Dynamic Hazard Management System for Manufacturing Sites Based on Industrial Internet of Things

Abstract

This invention discloses a dynamic control system for hazardous sources in manufacturing sites based on the Industrial Internet of Things (IIoT), relating to the field of industrial automation control and safety monitoring technology. The system collects comprehensive monitoring data through a multi-dimensional element sensing unit; a spatiotemporal synchronization processing unit constructs a dynamic spatiotemporal data cube; a nonlinear risk coupling analysis unit uses a coupling model to calculate a comprehensive risk index and identify hidden risks; a hierarchical early warning decision unit generates different levels of instructions based on the index; and a physical-level rigid execution unit achieves closed-loop control through audible and visual alarms, process parameter adjustments, and power cutoff. This invention solves the problems of a single evaluation mode, neglect of coupling effects, and lack of rigid control in response, achieving accurate identification and physical-level safety interception of dynamic hazardous sources, greatly improving the inherent safety level and response efficiency of the manufacturing site.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation control and safety monitoring, specifically relating to a dynamic management and control system for hazardous sources in manufacturing sites based on the Industrial Internet of Things. Background Technology

[0002] In the fields of intelligent manufacturing and the Industrial Internet, achieving deep perception and interconnection of production factors through Industrial Internet of Things (IIoT) technology is a core approach to improving safety on the manufacturing site. This model utilizes various sensing terminals to monitor key dimensions such as people, machines, materials, methods, and environment in real time during the production process, aiming to build a transparent and controllable production environment to effectively prevent various safety accidents and ensure the continuity of production operations.

[0003] Among these, dynamic control systems for hazardous sources in manufacturing sites are a core component of industrial safety. This technology primarily achieves accurate identification and dynamic early warning of potential hazards through real-time acquisition and intelligent analysis of multi-source data from the work site. Its core objective is to establish a closed-loop mechanism covering risk perception, evaluation, and response to address uncertainties in the production environment.

[0004] While existing technologies have made some progress in multi-dimensional monitoring, risk visualization, and visual recognition early warning, they still face core technological bottlenecks. Mainstream assessment models primarily employ the linear superposition of risks across various dimensions, failing to fully consider the non-linear coupling and amplification effects between multiple factors. This results in insufficient risk identification capabilities under complex operating conditions and a high risk of missed reports. Simultaneously, current control logic still heavily relies on manual intervention or graphical alerts, lacking rigid physical control measures for high-risk equipment at the risk response execution level, making it difficult to achieve immediate system-level protection in sudden emergencies. This technological state means that the system's perception accuracy and control strength for dynamic risks cannot match the real-time safety requirements of high-end manufacturing sites, thus becoming a key obstacle to improving the effectiveness of hazard source management in manufacturing sites. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic management and control system for hazardous sources in manufacturing sites based on the Industrial Internet of Things, in order to solve the problems in the existing technology that the assessment mode for hazardous sources in manufacturing sites is too simple, fails to fully consider the nonlinear coupling amplification effect between multiple factors, and lacks physical-level rigid control measures for risk response.

[0006] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0007] A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things (IIoT) includes:

[0008] The multi-dimensional element sensing unit is used to acquire full real-time monitoring data, including personnel behavior trajectories, equipment operating parameters, material status information, environmental physical indicators, and process operation procedures, by utilizing the industrial Internet of Things sensor network deployed on the manufacturing site.

[0009] The spatiotemporal synchronization processing unit is used to perform unified time-scale alignment and spatial coordinate transformation on all real-time monitoring data collected by the multi-dimensional element sensing unit, and to construct a dynamic spatiotemporal data cube oriented towards the manufacturing site.

[0010] The nonlinear risk coupling analysis unit is used to identify risk factors in each single dimension based on a dynamic spatiotemporal data cube, calculate the interaction gain coefficient between different risk factors, and solve the comprehensive risk index of the manufacturing site through a nonlinear coupling model.

[0011] The hierarchical early warning decision unit is used to compare the comprehensive risk index output by the nonlinear risk coupling analysis unit with the preset multi-level risk thresholds, and generate corresponding level control instructions based on the comparison results. The control instructions include reminder-level instructions, intervention-level instructions, and decision-making-level instructions.

[0012] The physical-level rigid actuator is used to respond to the control commands generated by the hierarchical early warning decision unit. By controlling the audible and visual alarm terminals, process parameter adjustment modules, and hardware-level power cut-off devices in the industrial site, it realizes closed-loop control of hazardous sources in the manufacturing site.

[0013] Furthermore, the multi-dimensional element sensing unit is equipped with various types of sensing nodes; personnel behavior trajectories are obtained through interaction between ultra-wideband positioning base stations deployed above the work area and positioning tags worn by workers; equipment operating parameters are collected through a fieldbus gateway connected to the industrial control system, including spindle speed, motor current, bearing temperature, and vibration spectrum; material status information is obtained by reading electronic tags attached to material packaging using an RFID reader; environmental physical indicators are obtained through a multi-functional environmental monitoring terminal integrating dust concentration sensors, combustible gas detectors, ambient temperature sensors, and humidity sensors; and process operation flows are obtained by capturing work scenes with machine vision cameras deployed at key workstations and performing action recognition using an edge computing module.

[0014] Furthermore, the execution process of the spatiotemporal synchronization processing unit is as follows: obtain the local timestamp of each sensing node, and perform millisecond-level clock synchronization on all sensing nodes based on a precise time protocol; establish a three-dimensional spatial coordinate system with the reference zero point of the manufacturing site as the origin, and map the data with positional attributes obtained by each sensing node to the unified three-dimensional spatial coordinate system; according to the preset time step, fill the synchronized data into the corresponding spatial voxel grid in chronological order to form a dynamic spatiotemporal data cube representing the real-time status of the manufacturing site.

[0015] Furthermore, the nonlinear risk coupling analysis unit identifies risk factors and calculates the comprehensive risk index in the following manner:

[0016] Step 1: Extract feature vectors from the dynamic spatiotemporal data cube and compare them with the preset safety benchmark library. When the feature vectors deviate from the range of the safety benchmark library by more than the preset tolerance value, it is determined that there is a single-dimensional risk factor.

[0017] Step 2: Construct a risk factor interaction matrix, where each element represents the degree of mutual influence when two specific risk factors coexist.

[0018] Step 3: Based on the historical accident case database, the interaction gain coefficient under different combinations of risk factors is extracted through multivariate nonlinear regression analysis. The interaction gain coefficient is used to characterize the portion of the destructive energy or degree of danger generated when multiple risk factors are spatially adjacent and temporally overlapping, which exceeds the simple linear superposition of the factors.

[0019] Step 4: Take the risk value of each single dimension as the independent variable, combine it with the corresponding interaction gain coefficient, and input it into the preset exponential coupling function to output the comprehensive risk index.

[0020] Furthermore, the multi-level risk thresholds preset in the hierarchical early warning decision unit include a first threshold, a second threshold, and a third threshold, with the third threshold being greater than the second threshold and the second threshold being greater than the first threshold. When the comprehensive risk index is greater than or equal to the first threshold and less than the second threshold, the risk level is determined to be low risk, and an alert-level instruction is generated. When the comprehensive risk index is greater than or equal to the second threshold and less than the third threshold, the risk level is determined to be medium risk, and an intervention-level instruction is generated. When the comprehensive risk index is greater than or equal to the third threshold, the risk level is determined to be high risk, and a decision-making level instruction is generated.

[0021] Furthermore, the physical-level rigid execution unit performs the following actions in response to different levels of control commands:

[0022] In response to the alert-level command, the sound and light alarms deployed at the work site are activated to issue a warning signal, and the location and type of risk are displayed on the large screen at the site to remind the workers to pay attention;

[0023] In response to intervention-level commands, parameter modification commands are sent to the controllers of relevant devices through the industrial IoT gateway to forcibly reduce the operating speed of the devices or reduce the workload, so that the manufacturing site can be restored to a controlled state.

[0024] In response to a shutdown command, it directly controls the trip coil of the electromagnetic contactor or circuit breaker connected in series in the equipment's power supply circuit to perform a physical-level power supply cut-off, forcibly stopping all high-risk operations.

[0025] Furthermore, the system also includes a digital twin monitoring unit, which is used to receive dynamic spatiotemporal data cubes and comprehensive risk indices, synchronously recreate the operating status of the manufacturing site in virtual space, and mark the distribution and coupling evolution trend of risk factors in real time in the form of heat maps; the digital twin monitoring unit is also used to perform virtual simulation verification of the action results of physical-level rigid execution units in order to evaluate the effectiveness of control measures.

[0026] Furthermore, the nonlinear risk coupling analysis unit is also used to: monitor the rate of change of the comprehensive risk index; when the comprehensive risk index does not reach the third threshold but its rate of change exceeds the preset mutation threshold, the system determines that there is a risk of instantaneous failure and automatically raises the warning level by 1 level.

[0027] Furthermore, the physical-level rigid execution unit has hardware-level self-testing and feedback functions; after receiving the control command and executing the action, the physical-level rigid execution unit monitors the current change of the actuator in real time through the current transformer to confirm whether the physical action has been truly completed; if the expected current change is not detected within a preset 500 milliseconds, the system will trigger the backup emergency braking logic through the backup communication link.

[0028] Furthermore, the system adopts an architecture that combines edge computing and a cloud platform. The multi-dimensional element perception unit, the spatiotemporal synchronization processing unit, and the nonlinear risk coupling analysis unit operate in the edge computing gateway deployed on the manufacturing site to ensure the real-time nature of risk identification and data processing latency of less than 20 milliseconds. The control instructions generated by the hierarchical early warning decision unit are simultaneously uploaded to the cloud platform for archiving and backup. The cloud platform performs periodic optimization and iteration on the interaction gain coefficient in the nonlinear coupling model based on the long-term accumulated risk control data.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] (1) By constructing a nonlinear risk coupling analysis unit, this invention completely changes the traditional safety management system's assessment mode of simply superimposing risk factors linearly. This solution deeply explores the interactive influence of multiple dimensions of risk factors such as people, machines, materials, methods, and environment within a specific time and space range. It introduces interactive gain coefficients and nonlinear coupling models, which can accurately identify those hidden risks that seem normal in a single dimension but may induce serious accidents under the coupling of multiple factors. This greatly improves the coverage and accuracy of hazard source identification under complex manufacturing conditions and effectively reduces the risk underreporting rate caused by nonlinear coupling amplification.

[0031] (2) The present invention designs a multi-level control mechanism that includes a physical-level rigid execution unit, realizing a technological leap from information reminders to physical interception; for high-risk states, the system no longer relies solely on warnings or manual confirmations from the software interface, but achieves rigid protection within seconds through a hardware-level power cut-off device; this physical-level power-off tripping mechanism ensures that in sudden extreme emergencies, the system can bypass complex software logic and potential human intervention delays, providing the last deterministic safety barrier for the manufacturing site.

[0032] (3) The present invention constructs a dynamic spatiotemporal data cube through a spatiotemporal synchronization processing unit, providing millisecond-level time alignment and high-precision spatial positioning support for the manufacturing site; the precise spatiotemporal correlation technology of the present invention makes the location of risk factors no longer a vague regional description, but can be accurate to specific spatial coordinates and precise occurrence time; combined with the edge computing architecture, it ensures extremely low latency throughout the entire link from perception to execution, enabling the control system to have real-time tracking and rapid response capabilities for dynamic hazard sources, greatly improving the inherent safety level of the high-end intelligent manufacturing site.

[0033] (4) By introducing a digital twin monitoring unit and a closed-loop feedback mechanism, this invention achieves visualization and verifiability of control effectiveness. The system can not only simulate risk evolution in real time in virtual space, but also ensure the implementation of physical-level rigid control commands by real-time monitoring of the current of the actuator. This dual verification mechanism and the continuous iteration capability of the cloud model enable the system to build a dynamic safety governance ecosystem that is self-aware, automatically decision-making, enforced, closed-loop feedback and continuously optimized, which significantly enhances the resilience of manufacturing enterprises in dealing with unknown and complex safety challenges.

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, embodiments of the present invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the overall technical solution architecture of the dynamic control system for hazardous sources in the manufacturing site based on the Industrial Internet of Things proposed in this invention.

[0037] Figure 2 This is a schematic diagram of the core principle framework of the nonlinear risk coupling analysis unit in this invention;

[0038] Figure 3 This is a logical flowchart of the spatiotemporal synchronization processing and dynamic spatiotemporal data cube construction in this invention;

[0039] Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between hierarchical early warning decision-making and physical-level rigid execution in this invention;

[0040] Figure 5 This is a logical flowchart of the digital twin monitoring and closed-loop feedback verification in this invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0042] Please refer to the attached document. Figure 1This embodiment provides a dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things (IIoT). This system is built upon the deep integration of the physical and digital spaces of a modern intelligent manufacturing plant. The system achieves deep perception of all production elements through an IIoT sensor network deployed at the bottom layer of the manufacturing site. Multi-dimensional element sensing units, acting as the system's sensing antennae, are equipped with various types of sensing nodes, aiming to capture subtle dynamics of the manufacturing site comprehensively. The acquisition of personnel behavior trajectories is accomplished collaboratively by ultra-wideband positioning base stations deployed above the work area and ultra-wideband positioning tags worn by the workers. The ultra-wideband positioning base stations employ nanosecond-level narrow pulse technology, calculating the three-dimensional spatial coordinates of personnel in the manufacturing site by measuring the time difference between the signals emitted by the positioning tags reaching different base stations. The positioning accuracy is maintained within 10cm, and the sampling frequency is set to 20 times per second, ensuring real-time tracking of personnel movements, bending, crossing restricted areas, and other behaviors. The acquisition of equipment operating parameters relies on a fieldbus gateway connected to the industrial control system. This gateway interacts with the programmable logic controller (PLC) on the manufacturing floor via shielded twisted-pair cables, strictly adhering to industrial communication protocols, and retrieving core operating indicators in real time, including spindle speed, motor current, bearing temperature, and vibration spectrum. The spindle speed data update cycle is 10 milliseconds, the motor current sampling accuracy reaches 0.01 amperes, and the bearing temperature is converted into a standard current signal of 4 to 20 milliamperes by a thermistor embedded in the bearing housing, which is then read by the fieldbus gateway. Vibration spectrum data is acquired by a piezoelectric accelerometer, and the raw waveform is processed by a fast Fourier transform within the fieldbus gateway to extract characteristic frequency points. Material status information is obtained by reading RFID tags attached to material packaging or turnover boxes using an RFID reader. The tags store the material's chemical stability, physical weight, storage time limit, and current process status code. Environmental physical indicators are monitored by a multi-functional environmental monitoring terminal integrating a dust concentration sensor, a combustible gas detector, an ambient temperature sensor, and a humidity sensor. Among them, the dust concentration sensor uses the laser scattering principle to output concentration data at the microgram per cubic meter level in real time; the combustible gas detector monitors the percentage of the lower explosive limit of volatile gases on site through a catalytic combustion sensor; and the ambient temperature and humidity sensors provide basic background parameters for environmental coupling risks. The digital identification of the process operation flow is completed by machine vision cameras deployed directly above key workstations. The machine vision cameras capture high-resolution images of the operation and push the video stream to the edge computing module via high-speed Ethernet. The edge computing module uses deep learning algorithms to perform feature matching on the operator's body movements and the interaction process of the operating tools, thereby determining whether the current actual process conforms to the preset process specifications.

[0043] Combined with appendix Figure 3The spatiotemporal synchronization processing unit receives all raw data streams from the multi-dimensional element sensing unit. Due to crystal oscillator frequency drift and network transmission jitter between different sensing nodes, the spatiotemporal synchronization processing unit first obtains the local timestamp of each sensing node and performs millisecond-level clock synchronization on all sensing nodes based on a precise time protocol to ensure that the entire system has a unified absolute time reference with a clock synchronization accuracy better than 1 millisecond. In the spatial dimension, the spatiotemporal synchronization processing unit establishes a three-dimensional Cartesian spatial coordinate system with the physical ground reference zero point of the manufacturing site as the origin. For the personnel positions obtained by ultra-wideband positioning tags, the material positions determined by RFID readers, and the positions of fixedly deployed sensors, the spatiotemporal synchronization processing unit uses a spatial coordinate transformation matrix to map the data with positional attributes obtained by each sensing node to a unified three-dimensional spatial coordinate system. Subsequently, according to a preset time step of 50 milliseconds, the spatiotemporal synchronization processing unit fills the synchronized multi-source heterogeneous data into the corresponding spatial voxel grids in chronological order. Each spatial voxel grid represents a 0.5m × 0.5m × 0.5m cubic area of ​​the manufacturing site. This data organization form a dynamic spatiotemporal data cube that represents the real-time status of the manufacturing site. This enables any coordinate point in the system to have a strongly correlated structured attribute for personnel, equipment, materials, environment, and process status at any time, providing a standardized data foundation for subsequent risk coupling analysis.

[0044] Please refer to the attached document. Figure 2 The nonlinear risk coupling analysis unit is the core intelligent hub of this system, and its operation logic is divided into four key stages.

[0045] In step 1, the nonlinear risk coupling analysis unit extracts feature vectors from the dynamic spatiotemporal data cube. The feature vector is a sequence of floating-point numbers with 32 dimensions, encompassing real-time observations of various sensor parameters. The nonlinear risk coupling analysis unit compares this feature vector with a safety benchmark library stored in local high-speed memory in real time. The safety benchmark library records the value ranges and fluctuation envelopes of each parameter under normal operating conditions. When the value of a certain dimension in the feature vector deviates from the preset range of the safety benchmark library by more than a preset 5% tolerance value, the system determines that a single-dimensional risk factor exists within that spatial voxel. For example, when the bearing temperature exceeds the preset upper limit of 65℃, or when the personnel trajectory coordinates enter the prohibited area of ​​the robot's operating radius, the corresponding single-dimensional risk factor is activated.

[0046] In step 2, the nonlinear risk coupling analysis unit constructs a risk factor interaction matrix. This matrix is ​​a symmetric matrix, where each element represents the degree of mutual influence when two specific risk factors coexist in the same or adjacent voxels. The rows and columns of the matrix correspond to unsafe acts by personnel, abnormal equipment conditions, hazardous properties of materials, adverse environmental conditions, and process violations, respectively.

[0047] In step 3, the system, based on a historical accident case database, extracts the interaction gain coefficients under different combinations of risk factors through multivariate nonlinear regression analysis. The interaction gain coefficient characterizes the portion of destructive energy or danger generated when multiple risk factors are spatially adjacent and temporally overlapping, exceeding the simple linear superposition of the factors. This system introduces a nonlinear coupling model to quantitatively describe this amplification effect, and its calculation logic follows the formula below:

[0048]

[0049] In the above formula, the interaction gain coefficient The calculation logic is as follows: the basic weight coefficients based on the interaction of risk factors. The term is determined by multiplying by the comprehensive logarithm term. Specifically, the comprehensive logarithm term is the natural logarithm of the target summation value, which is the sum of a constant 1 and a spatial coupling factor; the spatial coupling factor is obtained by multiplying the spatial overlap term and the spatial distance attenuation term. The spatial overlap term is characterized as follows: , , These represent the sets of influence ranges of the two risk factors in a three-dimensional coordinate system, with the numerator being... The denominator represents the intersection of the influence ranges of two risk factors. This represents the total spatial volume involved in this workstation. The spatial distance attenuation term is represented by the exponential part of the natural logarithm base e. This is used to describe how risk factors decay with increasing spatial distance, where... The Euclidean distance between the centroids of two risk factors. The characteristic length represents spatial correlation. Using this formula, the nonlinear risk coupling analysis unit can detect hidden high-risk states where a single parameter has not reached the alarm threshold, but which are caused by energy accumulation or spatiotemporal overlap due to the interplay of multiple factors.

[0050] In step 4, the nonlinear risk coupling analysis unit uses the risk value of each individual dimension as an independent variable, combines it with the corresponding interaction gain coefficient, and inputs it into a preset exponential coupling function to output the final comprehensive risk index. The comprehensive risk index is a continuous scalar value between 0 and 100, used to characterize the overall threat level of the manufacturing site. Its calculation principle is as follows:

[0051]

[0052] In this description of the principle, the comprehensive risk index R consists of two parts, wherein... This represents the total number of single-dimensional risk factors. The first part is the risk factors for each single dimension. With corresponding weight coefficients The first part is a linear weighted summation, reflecting the basic accumulation of risk; the second part is the interaction gain coefficient calculated based on the aforementioned steps. risk factors that appear in pairs and The nonlinear weighting reflects the coupled amplification effect of risks. This algorithm structure ensures that when both unauthorized operations and equipment overheating occur simultaneously on-site, the comprehensive risk index provided by the system will be much higher than the sum of the values ​​when either occurs alone.

[0053] Combined with appendix Figure 4 The hierarchical early warning decision unit receives the comprehensive risk index output by the nonlinear risk coupling analysis unit in real time. The hierarchical early warning decision unit has three preset risk thresholds: Threshold 1, Threshold 2, and Threshold 3. Based on the safety level requirements of the manufacturing site, Threshold 1 is set to 35, Threshold 2 to 65, and Threshold 3 to 85. The hierarchical early warning decision unit uses hard real-time comparison logic. When the comprehensive risk index is greater than or equal to 35 and less than 65, the system determines that the current manufacturing site is in a low-risk state, with potential danger in its nascent stage, and the hierarchical early warning decision unit generates a reminder-level instruction. When the comprehensive risk index is greater than or equal to 65 and less than 85, the risk level is determined to rise to medium risk, meaning that risk factors have begun to produce significant coupling interactions and are highly likely to evolve into an accident; at this time, an intervention-level instruction is generated. When the comprehensive risk index is greater than or equal to 85, the risk level is determined to be high risk, indicating that the manufacturing site is on the verge of extreme danger and operations must be stopped immediately; at this time, a stop-loss-level instruction is generated. While generating control instructions, the hierarchical early warning decision unit encrypts and encapsulates the feature vector, spatiotemporal coordinates, and coupling calculation process of the triggering instructions to form a safety event message.

[0054] The physical-level rigid actuator is the final link in this system's closed-loop control, directly impacting the physical equipment on the production floor. The physical-level rigid actuator executes actions with varying degrees of force depending on the level of control commands. In response to alert-level commands, the physical-level rigid actuator drives audible and visual alarms deployed in key passageways and control panels on the work floor to emit high-decibel warning audio signals at a frequency of twice per second, accompanied by flashing red lights. Simultaneously, multiple 4K resolution large screens on-site display the precise coordinates of the risk's location, the type of risk factor involved, and suggested avoidance actions in real time, prompting on-site personnel to conduct self-checks. In response to intervention-level commands, the physical-level rigid actuator sends parameter modification commands to the controllers of relevant equipment via an industrial IoT gateway. This command has a high priority and can override routine commands from on-site operators. By modifying the output frequency of the frequency converter or the setpoint of the proportional-integral-derivative controller, the system forcibly reduces the operating speed of rotating equipment, for example, reducing the spindle speed to 20% of its rated speed, or reducing the working pressure of the actuator by adjusting the hydraulic servo valve, thereby restoring the manufacturing floor to a controlled and safe state by reducing the system's internal energy. In response to a shutdown command, the physical-level rigid actuator employs the most stringent rigid protection measures. Its output channel is directly connected to the trip coil of the electromagnetic contactor or circuit breaker connected in series in the main power supply circuit of the equipment. When a shutdown command is issued, the physical-level rigid actuator outputs a 24-volt current pulse lasting 500 milliseconds, driving the trip coil to actuate and causing the contacts of the electromagnetic contactor to rapidly separate within 30 milliseconds, thereby cutting off the power supply to the equipment at the physical hardware level. This mechanism bypasses complex application-layer software logic, achieving a mandatory and irreversible physical stop for high-risk operational activities.

[0055] The physical-level rigid actuator also possesses hardware-level self-testing and feedback functions to ensure the accurate implementation of control commands. After receiving and executing a control command, the physical-level rigid actuator monitors the current changes in the power circuit in real time through a current transformer installed in the distribution cabinet. For example, after issuing a power cut-off command, the system expects to detect a sudden drop in current from the normal operating value to 0 amperes. The microprocessor inside the physical-level rigid actuator analyzes the current sampling data in real time to confirm whether the physical action has been truly completed. If the expected current drop edge is not detected within the preset 500-millisecond response time, the system will determine that the main actuator has malfunctioned, such as contactor contact sticking. In this case, the system will activate the backup emergency braking logic located upstream of the power distribution via the backup industrial wireless communication link to ensure the redundancy and reliability of safety protection.

[0056] Please refer to the attached document. Figure 5This system also integrates a digital twin monitoring unit. This unit receives, in real-time, dynamic spatiotemporal data cubes generated by the spatiotemporal synchronization processing unit and the comprehensive risk index output by the nonlinear risk coupling analysis unit via a high-speed data bus. In the virtual space, the digital twin monitoring unit synchronously recreates the operational status of the manufacturing site based on pre-constructed high-precision 3D modeling data. Every piece of equipment, every moving personnel, and every material has a corresponding 3D digital twin in the virtual space. The digital twin monitoring unit uses a dynamic heatmap to mark the distribution of risk factors in real-time at the ground and spatial levels of the virtual manufacturing site. The color depth of the heatmap is determined by the comprehensive risk index, with red areas representing high-coupling risk zones and blue areas representing safe zones. This visualization method allows managers to intuitively observe the spatiotemporal evolution trend of risk factors and identify critical paths of risk flow. Furthermore, the digital twin monitoring unit is also used for virtual simulation verification of the action results of physical-level rigid actuators. When the system takes measures to reduce speed or cut off power, the digital twin monitoring unit uses its built-in physics engine to simulate the gliding trajectory of the equipment under inertia and the decline curve of environmental parameters, and compares it with the actual data fed back by the sensors, thereby assessing the effectiveness and timeliness of the control measures.

[0057] The nonlinear risk coupling analysis unit also possesses the ability to perceive the speed of risk evolution, i.e., to monitor the rate of change of the comprehensive risk index. In actual manufacturing scenarios, certain sudden hazards may cause the risk index to surge dramatically in a very short period of time. When the current value of the comprehensive risk index has not reached the third threshold, but its first derivative, i.e., the rate of change, exceeds the preset abrupt change threshold of 15 index units per second, the system determines that there is a risk of instantaneous failure at the manufacturing site, indicating that uncontrollable catastrophic consequences are about to occur. At this time, the graded early warning decision unit will activate the escalation response logic, automatically raising the current early warning level by one level. For example, the original medium-risk intervention level instruction can be directly upgraded to a high-risk decisive level instruction, gaining an extremely valuable millisecond-level window for accident prevention.

[0058] This system adopts a distributed architecture that integrates edge computing and a cloud platform. Modules with extremely high real-time requirements, such as the multi-dimensional element perception unit, the spatiotemporal synchronization processing unit, and the nonlinear risk coupling analysis unit, all run on edge computing gateways deployed at the edge of the manufacturing site. The edge computing gateways utilize high-performance embedded processors and hardware acceleration units to ensure that the end-to-end latency from perception data input to risk analysis output is less than 20 milliseconds, meeting the real-time requirements of dynamic control at the manufacturing site. Simultaneously, control commands generated by the hierarchical early warning decision-making unit, real-time monitoring data streams, and execution feedback results are all uploaded to a cloud platform located in the enterprise's central computer room or an internet data center for full archiving and backup via an encrypted secure transmission tunnel. The cloud platform leverages its powerful computing and storage advantages, based on massive amounts of historical risk control data accumulated over a long period, and uses offline machine learning algorithms to periodically optimize and iterate the interaction gain coefficients and weight coefficients of each risk factor in the nonlinear coupling model. Every 24 hours, the cloud platform distributes the generated optimized model parameter package to the edge computing gateway, enabling continuous evolution of the intelligence level of the control system.

[0059] In specific precision electronics manufacturing environments, this system provides more targeted implementation solutions for cleanliness, electrostatic interference, and hazard control in confined spaces.

[0060] Please refer to the attached document. Figure 1 In this embodiment, the multi-dimensional element sensing unit further incorporates a high-precision electrostatic potentiometer and a micro / nano-scale particle size counter. Because the electronic manufacturing environment is extremely sensitive, the electrostatic potentiometer monitors the potential fluctuations on the workbench surface at a sampling rate of 100 times per second, preventing the risk of fire or product damage due to electrostatic discharge. In addition to using ultra-wideband positioning, personnel behavior trajectory acquisition also incorporates a ground pressure sensor array. This pressure sensor array is embedded beneath the cleanroom floor and, by sensing the pressure distribution of personnel footprints, can help determine whether workers are engaging in unauthorized rapid running or handling excessively heavy materials.

[0061] Combined with appendix Figure 3 At this point, the spatiotemporal synchronization processing unit refines the precision of the three-dimensional spatial coordinate system to the 5cm level. Each voxel in the dynamic spatiotemporal data cube not only stores physical coordinates and timestamps, but also adds attributes such as electrostatic potential level and airflow organization velocity dimension. The spatiotemporal synchronization processing unit preprocesses the multi-source sensor data by introducing a Kalman filter algorithm to remove data noise caused by high-frequency vibrations of the cleanroom air conditioning system, ensuring that the manufacturing site status represented by the data cube has extremely high confidence.

[0062] Please refer to the attached document. Figure 2The nonlinear risk coupling analysis unit pays particular attention to the nonlinear relationship between environmental humidity and static electricity accumulation when dealing with risks in precision electronics manufacturing. In step 3, when calculating the interaction gain coefficient, the nonlinear coupling model introduces a humidity correction term. When the environmental humidity is below 30%, the interaction gain coefficient between the static electricity factor and the personnel movement speed factor will increase exponentially. This means that in a dry environment, even slight personnel movement can induce extremely high potential jump risks. The system uses the following formula to finely characterize the risk interaction:

[0063]

[0064] In this formula, The basic proportional coefficient, This represents the percentage of ambient humidity collected in real time. As the humidity H decreases, the fractional term increases rapidly, significantly increasing the weight of electrostatic-related risk factors. This represents the rate of change of the monitored electrostatic potential. denoted as the voltage sensitivity coefficient. Through this nonlinear mapping, the system can sensitively detect minute but potentially fatal environmental coupling risks in precision manufacturing environments.

[0065] Combined with appendix Figure 4 The tiered early warning decision-making unit has adjusted the meaning of control commands for precision manufacturing scenarios. In response to intervention-level commands, the physical-level rigid execution unit, in addition to adjusting equipment parameters, automatically increases the power of the ion fans in the cleanroom to neutralize accumulated static electricity by releasing reverse ions. Simultaneously, the system forcibly increases the humidity of the local space to a safe threshold range by adjusting the valve opening of the air circulation system. In response to shutdown-level commands, the physical-level rigid execution unit not only cuts off the power supply to the main production line but also simultaneously activates the pre-pressurization circuit of the emergency inert gas fire suppression system and opens audible and visual guidance signs at key exit passages to guide personnel to evacuate along predetermined routes.

[0066] To prevent tool breakage or workpiece damage caused by sudden power loss from a precision machining spindle operating at high speed when executing a power-off command, the physical-level rigid actuator integrates an energy-consumption braking module. At the instant the power-off command is triggered, the physical-level rigid actuator first controls the energy-consumption braking contactor to engage, rapidly converting the motor's kinetic energy into heat energy through a high-power braking resistor, achieving a smooth and rapid shutdown with a stop time controlled within 0.8 seconds. Then, it completely disconnects the main circuit contactor. This two-stage rigid actuator strategy ensures personnel safety while minimizing the risk of mechanical damage to production equipment.

[0067] Please refer to the attached document. Figure 5In this embodiment, the digital twin monitoring unit adds the ability to simulate microscopic environmental fields. Using computational fluid dynamics algorithms, the digital twin monitoring unit simulates the drift path and accumulation area of ​​dust particles in a cleanroom in real time within a virtual space. When the system detects an increase in the risk index at a certain workstation, the digital twin monitoring unit can display the coverage area of ​​the risk spread on a heat map 30 seconds in advance using a predictive algorithm, providing proactive decision support for on-site safety supervisors.

[0068] In this embodiment, a higher-frequency data synchronization strategy is employed between the edge computing gateway and the cloud platform. Due to the extremely rapid evolution of risks in precision manufacturing, the edge computing gateway pushes 10 sets of key feature vector snapshots to the cloud every second. The cloud platform utilizes a risk assessment model based on deep reinforcement learning to dynamically fine-tune the risk thresholds for each workstation. For example, when it detects that the antistatic coating thickness of the raw materials is too thin for the day, the cloud platform automatically lowers the second and third thresholds for all relevant workstations, putting the control system in a more sensitive defensive state. This mechanism of dynamically adjusting thresholds based on material properties demonstrates the system's adaptive control capabilities in complex and ever-changing manufacturing environments.

[0069] In large-scale heavy equipment manufacturing sites, such as shipbuilding or large steel structure processing, the scale and spatial span of hazard sources increase significantly.

[0070] Please refer to the attached document. Figure 1 In this embodiment, the multi-dimensional element perception unit is extended for large-scale spaces. The acquisition of personnel behavior trajectories employs a fusion solution of long-range lidar and ultra-wideband positioning. The lidar is deployed high on the factory's pillars, achieving continuous tracking of multiple work targets over a large area through point cloud scanning at 300,000 points per second. Equipment operating parameter collection covers comprehensive information from the large overhead crane, CNC cutting machine, and welding robot. The overhead crane's operating data is acquired through a wireless industrial communication gateway, including hook height, lifting weight, trolley travel position, and wind speed load. The welding robot's monitoring data includes shielding gas flow rate, arc voltage, and weld seam tracking deviation.

[0071] Combined with appendix Figure 3 The dynamic spatiotemporal data cube constructed by the spatiotemporal synchronization processing unit possesses multi-scale characteristics. For the core processing area, the voxel grid size is set to 0.2 meters; for the large storage and transshipment area, the voxel grid size is set to 2 meters. The system utilizes multi-level grid indexing technology to ensure efficient data retrieval and risk correlation performance even in large-scale spaces. Regarding time synchronization, the system introduces the BeiDou satellite time signal as a reference time source, ensuring that the time error between hundreds of sensing nodes distributed across tens of thousands of square meters of factory space is strictly controlled within 500 microseconds.

[0072] Please refer to the attached document. Figure 2 The nonlinear risk coupling analysis unit enhances the model to address common risks in heavy manufacturing, such as lifting operations and overlapping construction activities. When calculating the comprehensive risk index, the system focuses on the triple combined effects of the suspended height of the heavy object, the density of personnel on the ground, and the on-site wind speed. By introducing the gravitational potential energy function and momentum estimation model, the nonlinear risk coupling analysis unit can predict the impact range in extreme situations such as wire rope breakage or hook slippage. A kinetic energy impact term has been added to its coupling calculation logic to ensure precise control of kinetic energy release hazards.

[0073] Combined with appendix Figure 4 The control commands generated by the tiered early warning decision unit have stronger linkage capabilities. In response to intervention-level commands, the system not only limits the vehicle's operating speed but also coordinates with the mobile work platform below to automatically move to a safe area. In response to stop-start commands, the physical-level rigid actuator directly acts on the vehicle's brake solenoid valve and hydraulic lock to forcibly lock the mechanical energy. For ongoing welding operations, the stop-start command immediately cuts off the welding power supply and shuts off the protective gas solenoid valve to prevent the fire from spreading.

[0074] Physically rigid actuators employ a distributed deployment architecture in large-scale field applications. Each core equipment cluster is paired with a rigid actuator terminal. These terminals maintain a heartbeat connection via a ring-shaped industrial Ethernet network. Once a halt-level command is triggered, nearby actuators can coordinate their actions, such as interlocking and shutting down adjacent production lines to prevent cascading failures. The physically rigid actuators utilize built-in backup power modules to ensure continued safety monitoring and emergency operation capabilities even during plant power outages.

[0075] Please refer to the attached document. Figure 5 The digital twin monitoring unit incorporates augmented reality (AR) technology, overlaying real-time calculated risk heat maps onto AR glasses worn by on-site inspectors. As inspectors move around, they can visually see hidden risk areas, such as invisible laser cutting paths or high-pressure gas leak risk zones. The digital twin monitoring unit also uses 3D sound field simulation to guide management personnel to focus on the source of abnormal sounds, improving the efficiency of risk perception in large and complex environments.

[0076] In heavy manufacturing scenarios, the cloud platform undertakes a larger-scale model training task. Due to the long manufacturing cycle and extremely complex processes of heavy equipment, the cloud platform utilizes long short-term memory neural networks to deeply analyze equipment failures and safety incidents accumulated over the past three years. By identifying the long-range correlation between equipment wear patterns and personnel fatigue cycles, the system can predict potential high-risk work windows 3 to 5 days in advance. The cloud platform distributes predictive control strategy packages to guide the nonlinear risk coupling analysis units on the edge to strengthen the monitoring intensity of specific factors within specific time periods.

[0077] In summary, this invention constructs a dynamic control system covering the entire chain from perception and analysis to decision execution through deep collaboration of multi-dimensional element perception units, spatiotemporal synchronization processing units, nonlinear risk coupling analysis units, hierarchical early warning decision-making units, and physical-level rigid execution units. The system accurately identifies the interactive amplification effect of multi-dimensional risk factors through a nonlinear coupling model, solving the problem of single and one-sided risk assessment in traditional solutions. Through millisecond-level spatiotemporal synchronization and data cube technology, it achieves transparent and precise state characterization of the manufacturing site. Through a hardware-level power cutoff mechanism, it ensures rigid physical protection by bypassing software risks in extreme and critical moments. This dynamic control solution based on the Industrial Internet of Things (IIoT) significantly improves the inherent safety level of the manufacturing site, providing solid technical support for safe production in high-end intelligent manufacturing.

[0078] In this embodiment, the functionality of each unit relies on an industrial-grade hardware platform. The edge computing gateway employs a 16-core processor with a clock speed of 2.4 GHz and at least 64 gigabytes of memory, ensuring extremely fast response for nonlinear regression algorithms and spatial voxel operations. The electromagnetic tripping mechanism of the physical-grade rigid execution unit uses high-performance magnetic materials to ensure reliability under 3 million cycles of operation. System communication uses fully shielded Cat 6e network cables and industrial-grade wireless access points, maintaining a high-quality connection with a packet loss rate of less than one in ten thousand even in complex electromagnetic environments. This deep integration and optimization from chips, algorithms, hardware interfaces to physical execution mechanisms jointly ensures the highly reliable operation of the dynamic control system for hazardous sources in the manufacturing site based on the Industrial Internet of Things, effectively eliminating blind spots in safety management and achieving closed-loop control and precise strikes against hidden risks.

[0079] The system architecture of this invention possesses strong scalability, applicable not only to the aforementioned precision electronics and heavy equipment manufacturing, but also to a wide range of fields such as automobile assembly, chemical production, and aerospace. Addressing specific risk points in different industries, the system can quickly adapt to new operational scenarios simply by updating feature parameters in the safety benchmark library and reconfiguring the initial weights of the interaction gain coefficients on the cloud platform, demonstrating high technical versatility and market application potential. By introducing digital twin and edge cloud collaborative technologies, this invention constructs a comprehensive, all-weather safety defense system for future lighthouse factories and unmanned workshops, significantly reducing the economic losses and social impact of safety accidents on manufacturing enterprises, demonstrating significant social and economic benefits.

[0080] During the long-term operation of the system, data consistency and security are also key management priorities. The tiered early warning decision-making unit employs digital signature technology based on public key infrastructure when generating instructions to ensure that control instructions are not maliciously tampered with or missent during transmission. Before executing any action, the physical-level rigid execution unit performs a second two-way handshake with the edge computing gateway to confirm the legality and timeliness of the instruction. This robust communication security mechanism, combined with hardware-level mandatory execution, constitutes a smart, efficient, robust, and secure manufacturing site defense network, achieving a high degree of unity between technological advancement and engineering reliability.

[0081] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The various sensors, controllers, actuators, and communication protocols referenced in the embodiments of the present invention can all be implemented using commercially available components that meet performance standards, and do not depend on any specific, exclusive single product, which further ensures the feasibility of the technical solution of the present invention. The description of each embodiment in this specification aims to reveal the core design ideas and engineering implementation path of the present invention. The numerical values, parameters, and logical judgment logic involved should be finely tuned according to the specific industrial site environment to achieve the best control effect. The implementation of a dynamic control system for hazardous sources in manufacturing sites based on the Industrial Internet of Things marks a profound transformation in safety management in the manufacturing industry from passive response to proactive defense, from fuzzy assessment to precise control, and from information alerts to physical interception.

Claims

1. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, characterized in that, include: The multi-dimensional element sensing unit is used to acquire full real-time monitoring data, including personnel behavior trajectories, equipment operating parameters, material status information, environmental physical indicators, and process operation procedures, by utilizing the industrial Internet of Things sensor network deployed on the manufacturing site. The spatiotemporal synchronization processing unit is used to obtain the local timestamp of each sensing node and perform millisecond-level clock synchronization of all sensing nodes based on a precise time protocol. A three-dimensional spatial coordinate system is established with the reference zero point of the manufacturing site as the origin, and the data with positional attributes acquired by each sensing node are mapped to the unified three-dimensional spatial coordinate system. According to the preset time step, the synchronized data is filled into the corresponding spatial voxel grid in chronological order to construct a dynamic spatiotemporal data cube for the manufacturing site. The nonlinear risk coupling analysis unit is used to extract feature vectors based on the dynamic spatiotemporal data cube, and to identify risk factors in each single dimension by comparing the feature vectors with a preset safety benchmark library; and to construct a risk factor interaction matrix to determine the degree of mutual influence when different risk factors exist simultaneously in the same spatial voxel or adjacent voxels. Furthermore, by combining the interaction gain coefficient extracted from the historical accident case database through multivariate nonlinear regression analysis, a comprehensive risk index characterizing the overall threat level of the manufacturing site is calculated through a nonlinear coupling model. The graded early warning decision unit is used to compare the comprehensive risk index with preset multi-level risk thresholds and generate corresponding level control instructions based on the comparison results. The control instructions include reminder-level instructions, intervention-level instructions, and decision-making-level instructions. The physical-level rigid actuator is used to respond to the control commands by driving the audible and visual alarms deployed at the work site to issue early warning signals, or by sending parameter modification commands to the controllers of relevant equipment through the industrial IoT gateway to forcibly reduce the operating speed of the equipment or reduce the working pressure, or by directly controlling the trip coil of the electromagnetic contactor or circuit breaker connected in series in the power supply circuit of the equipment to perform hardware-level power supply cut-off, thereby realizing closed-loop control of hazardous sources at the manufacturing site.

2. The dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things as described in claim 1, characterized in that, The multi-dimensional element sensing unit is equipped with sensing nodes; the sensing nodes include: Ultra-wideband positioning base stations deployed above the work area are used to interact with positioning tags worn by workers to obtain their movement trajectories. A fieldbus gateway connected to an industrial control system is used to collect equipment operating parameters, including spindle speed, motor current, bearing temperature, and vibration spectrum. Radio frequency identification (RFID) readers are used to read electronic tags attached to material packaging to obtain material status information; The all-in-one environmental monitoring terminal integrates a dust concentration sensor, a combustible gas detector, an ambient temperature sensor, and a humidity sensor to acquire environmental physical indicators. Machine vision cameras are deployed at key workstations and work with edge computing modules to perform motion recognition, thereby acquiring process operation flow data.

3. The dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things as described in claim 1, characterized in that, The execution process of the spatiotemporal synchronization processing unit includes: synchronizing the clocks of each sensing node based on a precision time protocol to ensure that the clock synchronization accuracy of the entire system is better than 1 millisecond; establishing a three-dimensional Cartesian spatial coordinate system with the ground reference zero point of the manufacturing site as the origin, and using a spatial coordinate transformation matrix to map the data with positional attributes acquired by each sensing node into the three-dimensional Cartesian spatial coordinate system; and filling the synchronized multi-source heterogeneous data into a spatial voxel grid of a preset size according to a preset time step of 50 milliseconds, where each spatial voxel grid represents a 0.5m x 0.5m x 0.5m cubic area of ​​the manufacturing site.

4. The dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things as described in claim 1, characterized in that, The process of identifying risk factors by the nonlinear risk coupling analysis unit includes: extracting a floating-point number sequence containing 32 dimensions as a feature vector from the dynamic spatiotemporal data cube; comparing the feature vector with a security benchmark library stored in a local high-speed memory in real time; and determining that a single-dimensional risk factor exists in the spatial voxel when the value of any dimension in the feature vector deviates from the preset range of the security benchmark library by more than a preset 5% tolerance value.

5. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, as described in claim 1, is characterized in that... The interaction gain coefficient is determined by multiplying the basic weight coefficient of the risk factor interaction by a comprehensive logarithmic term; wherein, the comprehensive logarithmic term is the natural logarithm of the target summation value, the target summation value is the sum of a constant 1 and a spatial coupling factor; the spatial coupling factor is the product of a spatial overlap term and a spatial distance attenuation term; the spatial overlap term is the ratio of the intersection of the sets of influence ranges of the two risk factors in the three-dimensional spatial coordinate system to the total spatial volume involved in the workstation; the spatial distance attenuation term is an exponential function value determined based on the Euclidean distance between the centroids of the two risk factors and a preset spatial correlation feature length.

6. The dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things as described in claim 1, characterized in that, The comprehensive risk index consists of a linear weighted part and a non-linear weighted part; the linear weighted part is the weighted sum of each single-dimensional risk factor and its corresponding weight coefficient, which is used to reflect the basic accumulation of risk; the non-linear weighted part is based on the interaction gain coefficient to non-linearly weight paired risk factors, which is used to reflect the coupling amplification effect of risk in the spatiotemporal range.

7. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, as described in claim 1, is characterized in that... The multi-level risk thresholds preset in the hierarchical early warning decision unit include a first threshold, a second threshold, and a third threshold, wherein the third threshold is greater than the second threshold, and the second threshold is greater than the first threshold. When the comprehensive risk index is greater than or equal to the first threshold and less than the second threshold, the risk level is determined to be low risk, and an alert-level instruction is generated. When the comprehensive risk index is greater than or equal to the second threshold and less than the third threshold, the risk level is determined to be medium risk, and an intervention-level instruction is generated. When the comprehensive risk index is greater than or equal to the third threshold, the risk level is determined to be high risk, and a control level instruction is generated.

8. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, as described in claim 1, is characterized in that... The physical-level rigid actuator has hardware-level self-testing and feedback functions; after receiving the interruption-level command and executing the action, the physical-level rigid actuator monitors the current change of the power circuit in real time through the current transformer; when the expected current falling edge is not detected within the preset 500 millisecond response time, it is determined that the main actuator has failed, and the backup emergency braking logic set up upstream of the power distribution is triggered through the backup communication link.

9. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, as described in claim 1, is characterized in that... The system also includes a digital twin monitoring unit, used to receive the dynamic spatiotemporal data cube and the comprehensive risk index; The digital twin monitoring unit synchronously recreates the operating status of the manufacturing site in virtual space and marks the distribution and coupling evolution trend of risk factors in real time in the form of a dynamic heat map. The digital twin monitoring unit is also used to perform virtual simulation verification of the action results of the physical-level rigid execution unit using the built-in physics engine in order to evaluate the effectiveness of control measures.

10. A dynamic control system for hazardous sources in a manufacturing site based on the Industrial Internet of Things, as described in claim 1, is characterized in that, The system adopts a distributed architecture that combines edge computing and a cloud platform. The multi-dimensional element perception unit, the spatiotemporal synchronization processing unit, and the nonlinear risk coupling analysis unit operate in an edge computing gateway deployed at the edge of the manufacturing site, and the end-to-end latency from perception data input to risk analysis output is less than 20 milliseconds. The cloud platform is used to archive and back up control instructions and real-time monitoring data streams, and to optimize and iterate the interaction gain coefficient on a 24-hour cycle based on historical data using offline algorithms.

Images