latency depend on the host's system software and processor/memory speed; Ethernet information transmission latency depends on frame length and network bandwidth; propagation latency is determined by transmission distance and signal propagation speed in the medium; receiving station MAC layer queuing latency depends on the time it takes for the receiving station processor to respond to MAC layer interrupts; and sending station MAC layer queuing latency is determined by the MAC layer protocol.
In today's rapidly developing computer networks, the proportion of high-level processing and queuing delays in the overall time latency is decreasing due to the increasing speed of host processors/memory and the continuous improvement of system software. Furthermore, this latency is predictable and will not be discussed further in deterministic analysis. In practical applications, data frames with high real-time requirements are relatively short, and the widespread use of 100M or higher-speed Ethernet makes information transmission delay negligible and predictable. EPA systems are divided into different micro-segments, making propagation delays determined by transmission distance and signal propagation speed negligible and still predictable. The MAC layer queuing delay at the receiving station, determined by the interrupt handling capability of the receiving station processor, is also predictable. Only the sending station queuing delay, determined by the MAC layer protocol, is constrained by the Media Access Protocol (MAC). This is the origin of the term "deterministic network" for the free-competition mechanism of traditional Ethernet. Adopting full-duplex switched Ethernet changes the CSMA/CD mechanism, making information transmission delay solely due to the switch's latency; the speed and efficiency of network communication depend on the switch. To minimize the impact of switches on overall network speed and efficiency, network topology should avoid communication across multiple switches as much as possible, and devices that frequently exchange data should be placed on the same network segment. Therefore, it can be concluded that, given the current state of network technology, the processing speed of the field intelligent processing unit and the MAC layer media access scheduling method of the transmitting station are the main causes of transmission delay.
3.2 EPA Communication Scheduling
The EPA employs three communication methods to meet different information transmission requirements: client/server, report distribution, and publisher/subscriber, respectively satisfying various needs such as high real-time performance, reliability, and saving network resources.
Data transmitted in the EPA network includes basic equipment information, equipment configuration information, process measurement and control information, and alarm information. Based on the characteristics of information transmission over the network, it can be divided into periodic and aperiodic information. Periodic information mainly consists of measurement and control information, which is sent periodically, usually triggered at certain time intervals. It has high real-time requirements and deadlines for execution. The transmission of measurement and control information also has a certain order, and the information flow has a clear directionality. Aperiodic information mainly consists of user operation commands, configuration information, and alarms. These are short, randomly triggered, and typically range from a few bits to tens of bytes, with less information volume. Among aperiodic information, equipment configuration information does not have high real-time requirements but must ensure reliability. Alarm information, however, has high real-time requirements. Generally, real-time data is relatively short; for most real-time data, only the latest data is meaningful, so retransmission is not required. In industrial applications, periodic information is more common, while aperiodic information is less so.
In the EPA system, within each control area (micro-segment), field intelligent devices communicate with each other, and communication between different micro-segments is forwarded by the EPA bridge. Within a micro-segment, all devices communicate periodically. A communication cycle consists of two phases: a periodic message transmission phase (Tp) and an aperiodic message transmission phase (Tn). Periodic messages have the highest priority, and each device uses a producer/consumer communication model to send messages in a pre-configured order. When the node execution time arrives, the field device broadcasts the information to be sent to the entire segment, which can be seen by all device nodes within the segment. Only the corresponding consumer node receives the data. Therefore, periodically transmitted data has time determinism. The scheduling of aperiodic information always occurs between two periodic scheduling sessions, so the duration of aperiodic information transmission cannot exceed this limit. For short data frame information such as alarm information, it can be guaranteed to be completed within the specified time limit. However, for transmission requirements such as system configuration information and uploaded programs, which involve larger data volumes and require guaranteed reliability, the entire transmission process must rely on increased network speed and processing speed of field intelligent processing units. Therefore, this must be carefully considered when designing micro-segments. For randomly generated non-periodic messages within the network segment, the bus arbitrator manages them according to priority through a polling method. Only when a node receives a transmission permission that meets the priority conditions is the non-periodic message allowed to be sent. During the non-periodic message transmission phase, the EPA bus arbitrator can automatically adjust the token priority to ensure that the message is sent smoothly within the specified time range, guaranteeing reliable system operation.
4. Conclusion
Analysis of the EPA communication structure and scheduling strategy reveals that EPA industrial Ethernet offers a certain degree of determinism in response time. It allows for careful design of the network architecture, selection of key network components, and selection of field intelligent devices, tailored to the specific needs and performance requirements of enterprises, thereby meeting industrial control requirements. Improving the real-time performance of the control network is fundamental to ensuring its determinism. The author believes that the key tasks at present are to improve the processing speed and intelligence level of field devices, while simultaneously perfecting the network's clock synchronization.
References
[1] Gao Lu, et al. EPA network architecture. Computer Engineering [J], 2004.9:81-83
[2] Feng Dongqin, et al. Research on real-time communication model of industrial control network based on Ethernet [J]. Journal of Instrumentation, 2003, 8: 715-718
[3] Gu Hongjun, et al. Real-time characteristics analysis and data transmission technology of network control systems. Computer Engineering and Applications, 2001, 6: 38-41
[4] Chen Lei, et al. Research on real-time characteristics of Ethernet in industrial applications [J]. Journal of Zhejiang University (Engineering Science), 2004, 6: 670-675
[5] Tang Hongru. Constructing a real-time communication network platform for field use with Ethernet+TCP/IP. Computer Applications Research, 2001, 3: 45-48
