Industrial control system and method based on heterogeneous computing power direct drive and all-fiber flattening architecture

Abstract

The invention discloses an industrial control system and method based on heterogeneous computing power direct drive and an all-fiber flattening architecture, and belongs to the technical field of industrial automation control. The system comprises a central intelligent computing cluster, a high-speed routing and security gateway, a plurality of intelligent execution terminals and an all-fiber communication network. The central intelligent computing cluster directly generates a control waveform, an intelligent execution terminal is directly driven through an all-fiber network, closed-loop control is achieved, and a traditional PLC and a traditional middle layer are abandoned. The high-speed router and the security gateway are integrated with a hardware fusing module, and security protection and nanosecond clock synchronization are provided. The bypass data mirror image node can record full amount of original data in a lossless mode. The problems of hierarchical redundancy, limited real-time performance, insufficient edge computing power, data value loss and the like of a traditional industrial control system are solved, flat direct drive control with high real-time performance, high reliability and high safety is achieved, and the method is suitable for high-end manufacturing and intelligent industrial scenes.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to an industrial control system and method based on heterogeneous computing power direct drive and an all-fiber flat architecture. Background Technology

[0002] Existing industrial control systems generally adopt a pyramid-shaped hierarchical architecture based on the ISA-95 standard, typically including, from bottom to top, a field input / output (I / O) layer, a programmable logic controller (PLC) control layer, a supervisory control and data acquisition (SCADA) layer, and a manufacturing execution system (MES) layer. While this architecture has played a crucial role in the development of industrial automation, it has several inherent limitations and struggles to fully adapt to the new demands of high-end manufacturing and intelligent transformation. These limitations are specifically reflected in the following aspects: System-level redundancy and limited real-time performance: Data needs to be forwarded through multiple layers of devices and converted between different protocols during transmission, which leads to longer communication paths, increased latency, and difficulty in accurate prediction and control. Therefore, it cannot meet the stringent real-time requirements of millisecond-level or even sub-millisecond-level closed-loop control for highly dynamic processes.

[0003] Insufficient edge computing power and simplistic control strategies: PLCs, which are located in the core control position, are limited by their embedded architecture and computing power, making it difficult to support and run computationally intensive models (such as computational fluid dynamics simulation CFD, deep neural networks DNN, etc.) in real time. This prevents advanced control algorithms from being implemented at the control layer, and in most scenarios, they can only rely on simple logic such as the classic proportional-integral-derivative (PID).

[0004] The physical layer relies on copper cables, which pose challenges to reliability and cost: Traditional distributed control systems (DCS) rely on a large number of copper cables for signal transmission. This not only involves complex cabling projects and high material and maintenance costs, but also makes them susceptible to interference in complex industrial electromagnetic environments, affecting the stability of signal transmission and the overall reliability of the system.

[0005] Data value loss hinders intelligent upgrades: Due to bandwidth and storage limitations, the high-frequency, high-precision raw data collected by the underlying sensors is often significantly downsampled or filtered before transmission (e.g., only the average value is uploaded at second intervals), resulting in the loss of key high-frequency feature information containing equipment status and process dynamics, which hinders the training of artificial intelligence models and process optimization based on complete data. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an industrial control system and method based on heterogeneous computing power direct drive and all-fiber flat architecture.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An industrial control system based on heterogeneous computing power direct drive and an all-fiber flat architecture includes: The central intelligent computing cluster has high-throughput parallel computing capabilities and is equipped with fiber optic communication interfaces for running real-time simulation models, deep learning inference models, and generating control waveforms. The high-speed router and security gateway, built on programmable logic devices, connects the central intelligent computing cluster and several lower-level nodes via fiber optic links, and is used for hardware routing of instructions, protocol parsing, and security boundary control. Several intelligent actuators are distributed in the industrial field, equipped with fiber optic communication interfaces and digital-to-analog / analog-to-digital conversion interfaces, used to directly drive field actuators and collect raw sensor data; A full-fiber communication network connects the central intelligent computing cluster, high-speed routers and security gateways, and all intelligent execution terminals, and uses a proprietary high-speed serial protocol that does not rely on the Ethernet MAC layer. The system architecture does not include a programmable logic controller or intermediate relay layer; instead, the central intelligent computing cluster directly performs closed-loop control of the intelligent execution terminal across all layers.

[0008] Furthermore, the all-fiber communication network adopts a star or tree topology, wherein the high-speed router and security gateway serve as the network's master clock station, broadcasting a global nanosecond-level synchronization clock signal to all the intelligent execution terminals through physical layer encoding.

[0009] Furthermore, the high-speed router and security gateway integrates a hardware circuit breaker module, which stores an unwriteable physical security threshold. When the control command issued by the central intelligent computing cluster exceeds the physical security threshold, the high-speed router and security gateway directly blocks the command at the hardware logic layer and executes a preset security action.

[0010] Furthermore, the system also includes: a bypass data mirroring and storage node, connected to the mirror port of the high-speed router and security gateway, for lossless recording of high-frequency raw waveform data from the intelligent execution terminal in bypass listening mode, and for data storage and persistence independently of the control loop.

[0011] Furthermore, the proprietary high-speed serial protocol uses a fixed-length frame structure for data transmission, and the fixed-length frame structure includes at least a frame header, timestamp, node identifier, data payload, and check field.

[0012] Furthermore, the central intelligent computing cluster is a heterogeneous computing architecture, containing at least a central processing unit and a graphics processing unit, and is equipped with a fiber optic communication acceleration card.

[0013] An industrial control method based on heterogeneous computing power direct drive and an all-fiber flat architecture includes the following steps: S1. Holographic perception: The intelligent execution terminal collects the raw waveform data of the field sensors at a high sampling rate, and transmits it to the high-speed router and security gateway through the all-fiber communication network without compression. S2. High-speed routing: After receiving the original waveform data, the high-speed router and security gateway perform parallel unpacking through hardware logic and directly map the data stream to the memory space of the central intelligent computing cluster. S3. Simulation Decision: The central intelligent computing cluster uses the real-time received data as boundary conditions to run multiphysics simulation or artificial intelligence inference models and calculate the control waveform at the next moment. S4. Direct Execution: The central intelligent computing cluster sends the control waveform through the all-fiber communication network, and after routing through the high-speed router and security gateway, it is directly written into the control register of the intelligent execution terminal to drive the actuator.

[0014] Furthermore, in S3, the simulation decision-making process includes a real-time mapping from simulation to reality, directly converting the calculation results of multiphysics simulation into control quantities for motors, valves, or pumps.

[0015] Compared with the prior art, the beneficial effects of this application are as follows: 1. This invention uses a centralized heterogeneous computing power unit to replace the traditional PLC control layer, realizes direct closed-loop control of field equipment by high-performance computing resources, builds an integrated architecture of "computing power is control", and completely eliminates the delay caused by multi-level forwarding and protocol conversion.

[0016] 2. Replace traditional copper cables and Ethernet with an all-fiber network based on a high-speed serial protocol to achieve nanosecond-level clock synchronization and highly interference-resistant transmission at the physical layer, ensuring that control commands and data streams have deterministic and extremely low end-to-end latency.

[0017] 3. Integrate an FPGA-based hardware fuse module into the high-speed router and security gateway, with a preset, tamper-proof physical security threshold. This module can intercept abnormal control commands in real time without the need for upper-layer computing power intervention, forming a security protection boundary independent of the control loop.

[0018] 4. By using bypass mirroring and storage nodes, high-frequency, high-resolution raw waveform data from the field is captured in full through listening, realizing "black box" persistent storage of control process data, providing a complete data foundation for artificial intelligence model training and process traceability. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a topology diagram of the system of the present invention.

[0021] Figure 2 This is a timing diagram showing the control signals and data flow. Detailed Implementation

[0022] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] One specific embodiment of the present invention relates to an industrial control system based on heterogeneous computing power direct drive and an all-fiber flat architecture. This system completely abandons the traditional pyramidal architecture that includes programmable logic controllers and intermediate relay layers, and instead adopts a highly integrated three-layer flat structure.

[0024] Reference Figure 1 , Figure 2 The core hardware of this system includes a central intelligent computing cluster, high-speed routers and security gateways, and several intelligent execution terminals distributed across the industrial site. All components are interconnected through a full-fiber communication network. The central intelligent computing cluster, serving as the system's control brain, typically employs a heterogeneous computing architecture including a central processing unit and a graphics processing unit, and is equipped with dedicated fiber optic communication acceleration cards (such as those based on PCIe interfaces). It possesses high-throughput parallel computing capabilities, specifically designed for running real-time simulation models of multiphysics fields such as virtual factory digital twins, computational fluid dynamics, and chemical reaction kinetics, as well as executing complex deep learning inference tasks. Based on this, it directly generates the precise control waveforms required to drive the underlying equipment, achieving a fundamental shift of "computing power equals control."

[0025] The high-speed routing and security gateway is a critical hub node built on field-programmable logic devices (FPGAs). It connects to the central intelligent computing cluster upstream and all intelligent execution terminals downstream via fiber optic links. Its core functions include high-speed routing of data streams at the hardware logic level, parsing of proprietary communication protocols, and execution of crucial security boundary controls. The gateway integrates a hardware fuse module that stores non-rewritable physical safety thresholds (such as maximum permissible temperature, pressure, and rotational speed). If control command parameters from the central intelligent computing cluster exceed these preset thresholds, the gateway will directly block the abnormal command at the hardware logic level without relying on upper-layer software decisions, triggering preset safety actions such as outputting a safety status or cutting off power, thus constructing an intrinsically safe defense independent of the control loop. Furthermore, the gateway also acts as the clock master for the all-fiber network. In networks employing star or tree topologies, it broadcasts a global nanosecond-level synchronization clock signal to all intelligent execution terminals using physical layer coding technology, ensuring strict synchronization of all devices at an extremely fine time granularity. Intelligent actuators are embedded nodes deployed next to devices such as motors, pumps, valves, and sensors. They are developed based on system-on-a-chip or small-scale programmable logic devices. Their design is extremely simple and does not have a complex operating system. Their core function is to provide a high-speed fiber optic communication interface and a high-precision digital-to-analog and analog-to-digital conversion interface, so that they can directly acquire raw waveform data from sensors without any compression or downsampling and directly drive actuators to complete high dynamic response.

[0026] The all-fiber communication network connecting all the aforementioned nodes employs a proprietary high-speed serial protocol that does not rely on the Ethernet MAC layer. This protocol typically uses a fixed-length frame structure to ensure deterministic transmission; each data frame includes at least a frame header, a high-precision timestamp, a node identifier, a data payload, and a checksum field. This streamlined design eliminates address resolution, routing lookups, and complex retransmission mechanisms found in the traditional TCP / IP protocol stack, enabling efficient data transmission at both the physical and data link layers. This ensures deterministic and extremely low end-to-end latency throughout the entire path from control command issuance to execution feedback. To further enhance system observability and data value, an optional bypass data mirroring and storage node can be connected to the mirror port of the high-speed router and security gateway. This node operates in bypass listening mode, capable of losslessly recording all high-frequency raw waveform data flowing through the gateway and independently storing and persisting the data long-term, functioning as a "data black box" for industrial control. This provides a complete data foundation for subsequent AI model iteration training, process optimization, and accident tracing.

[0027] The typical implementation process of the industrial control method based on the above system is as follows. First, in the holographic perception stage, intelligent execution terminals deployed on-site acquire raw analog signals from sensors (such as vibration, pressure, and temperature sensors) at extremely high sampling rates (e.g., hundreds of kHz), convert them into digital waveform data, and transmit them directly and transparently to the high-speed router and security gateway via an all-fiber communication network without any online compression processing. Next, in the high-speed routing stage, the gateway utilizes its hardware logic parallel processing capabilities to quickly unpack the massive amounts of received data and, according to predefined mapping relationships, efficiently writes the unpacked data stream into the designated memory or video memory space of the central intelligent computing cluster through methods such as direct memory access, greatly reducing data preparation latency. Subsequently, in the simulation decision-making stage, the central intelligent computing cluster uses the real-time received on-site data as boundary conditions, injecting it into the running multiphysics simulation model or artificial intelligence inference model for real-time calculation and prediction. This process involves a real-time mapping from simulation to reality. For example, the flow rate changes calculated by fluid dynamics simulation are directly and accurately converted into specific control quantities for motor speed or valve opening, and detailed drive waveforms for the next control moment are generated within an extremely short period (e.g., milliseconds). Finally, in the direct-drive execution phase, the central intelligent computing cluster transmits this control waveform through a fiber optic network. After security verification and hardware routing via high-speed routing and a security gateway, it is directly written into the control register of the target intelligent execution terminal. The execution terminal then outputs precise analog signals or drive pulses through its digital-to-analog converter based on the received waveform data, directly controlling actuators such as motors and valves to complete the specified actions, thus forming a millisecond-level high-speed closed loop directly driven by powerful central computing power, from perception and decision-making to execution.

[0028] In summary, this specific implementation method, by constructing a flat direct-drive architecture of "central heterogeneous computing power - fiber optic security gateway - field intelligent terminal" and cooperating with a customized deterministic fiber optic communication protocol, not only completely eliminates the latency bottleneck of traditional multi-layer architecture and realizes high-performance real-time direct control of complex industrial processes, but also fundamentally improves the security and evolvability of the system through hardware-level security circuit breaking and full data bypass recording.

[0029] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An industrial control system based on heterogeneous computing power direct drive and an all-fiber flat architecture, characterized in that, include: The central intelligent computing cluster has high-throughput parallel computing capabilities and is equipped with fiber optic communication interfaces for running real-time simulation models, deep learning inference models, and generating control waveforms. The high-speed router and security gateway, built on programmable logic devices, connects the central intelligent computing cluster and several lower-level nodes via fiber optic links, and is used for hardware routing of instructions, protocol parsing, and security boundary control. Several intelligent actuators are distributed in the industrial field, equipped with fiber optic communication interfaces and digital-to-analog / analog-to-digital conversion interfaces, used to directly drive field actuators and collect raw sensor data; A full-fiber communication network connects the central intelligent computing cluster, high-speed routers and security gateways, and all intelligent execution terminals, and uses a proprietary high-speed serial protocol that does not rely on the Ethernet MAC layer. The system architecture does not include a programmable logic controller or intermediate relay layer; instead, the central intelligent computing cluster directly performs closed-loop control of the intelligent execution terminal across all layers.

2. The system according to claim 1, characterized in that, The all-fiber communication network adopts a star or tree topology, wherein the high-speed router and security gateway serve as the network's master clock station, broadcasting a global nanosecond-level synchronization clock signal to all the intelligent execution terminals through physical layer encoding.

3. The system according to claim 1, characterized in that, The high-speed router and security gateway integrates a hardware circuit breaker module, which stores an unwriteable physical security threshold. When the control command issued by the central intelligent computing cluster exceeds the physical security threshold, the high-speed router and security gateway directly blocks the command at the hardware logic layer and executes a preset security action.

4. The system according to claim 1, characterized in that, Also includes: The bypass data mirroring and storage node is connected to the mirror port of the high-speed router and security gateway. It is used to record high-frequency raw waveform data from the intelligent execution terminal without loss in bypass listening mode, and to store and persist the data independently of the control loop.

5. The system according to claim 1, characterized in that, The proprietary high-speed serial protocol uses a fixed-length frame structure for data transmission. The fixed-length frame structure includes at least a frame header, timestamp, node identifier, data payload, and check field.

6. The system according to claim 1, characterized in that, The central intelligent computing cluster is a heterogeneous computing architecture, which includes at least a central processing unit and a graphics processing unit, and is equipped with a fiber optic communication acceleration card.

7. An industrial control method based on the system according to any one of claims 1 to 6, characterized in that, Includes the following steps: S1. Holographic perception: The intelligent execution terminal collects raw waveform data from the field sensors at a high sampling rate, and transmits it to the high-speed router and security gateway through the all-fiber communication network without compression. S2. High-speed routing: After receiving the original waveform data, the high-speed router and security gateway perform parallel unpacking through hardware logic and directly map the data stream to the memory space of the central intelligent computing cluster. S3. Simulation Decision: The central intelligent computing cluster uses the real-time received data as boundary conditions to run multiphysics simulation or artificial intelligence inference models and calculate the control waveform at the next moment. S4. Direct drive execution: The central intelligent computing cluster sends the control waveform down through the all-fiber communication network, and after routing through the high-speed router and security gateway, it is directly written into the control register of the intelligent execution terminal to drive the actuator.

8. The method according to claim 7, characterized in that, In S3, the simulation decision-making process includes a real-time mapping from simulation to reality, directly converting the calculation results of multiphysics simulation into control quantities for motors, valves, or pumps.

Images