Equipment real-time monitoring and intelligent scheduling system based on IoT and AI fusion

Abstract

The invention discloses an equipment real-time monitoring and intelligent scheduling system based on IoT and AI fusion, which comprises a sensing layer, a transmission layer, an edge computing layer, a cloud intelligent center layer and an application interaction layer, and is characterized in that the sensing layer maps heterogeneous data collected by a sensor to a unified space-time coordinate system by adopting a space-time encoder technology; the transmission layer is responsible for transmitting the data collected by the sensing layer to the edge computing layer and the cloud; the edge calculation layer calculates a three-dimensional risk field in real time based on the current operation state data of the equipment; the cloud intelligent center layer constructs digital twinborn bodies, and synchronizes the running state of equipment in real time; the application interaction layer presents equipment operation risk distribution through the risk field thermodynamic diagram; by constructing a multi-layer and multi-module collaborative architecture and fully applying advanced IoT and AI technologies, all-around real-time monitoring, accurate fault prediction and intelligent scheduling decision making of industrial equipment are realized, an intelligent, efficient and reliable operation environment is created for enterprises, and industrial production is promoted to be transformed into intelligentization and digitization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things, and in particular to a device real-time monitoring and intelligent scheduling system based on IoT and AI fusion. BACKGROUND

[0002] With the continuous advancement of Industry 4.0 and the concept of intelligent manufacturing, industrial enterprises have put forward higher and higher requirements for the monitoring and scheduling management of production equipment. The traditional device monitoring and scheduling system mainly has the following shortcomings: the traditional sensor network can only collect limited physical parameters such as temperature and pressure data, and it is difficult to effectively perceive the deep running state of the equipment and the invisible physical field information, and it cannot comprehensively and accurately reflect the real running condition of the equipment.

[0003] At the edge computing level, the traditional data processing method can only perform simple data aggregation and preliminary processing, and lacks the ability of deep mining and intelligent analysis of data, making it difficult to evaluate the running risk of the equipment and predict potential failures in real time.

[0004] The cloud and application interaction layer is usually independent of the edge computing layer and the bottom data collection layer, and the data sharing and collaborative work ability between systems is poor, resulting in serious information island problem, and it is difficult to realize global intelligent scheduling and optimization decision;

[0005] The traditional fault diagnosis method mainly relies on artificial experience and simple fault alarm system, and it is difficult to quickly and accurately locate the fault root cause and provide effective solutions, resulting in long equipment maintenance time and large production interruption loss;

[0006] In terms of production scheduling, the traditional scheduling system is mostly based on fixed rules and artificial experience, and lacks intelligent analysis and dynamic optimization ability of real-time production data and equipment state, and cannot adjust production plan and resource allocation in time according to actual situation, which is difficult to meet the demand of modern industrial production for efficient and flexible scheduling.

[0007] However, the equipment in modern industrial production process is increasingly complex, the production process is highly automated, and the production demand is variable, requiring the device real-time monitoring and intelligent scheduling system to have higher real-time, accuracy, intelligence and collaboration. The traditional system is difficult to meet these requirements, and an IoT and AI fusion-based device real-time monitoring and intelligent scheduling system is needed to overcome the above problems and improve the overall efficiency and competitiveness of industrial production. SUMMARY

[0008] The present application provides an IoT and AI fusion-based device real-time monitoring and intelligent scheduling system to solve the above technical problems.

[0009] The technical solution of the present application is as follows:

[0010] The real-time equipment monitoring and intelligent scheduling system based on the integration of IoT and AI includes the perception layer, transmission layer, edge computing layer, cloud intelligent hub layer, and application interaction layer, among which:

[0011] The perception layer uses spatiotemporal encoder technology to map heterogeneous data collected from different types of sensors at different times into a unified spatiotemporal coordinate system;

[0012] The transport layer is responsible for transmitting the data collected by the perception layer to the edge computing layer and the cloud;

[0013] The edge computing layer calculates the three-dimensional risk field in real time based on the current operating status data of the equipment. It uses containerized technology for deployment, enabling rapid construction, migration, and expansion of controllers, improving the flexibility and maintainability of the system.

[0014] The cloud-based intelligent core layer builds digital twins and synchronizes the actual operating status of equipment in real time. It uses multi-agent reinforcement learning algorithms to perform global optimization and scheduling of equipment maintenance sequences, production rhythm adjustments, and AGV path planning.

[0015] The application interaction layer clearly presents the distribution of equipment operation risks through intuitive risk field heat maps, and also displays the scheduling plan; the scheduling instructions generated by the intelligent scheduler are accurately distributed to each subsystem through the standard API gateway; based on the graph neural network, the equipment fault propagation path is deeply analyzed and traced to generate a "risk chain topology map" to accurately locate the initial failure point.

[0016] Preferably, the perception layer collects data by deploying multiple sensors for temperature, humidity, gas, water level, and vibration at key locations of industrial equipment;

[0017] Based on the vibration spectrum and current harmonic data generated by equipment operation, AI algorithms are used to generate invisible physical field parameters that cannot be directly obtained by traditional sensors.

[0018] A spatiotemporal encoder is used to map heterogeneous data collected from different types of sensors at different times into a unified spatiotemporal coordinate system.

[0019] Preferably, the transmission layer transmits the data collected by the perception layer to the edge computing layer and the cloud; it supports multiple transmission methods, including 4G / 5G wireless networks, fiber-optic wired networks, and internal private networks, which can be flexibly selected according to the actual scenarios and equipment distribution of the enterprise; at the same time, the transmission process adopts a lightweight protocol to ensure the efficiency and stability of data transmission.

[0020] Preferably, the edge computing layer calculates a three-dimensional risk field based on the current operating status data of the equipment. When the risk field value exceeds a preset threshold, an early warning mechanism is immediately triggered to promptly notify relevant personnel to take measures to avoid equipment failure and production accidents.

[0021] In an emergency, emergency control instructions are executed; containerization technology is used for deployment to achieve rapid construction, migration, and expansion of controllers, improving the flexibility and maintainability of the system.

[0022] Preferably, the cloud-based intelligent central layer uses the Internet of Things, big data and AI technologies to build a digital twin of the equipment, realizing a full range of digital mapping of the equipment's appearance, structure, performance and operating status; integrating MES work orders, WMS materials, and PCS process parameter data to break down information silos and achieve cross-system data fusion and sharing; and using multi-agent reinforcement learning algorithms to perform global optimization and scheduling of equipment maintenance sequences, production rhythm adjustments, and AGV path planning.

[0023] Preferably, the application interaction layer clearly presents the distribution of equipment operation risks through an intuitive risk field heat map, and displays the scheduling plan;

[0024] The dispatch instructions generated by the intelligent dispatcher are accurately distributed to the DCS, WMS, and MES subsystems through the standard API gateway to ensure that the systems work together to achieve automated and intelligent control of the production process.

[0025] Based on graph neural network technology, the propagation path of equipment failure is deeply analyzed and traced to generate a "risk chain topology map" to accurately locate the initial failure point, providing a key basis for fault repair and prevention.

[0026] Preferably, the perception layer includes a physical sensor array module, a virtual sensor engine module and a spatiotemporal alignment module, wherein:

[0027] The physical sensing array module deploys temperature and humidity sensors, gas sensors, water level sensors, and vibration sensors at key locations on industrial equipment and production sites to sense various physical parameters of the equipment's operating environment and the vibration state of the equipment itself in real time. The sensors convert the sensed physical quantities into electrical signals, which are then amplified, filtered, and pre-processed before being converted into digital signals via an analog-to-digital converter.

[0028] The virtual sensor engine module collects raw data from the physical sensor array, including vibration waveforms and current signals of the equipment, and performs preliminary processing on the data, including noise removal and signal filtering operations, to improve data quality;

[0029] Through fast Fourier transform, the frequency components and amplitude of the vibration signal are extracted from the original signal. Then, using convolutional neural network, the implicit physical field parameters inside the device are generated based on these characteristic parameters, including stress distribution field and thermal distribution field.

[0030] The spatiotemporal alignment module maps and coordinates the data collected from different sensors at different times so that the data can be represented and integrated in a unified spatiotemporal coordinate system.

[0031] Preferably, the edge computing layer includes a dynamic risk field engine module, a lightweight linkage controller module, an edge knowledge base module, and a data cache and preprocessing module, wherein:

[0032] The dynamic risk field engine module calculates a three-dimensional risk field in real time based on the collected equipment operating status data and continuously monitors the risk field value. Once it exceeds a preset threshold, an early warning mechanism is immediately triggered and an alarm notification is sent to relevant personnel so that timely measures can be taken to prevent equipment failures and production accidents.

[0033] The lightweight linkage controller module quickly responds and executes emergency control instructions in emergency situations to prevent accidents from escalating and ensure the safety of equipment and personnel. Containerization technology is used for deployment to enable the rapid construction, migration, and expansion of the lightweight linkage controller.

[0034] The edge knowledge base module is configured to store a library of historical device failure modes, which includes characteristic parameters, causes, solutions, and corresponding repair cases of common failures of various types of equipment. The edge knowledge base is regularly updated based on a federated learning algorithm.

[0035] The data cache and preprocessing module temporarily stores the data collected from the perception layer to ensure that the data will not be lost due to transmission problems, and provides complete and continuous data support for subsequent data processing and analysis; it performs preliminary processing on the cached data, including data cleaning, data completion, and data format conversion operations, to improve data quality and availability, and reduce the processing burden of the cloud-based intelligent center.

[0036] Preferably, the cloud-based intelligent hub layer includes a digital twin module, a multi-agent scheduler module, and an adaptive strategy library module, wherein:

[0037] The digital twin module integrates real-time device data from the perception layer, data processed by the edge computing layer, and business data from the enterprise information system to build a comprehensive and accurate digital model of the device;

[0038] Leveraging 3D modeling technology, physics engines, and data-driven approaches, we create a virtual image of the device, encompassing its appearance, structure, performance characteristics, and operational status. This ensures that the digital twin reflects the actual state of the physical device in real time.

[0039] Synchronize the operating status of physical equipment to the digital twin through a real-time data update mechanism;

[0040] Use visualization tools and interfaces to present digital twins to users in an intuitive way;

[0041] The multi-agent scheduler module allocates production tasks to different equipment, production lines and work cells in a reasonable manner according to production plans, order requirements, equipment status and resource availability factors, ensuring balanced and efficient task allocation;

[0042] Utilize path planning algorithms to plan optimal transportation and movement paths for AGVs and robots; dynamically adjust the production rhythm of production equipment based on the actual operating status of the equipment and the urgency of the production task, optimize the production process, improve production efficiency, and avoid equipment overload and idleness;

[0043] Adopting a multi-agent reinforcement learning algorithm, through collaboration and competition among agents, it continuously learns and optimizes the global scheduling strategy to improve the overall performance and adaptability of the system, and monitors and detects resource conflicts and task conflicts that may arise during the production process in real time.

[0044] The adaptive policy library module stores a large number of verified scheduling policies, fault handling policies, and resource optimization policies; regularly evaluates and updates the policies in the policy library. When encountering specific production scenarios and problems, the module quickly retrieves and calls the corresponding policies from the policy library based on the characteristics and requirements of the scenarios for processing. The called policies are adaptively adjusted based on the actual execution results and environmental changes.

[0045] Preferably, the application interaction layer includes an intelligent decision cabin module, a cross-system executor module, and a root cause analyzer module, wherein:

[0046] The intelligent decision-making module visualizes the risk data transmitted from the edge computing layer and the cloud-based intelligent hub layer, generates an intuitive risk field heat map, and displays intelligent scheduling plans in various visual forms such as charts, flow charts, and timelines. Users input different decision-making plans into the intelligent decision-making module based on the current production status and risk information, and the system evaluates the implementation effect of each plan through simulation operations.

[0047] The cross-system executor module receives scheduling instructions issued from the intelligent decision cabin module and the cloud intelligent hub layer, analyzes the instructions, extracts key information in the instructions, converts the analyzed instructions into instructions in specific formats and protocols that can be recognized and executed by each subsystem, and monitors the execution of the instructions by each subsystem in real time, and feeds back the execution results of each subsystem to the intelligent decision cabin module and the cloud intelligent hub layer, so that the system evaluates and adjusts the entire scheduling process.

[0048] The root cause analyzer module collects various operation data before and after equipment failure, constructs a graph model of the equipment and its operating environment based on a graph neural network algorithm, takes equipment components, sensors and production processes as nodes in the graph, and takes data correlation and fault propagation as edges, traces the propagation path of the fault by analyzing the characteristics of the nodes and edges in the graph, determines from which component and link the fault started to spread, and uses the graph embedding technology and classification algorithm of the graph neural network algorithm to evaluate the importance of the nodes in the graph model and identify the initial failure point that caused the failure, i.e., the root cause of the failure.

[0049] The present application has the following advantages:

[0050] 1. The present application optimizes and schedules the production tasks, equipment operation rhythm and material distribution through the multi-agent scheduler, so that the production process is smoother and more efficient, and the time waste caused by poor connection between production links, equipment idling and overloading is reduced.

[0051] 2. The present application realizes preventive maintenance of the equipment based on real-time evaluation and prediction of the equipment operation risk by the dynamic risk field engine, arranges maintenance work in advance before the equipment fails seriously, avoids high repair costs and equipment damage risks caused by sudden equipment failure, prolongs the service life of the equipment, and reduces the equipment operation and maintenance cost.

[0052] 3. The present application monitors the temperature, pressure, gas concentration and vibration parameters through various sensors in the perception layer, and evaluates the risk through the dynamic risk field engine, discovers potential safety hazards in time and issues warning signals.

[0053] 4. From the perception layer to the cloud-based intelligent hub layer, the system collects, transmits, stores, and analyzes equipment operation data and production process data in real time, providing production managers with comprehensive, accurate, and timely production information. This allows them to monitor production progress, equipment status, and product quality in real time, enabling refined management and control of the production process. Based on the digital twin module and root cause analyzer module, any link in the production process can be traced and analyzed.

[0054] 5. The present invention breaks down the information silos between various systems within the enterprise, and realizes the deep integration and sharing of equipment data, production data, and management data; by building a unified data platform and data model, it provides comprehensive and accurate data support for the production decision-making and operation management of the enterprise, and improves the digital management level of the enterprise; with the help of decision simulation and evaluation in the intelligent decision cabin module and intelligent optimization scheduling of the multi-agent scheduler, it provides a scientific and quantitative basis for the production decision-making of the enterprise, improves the accuracy and rationality of decision-making, promotes the development of the enterprise production model towards intelligence and automation, and enhances the market competitiveness of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a system block diagram of the real-time monitoring and intelligent scheduling system for equipment based on the integration of IoT and AI in the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] like Figure 1 As shown, the present invention provides a real-time monitoring and intelligent scheduling system for equipment based on the integration of IoT and AI, including a perception layer, a transmission layer, an edge computing layer, a cloud-based intelligent hub layer and an application interaction layer.

[0058] In the perception layer, the physical sensor array module deploys temperature and humidity sensors, gas sensors, water level sensors, and vibration sensors at key locations on industrial equipment and production sites to sense various physical parameters of the equipment's operating environment and the vibration status of the equipment itself in real time.

[0059] The sensed physical quantity is converted into an electrical signal through the sensor, and after amplification and filtering preprocessing, the analog signal is converted into a digital signal through an analog-to-digital converter;

[0060] The virtual sensor engine module collects raw data from the physical sensor array, including the device's vibration waveform and current signal, and performs preliminary processing on this data, including noise removal and signal filtering operations, to improve data quality.

[0061] Through fast Fourier transform, the frequency components and amplitude of the vibration signal are extracted from the original signal. Then, using convolutional neural network, the implicit physical field parameters inside the device are generated based on these characteristic parameters, including stress distribution field and thermal distribution field.

[0062] As an embodiment, when performing stress monitoring on a wind turbine blade, the virtual sensor engine module first collects raw vibration waveform data from a blade vibration sensor;

[0063] Perform wavelet denoising on these data to remove high-frequency noise interference;

[0064] Then, the FFT algorithm is used to perform spectrum analysis on the noise-reduced signal to obtain the frequency and amplitude characteristic parameters of the vibration signal;

[0065] These characteristic parameters are input into the trained CNN model to generate a heat map of the stress distribution inside the wind turbine blade. This allows us to visually see the stress concentration in different parts of the blade and detect potential fatigue crack failures in advance.

[0066] When performing wavelet denoising, the basic formula of wavelet transform is:

[0067] W(a,b)=(1 / √a)∫x(t)*ψ*((tb) / a)dt;

[0068] Where x(t) is the original signal, ψ(t) is the wavelet basis function, a is the scale factor, b is the translation factor, and W(a,b) is the coefficient after wavelet transformation. By selecting appropriate wavelet basis functions and scale and translation factors, the signal can be effectively decomposed to separate the noise component and the useful signal.

[0069] In the Fast Fourier Transform (FFT), for a discrete-time signal x(n), the spectrum X(k) is calculated as:

[0070] X(k)=Σx(n)*e^(-j2πkn / N), k=0,1,...,N-1;

[0071] Where N is the number of sampling points of the signal, n is the serial number of the sampling point, and j is the imaginary unit. FFT can be used to convert the time domain signal into the frequency domain signal to obtain the amplitude and phase information of different frequency components in the signal.

[0072] The space-time alignment module maps and coordinates the data collected from different sensors at different times, enabling representation and fusion in a unified space-time coordinate system, solving the space-time misalignment phenomenon caused by different time synchronization and spatial position differences, and providing a data basis for subsequent comprehensive analysis and processing;

[0073] As an example, for a large warehouse logistics center where multiple temperature and humidity sensors, smoke sensors, and cameras are deployed, due to the distribution of these devices in different locations, there may be slight differences in data collection time due to device clock differences. The space-time alignment module maps the collected data to a unified space-time coordinate system with the warehouse center as the reference by obtaining the location information and timestamp information of each device.

[0074] For example, convert temperature data T_i (collected at location (x_i, y_i, z_i), time t_i) to a representation in a unified space-time coordinate system, enabling effective fusion analysis of data from different locations and times. By determining whether a region has both temperature and humidity anomalies and smoke concentration increases at a specific time, potential fire hazards can be detected in a timely manner.

[0075] The transmission layer transmits the data collected by the perception layer to the edge computing layer and the cloud. It supports multiple transmission methods, including 4G / 5G wireless networks, optical fiber wired networks, and internal private networks, allowing flexible selection based on actual enterprise scenarios and device distribution. At the same time, lightweight protocols are used in the transmission process to ensure the efficiency and stability of data transmission.

[0076] In the edge computing layer, the dynamic risk field engine module calculates a three-dimensional risk field in real time based on the collected device operating state data, including temperature, pressure, vibration, and other sensor data. The calculation formula is:

[0077] Ψ=∑wi·σ(Si-τi(t));

[0078] Where Si represents the device state parameters collected by various sensors, τi(t) is a threshold value dynamically adjusted by LSTM based on device historical operating data, and the weight wi is optimized online through reinforcement learning algorithm to accurately assess the current operating risk of the device.

[0079] The risk field value is continuously monitored, and once it exceeds the preset threshold, the early warning mechanism is triggered, and an alarm notification is sent to relevant personnel to take timely measures to prevent device failure and production accidents.

[0080] As an example, in a reactor equipment monitoring scenario in chemical production, temperature sensors, pressure sensors, and vibration sensors are deployed to monitor the reactor's operating status. The dynamic risk field engine calculates the risk field value in real time based on the data collected by these sensors. Assuming the current temperature is T, the pressure is P, and the vibration amplitude is A, the corresponding risk calculation formula is:

[0081] Ψ=w1·σ(T-τ1(t))+w2·σ(P-τ2(t))+w3·σ(A-τ3(t));

[0082] Among them, τ1(t), τ2(t), and τ3(t) are the dynamic thresholds for temperature, pressure, and vibration, respectively, determined by LSTM after predicting the degradation trend of the reactor based on its historical operating data. w1, w2, and w3 are the weights of temperature, pressure, and vibration in risk assessment, respectively, which are continuously adjusted and optimized based on historical failure data and equipment importance factors through a reinforcement learning algorithm. When the calculated risk field value Ψ exceeds the preset safety threshold, the system immediately sends an alarm to the operator, indicating that there may be risks of equipment overheating, overpressure, and abnormal vibration, and timely inspection and treatment are required.

[0083] σ represents an activation function, such as the sigmoid function, which is used to map input values ​​to a specific range to ensure the rationality and comparability of risk field values. In this way, the dynamic risk field engine can comprehensively consider multiple equipment status parameters, comprehensively assess equipment operation risks, and provide timely warning responses.

[0084] The lightweight linkage controller module executes emergency control commands. In an emergency, if the equipment experiences abnormal conditions such as over-temperature, over-pressure, or leakage, it can quickly respond and execute emergency control commands, and accordingly open cooling devices, close valves, and stop equipment operations to prevent the accident from escalating and ensure the safety of equipment and personnel. The response delay is strictly controlled within 200ms to ensure the timeliness and effectiveness of control.

[0085] Containerization technology is used for deployment to enable rapid construction, migration, and expansion of lightweight linkage controller modules. Containerized deployment ensures stable operation of the controller in different environments, facilitates unified management and maintenance, and improves system flexibility and maintainability.

[0086] As an example, when monitoring and controlling the gas in the coal mine, when the gas concentration sensor detects that the gas concentration exceeds the standard, the lightweight linkage controller module quickly receives this emergency signal and responds within less than 200 ms, automatically turning on the ventilation equipment and increasing the ventilation volume to reduce the gas concentration and prevent gas explosion accidents; at the same time, the controller can also control the power equipment in the relevant area of the mine to be powered off according to the preset program to avoid electrical sparks causing gas explosion; through containerized deployment, the lightweight linkage controller can be quickly deployed and applied in different coal mine monitoring system environments to realize efficient linkage control of various coal mine safety equipment.

[0087] The edge knowledge base module sets up a storage device historical fault mode library, which includes characteristic parameters, fault causes, solutions and corresponding maintenance case information of common faults of various devices; when the device has an abnormality and the risk field value is abnormal, the edge knowledge base can quickly match similar fault modes in the library to provide reference for the on-site maintenance personnel, assist them to quickly locate the fault cause and take effective maintenance measures.

[0088] Based on the federated learning algorithm, the edge knowledge base is updated regularly. Specifically, the federated learning algorithm can continuously optimize the fault mode recognition model and solutions in the knowledge base through multi-client collaborative learning while protecting the privacy of enterprise data, ensuring the timeliness and accuracy of the knowledge base content to better adapt to changes in the device operating state and the emergence of new fault types.

[0089] As an example, in the robot welding workstation of an automobile manufacturing factory, the edge knowledge base module stores various fault modes of the robot welding equipment, including welding gun failure, mechanical arm jamming and welding parameter deviation; when a welding robot has an abnormal situation of welding quality decline, the risk field value calculated by the dynamic risk field engine increases, the system triggers an early warning, at this time, the edge knowledge base quickly matches the current device state data with the fault modes in the library and finds that the welding current instability may be caused by the electrode cap wear of the welding gun; the maintenance personnel refer to the corresponding maintenance cases and solutions provided by the knowledge base to replace the electrode cap and adjust the welding parameters in time, so that the robot returns to normal work; at the same time, through federated learning, the edge knowledge bases of multiple welding workstations in the factory learn from each other, continuously updating and improving the fault mode library; for example, when a robot at a certain site has a new welding defect fault mode caused by the change of welding material humidity and successfully solves it, this new fault mode and solution will be shared to the knowledge bases of other sites through the federated learning mechanism, improving the fault diagnosis and maintenance level of the robot welding equipment in the whole factory.

[0090] The data cache and preprocessing unit temporarily stores the data collected from the perception layer, ensures that the data will not be lost due to transmission problems, and provides complete and continuous data support for subsequent data processing and analysis;

[0091] And the cached data is preliminarily processed, including data cleaning, data padding, and data format conversion operations, to improve data quality and availability and reduce the processing burden of the cloud intelligent hub;

[0092] As an embodiment, in the wind turbine monitoring system of a wind farm, various sensors installed on the wind turbine continuously collect and upload data; the data cache and preprocessing module first temporarily stores these data, so that when network transmission fails, the data will not be lost, and after the network is restored, the uploading will continue; in the preprocessing stage, the collected wind speed data is cleaned to remove obvious error values caused by instantaneous sensor failure; for individual missing data points caused by transmission delay, linear interpolation method is used for padding; at the same time, the data collected by different sensors are converted into unified timestamp format and numerical unit format, so as to facilitate subsequent data fusion analysis and risk assessment, to ensure that the edge computing layer can accurately and efficiently process the wind turbine operation data and timely discover potential equipment failure risks.

[0093] In the cloud intelligent hub layer, the digital twin module integrates the real-time temperature, pressure, vibration data of the equipment from the perception layer, the data processed by the edge computing layer, and the work order information, material information, and process parameter data in MES, WMS, and PCS, to build a comprehensive and accurate digital model of the equipment;

[0094] Using 3D modeling technology, physical engine, and data-driven methods, a virtual mirror of the equipment is created, including the appearance, structure, performance characteristics, and running state of the equipment, and ensuring that the digital twin can reflect the actual state of the physical equipment in real time;

[0095] Through real-time data updating mechanism, the running state of the physical equipment is synchronized to the digital twin, including the start / stop state, running parameter changes, and fault state of the equipment, to ensure the consistency of the digital twin and the physical equipment in the time dimension;

[0096] Using visualization tools and interfaces, the digital twin is displayed to the user in an intuitive way, through 3D graphics, dashboards, and heat maps, to show the real-time running state, key performance indicators, and potential fault risk areas of the equipment, to facilitate the user to quickly understand the overall condition of the equipment;

[0097] As an example, in the stamping workshop of an automobile manufacturing factory, a digital twin of the stamping equipment is built, which integrates real-time temperature, pressure, and vibration data of the stamping machine, production order information in the MES system, process parameters, and material supply information in the WMS system; through 3D modeling, the appearance and internal structure of the stamping machine are accurately presented; during production, the digital twin synchronizes the running state of the stamping machine in real time, and when the temperature of a key component of the stamping machine abnormally rises, the digital twin immediately displays a high-temperature warning in the form of a heat map at the corresponding 3D model position, and updates the relevant temperature index values on the instrument panel; at the same time, combined with order information and material information, it can predict whether the material supply can meet the production demand in time, and whether the order task can be completed on time according to the current production progress and equipment state, providing decision support for production scheduling;

[0098] The multi-agent scheduler module reasonably allocates production tasks to different devices, production lines, and work units according to production plans, order demands, device states, and resource availability factors, ensuring the balance and efficiency of task allocation;

[0099] An improved A* algorithm is used to plan the optimal transportation and movement path for AGV and robot mobile devices, considering factors such as shortest path distance, shortest time, and avoiding high-risk areas;

[0100] According to the actual running state of the equipment and the urgency of the production task, the production rhythm of the production equipment is dynamically adjusted, the production process is optimized, the production efficiency is improved, and at the same time, equipment overload and idling are avoided;

[0101] A multi-agent reinforcement learning algorithm is used, through the cooperation and competition between agents, to continuously learn and optimize the global scheduling strategy, to improve the overall performance and adaptability of the system; specifically, the agent can represent different production elements such as equipment, production line, and material, and they make decisions and influence each other according to their own state and environmental feedback;

[0102] Real-time monitoring and detection of possible problems in the production process, such as multiple devices competing for the same material resources at the same time, multiple tasks being allocated to the same device at the same time, etc., through coordination mechanisms to solve conflicts in time and ensure the smooth progress of the production process;

[0103] As an example, for production line scheduling of an electronic manufacturing factory, the multi-agent scheduler module is responsible for coordinating multiple production links, including chip mounting, welding, testing; according to the production plan and order demand, different types of chip mounting tasks are allocated to multiple mounting production lines, and reasonable allocation is made in combination with the current state of the equipment; for material transportation, the scheduler uses an improved A* algorithm to plan the optimal path for AGV from the material warehouse to each production line, while considering avoiding high-risk areas such as equipment maintenance areas and personnel-intensive areas; during the production process, through multi-agent reinforcement learning algorithm, the agents corresponding to each production line constantly adjust the production rhythm and task execution order according to their own production progress, equipment state and material supply situation, in order to maximize the overall production efficiency; when the probability of failure of the equipment of a certain mounting production line increases, the corresponding agent will transfer part of the tasks to other normally operating production lines and adjust the subsequent production plan, while notifying the AGV to adjust the material transportation path to ensure that materials can be supplied to the new production line in time, minimizing the production interruption time;

[0104] The adaptive strategy library module stores a large number of verified scheduling strategies, fault handling strategies, and resource optimization strategies. These strategies can be empirical rules summarized based on historical data or intelligent strategies generated through machine learning algorithms. The strategies are classified and labeled according to different production scenarios, equipment types, and fault modes, facilitating quick retrieval and invocation.

[0105] The strategies in the strategy library are periodically evaluated and updated. Based on new production data, equipment operation conditions, and changes in business requirements, existing strategies are optimized and improved to ensure their effectiveness and applicability.

[0106] When encountering specific production scenarios and problems, the corresponding strategies are quickly retrieved and invoked from the strategy library for processing according to the characteristics and requirements of the scenario. When a specific fault occurs in the equipment, the corresponding fault handling strategy is invoked. When production tasks are urgent, the emergency scheduling strategy is invoked.

[0107] According to the actual execution effect and environmental changes, the invoked strategy is adaptively adjusted. If the strategy execution result does not achieve the expected effect, the strategy is modified and optimized through a feedback mechanism to improve the adaptability and robustness of the strategy.

[0108] As an example, in a machining workshop, the adaptive strategy library module stores scheduling strategies and fault handling strategies for different types of machine tools. When a milling machine experiences a spindle temperature overheat fault, the system calls the milling machine's spindle temperature overheat fault handling strategy from the strategy library based on the fault characteristics. The strategy includes reducing the cutting speed, increasing the coolant flow, and inspecting the spindle bearings. After implementing these measures, the system monitors changes in the equipment status in real time. If the temperature is still not effectively controlled, the strategy is adjusted through a feedback mechanism, further adjusting the cutting parameters and arranging more detailed equipment inspections. At the same time, the strategy library records the fault handling process and results, providing richer experience data for subsequent handling of similar faults. In terms of production scheduling, when the workshop receives a batch of urgent orders and needs to complete a large number of parts processing tasks in a short period of time, the system calls the urgent order scheduling strategy from the strategy library. The strategy includes prioritizing the allocation of high-efficiency equipment, adjusting production priorities, and increasing overtime production time. The scheduling strategy is dynamically adjusted based on the actual production progress and equipment operation status, and tasks are reallocated based on the actual processing efficiency and fault conditions of the equipment to ensure that urgent orders can be completed on time.

[0109] In the application interaction layer, the intelligent decision-making module visualizes the risk data transmitted from the edge computing layer and the cloud intelligent hub layer to generate an intuitive risk field heat map. The heat map uses different shades of color to indicate the risk level of equipment and production areas, helping users quickly identify high-risk areas.

[0110] Display intelligent scheduling plans in the form of charts, flow charts, and timelines, including equipment maintenance plans, production task adjustment plans, and material distribution routes, so that users can clearly understand the scheduling arrangements;

[0111] Based on the current production status and risk information, users input different decision-making plans into the intelligent decision-making module. The system evaluates the implementation effect of each plan through simulation calculations, including its impact on production efficiency, cost, and equipment utilization, providing a quantitative basis for users to select the best decision-making plan.

[0112] The user adjusts and confirms the scheduling plan according to the actual situation and issues instructions through a friendly human-computer interaction interface. The system then transmits the instructions to the cross-system actuators for corresponding operations.

[0113] As an example, in the production monitoring center of a smart manufacturing plant, the intelligent decision-making module displays a heat map of the risk field for the entire production workshop. The map shows that a certain processing equipment area is marked as a red high-risk area, indicating that the equipment may be at risk of failure. At the same time, the system displays multiple maintenance scheduling options for the equipment, including immediate shutdown for repairs, repairs after the current task is completed, and adjustments to production tasks to avoid critical processes. If the user chooses the option of repairing after the current task is completed, the intelligent decision-making module, through simulation evaluation, shows that this option can reduce production losses by 30% and maintain equipment utilization at a reasonable level. After confirmation, the user issues an instruction on the decision-making module interface, and the system transmits the instruction to the cross-system actuator, notifying the equipment maintenance personnel to perform repairs at the specified time and adjust the relevant production task schedule.

[0114] The cross-system actuator module receives the dispatch instructions from the intelligent decision cabin module and the cloud intelligent central layer, parses the instructions, and extracts key information from the instructions, including equipment operating parameters, production task adjustment requirements, and material distribution instructions;

[0115] Convert the parsed instructions into instructions of a specific format and protocol that can be recognized and executed by each subsystem, ensuring that the instructions can be correctly transmitted and executed between different systems;

[0116] Real-time monitoring of each subsystem's execution of instructions, including whether the equipment has adjusted its operating parameters according to the instructions, whether the production tasks have been executed according to the new plan, and whether the materials have been delivered according to the designated routes, to ensure that the instructions can be executed accurately and timely;

[0117] Feedback the execution results of each subsystem to the intelligent decision-making module and the cloud-based intelligent central layer so that the system can evaluate and adjust the entire scheduling process; if there is a failure or delay in the execution of an instruction during the execution process, an alarm message will be sent to the relevant modules in a timely manner;

[0118] As an example, in the production scheduling system of a chemical production enterprise, the cross-system actuator receives a scheduling instruction from the intelligent decision-making cabin, requiring the DCS system to adjust the feed rate of a certain reactor, and at the same time notifying the WMS system to trigger material scheduling, urgently delivering a certain key raw material from the warehouse to the area where the reactor is located, and requiring the MES system to reschedule subsequent production work orders. The cross-system executor first parses the instructions and extracts the specific operational requirements for DCS, WMS and MES; then, it converts the instructions into the corresponding communication protocols and data formats of each system, converts the DCS instructions into control signals of the SCADA system, and converts the WMS and MES instructions into corresponding API requests; during the instruction execution process, the cross-system executor monitors the execution status of each system in real time, such as whether the DCS has adjusted the opening of the reactor's feed valve, whether the WMS has arranged the AGV for material distribution, and whether the MES has re-arranged the production task sequence; if it is found that an AGV has a fault during the material distribution process, resulting in a delivery delay, the cross-system executor immediately feeds back the abnormal information to the intelligent decision cabin module and the cloud-based intelligent central layer. The system readjusts the scheduling plan based on the feedback, arranges spare AGVs for distribution, and adjusts the priority of production tasks;

[0119] The root cause analyzer module collects various operational data before and after equipment failures, including physical parameters collected by sensors, equipment operation logs, and production plan data. It cleans and normalizes this data to remove noise and invalid information and extract characteristic data related to the failure.

[0120] Based on a graph neural network algorithm, a graph model of the equipment and its operating environment is constructed, with equipment components, sensors, and production processes as nodes in the graph, and data associations and fault propagation relationships as edges. By analyzing the characteristics of the nodes and edges in the graph, the fault propagation path can be traced to determine the component and link from which the fault originated.

[0121] Using graph embedding technology and classification algorithms from graph neural networks, we assess the importance of nodes in the graph model and identify the initial failure point, or root cause, that leads to the fault. We then combine domain knowledge and a failure mode database to further verify and confirm the accuracy of the root cause.

[0122] The system presents the root cause of the fault and its propagation path to users in the form of topology diagrams and tree diagrams, and generates a detailed fault analysis report, including the root cause of the fault, the fault propagation process, the affected equipment and production link information, providing guidance for equipment maintenance and production recovery;

[0123] As an example, when used in a substation in a power system, a transformer fails, causing a power outage. The root cause analyzer module first collects the voltage, current, oil temperature, and oil level sensor operating data before and after the transformer failure, as well as the transformer's operation log and maintenance records. After preprocessing this data, a graph neural network algorithm is used to construct an equipment graph model of the substation, with the transformer, circuit breaker, relay protection device, and power line as nodes in the graph, and data association and fault propagation relationships as edges. Through graph neural network analysis, it is found that the fault propagation path starts from a winding of the transformer. Due to aging of the winding insulation, a short circuit occurs, which in turn causes a sharp increase in the transformer's oil temperature, ultimately causing the transformer protection device to trip. The root cause analyzer module presents this root cause location result to the operation and maintenance personnel in the form of a topological map and generates a fault analysis report that details the starting point, propagation process, and impact range of the fault. Based on the report, the operation and maintenance personnel conduct targeted inspection and replacement of the transformer winding to quickly restore power supply.

[0124] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

Claims

1. A real-time monitoring and intelligent dispatching system for equipment based on the integration of IoT and AI, featuring: It includes the perception layer, transmission layer, edge computing layer, cloud intelligent hub layer and application interaction layer, among which: The perception layer uses spatiotemporal encoder technology to map heterogeneous data collected from different types of sensors at different times into a unified spatiotemporal coordinate system; The transport layer is responsible for transmitting the data collected by the perception layer to the edge computing layer and the cloud; The edge computing layer calculates the three-dimensional risk field in real time based on the current operating status data of the equipment. It uses containerized technology for deployment, enabling rapid construction, migration, and expansion of controllers, improving the flexibility and maintainability of the system. The cloud-based intelligent core layer builds digital twins and synchronizes the actual operating status of equipment in real time. It uses multi-agent reinforcement learning algorithms to perform global optimization and scheduling of equipment maintenance sequences, production rhythm adjustments, and AGV path planning. The application interaction layer clearly presents the distribution of equipment operation risks through intuitive risk field heat maps, and also displays the scheduling plan; the scheduling instructions generated by the intelligent scheduler are accurately distributed to each subsystem through the standard API gateway; based on the graph neural network, the equipment fault propagation path is deeply analyzed and traced to generate a "risk chain topology map" to accurately locate the initial failure point.

2. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 1 is characterized by: The perception layer collects data by deploying temperature, humidity, gas, water level, and vibration sensors at key locations of industrial equipment. Based on the vibration spectrum and current harmonic data generated by equipment operation, AI algorithms are used to generate invisible physical field parameters that cannot be directly obtained by traditional sensors. A spatiotemporal encoder is used to map heterogeneous data collected from different types of sensors at different times into a unified spatiotemporal coordinate system.

3. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 1 is characterized by: The transmission layer transmits the data collected by the perception layer to the edge computing layer and the cloud; It supports multiple transmission methods, including 4G / 5G wireless networks, fiber-optic wired networks, and internal private networks, which can be flexibly selected based on the actual scenarios and equipment distribution of the enterprise; at the same time, the transmission process adopts a lightweight protocol to ensure the efficiency and stability of data transmission.

4. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 1 is characterized by: The edge computing layer calculates the three-dimensional risk field in real time based on the current operating status data of the equipment. When the risk field value exceeds the preset threshold, the early warning mechanism is immediately triggered to promptly notify relevant personnel to take measures to avoid equipment failure and production accidents. In an emergency, emergency control instructions are executed; containerization technology is used for deployment to achieve rapid construction, migration, and expansion of controllers, improving the flexibility and maintainability of the system.

5. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 1 is characterized by: The cloud-based intelligent hub layer uses the Internet of Things, big data, and AI technologies to build a digital twin of the equipment, achieving a comprehensive digital mapping of the equipment's appearance, structure, performance, and operating status; integrating MES work orders, WMS materials, and PCS process parameter data to break down information silos and achieve cross-system data fusion and sharing; and using a multi-agent reinforcement learning algorithm to perform global optimization and scheduling of equipment maintenance sequences, production rhythm adjustments, and AGV path planning.

6. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 1 is characterized by: The application interaction layer clearly presents the distribution of equipment operation risks through an intuitive risk field heat map, while also displaying the scheduling plan; The dispatch instructions generated by the intelligent dispatcher are accurately distributed to the DCS, WMS, and MES subsystems through the standard API gateway to ensure that the systems work together to achieve automated and intelligent control of the production process. Based on graph neural network technology, the propagation path of equipment failures is deeply analyzed and traced to generate a "risk chain topology map" to accurately locate the initial failure point, providing a key basis for fault repair and prevention.

7. The real-time monitoring and intelligent scheduling system for equipment based on the integration of IoT and AI according to claim 2 is characterized by: The perception layer includes a physical sensor array module, a virtual sensor engine module, and a spatiotemporal alignment module, wherein: The physical sensing array module deploys temperature and humidity sensors, gas sensors, water level sensors, and vibration sensors at key locations on industrial equipment and production sites to sense various physical parameters of the equipment's operating environment and the vibration state of the equipment itself in real time. The sensors convert the sensed physical quantities into electrical signals, which are then amplified, filtered, and pre-processed before being converted into digital signals via an analog-to-digital converter. The virtual sensor engine module collects raw data from the physical sensor array module, including the vibration waveform and current signal of the device, and performs preliminary processing on the data, including noise removal and signal filtering operations, to improve data quality; Through fast Fourier transform, the frequency components and amplitude of the vibration signal are extracted from the original signal. Then, using convolutional neural network, the implicit physical field parameters inside the device are generated based on these characteristic parameters, including stress distribution field and thermal distribution field. The spatiotemporal alignment module maps and coordinates the data collected from different sensors at different times so that the data can be represented and integrated in a unified spatiotemporal coordinate system.

8. The device real-time monitoring and intelligent scheduling system based on the integration of IoT and AI according to claim 4 is characterized by: The edge computing layer includes a dynamic risk field engine module, a lightweight linkage controller module, an edge knowledge base module, and a data cache and preprocessing module, wherein: The dynamic risk field engine module calculates a three-dimensional risk field in real time based on the collected equipment operating status data and continuously monitors the risk field value. Once it exceeds a preset threshold, an early warning mechanism is immediately triggered and an alarm notification is sent to relevant personnel so that timely measures can be taken to prevent equipment failures and production accidents. The lightweight linkage controller module responds quickly and executes emergency control instructions in emergency situations. To prevent accidents from escalating and ensure the safety of equipment and personnel; containerization technology is used for deployment to enable the rapid construction, migration, and expansion of lightweight linkage controllers; The edge knowledge base module is configured to store a library of historical device failure modes, which includes characteristic parameters, causes, solutions, and corresponding repair cases of common failures of various types of equipment. The edge knowledge base is regularly updated based on a federated learning algorithm. The data cache and preprocessing module temporarily stores the data collected from the perception layer to ensure that the data will not be lost due to transmission problems, and provides complete and continuous data support for subsequent data processing and analysis; it performs preliminary processing on the cached data, including data cleaning, data completion, and data format conversion operations, to improve data quality and availability, and reduce the processing burden of the cloud-based intelligent center.

9. The real-time monitoring and intelligent scheduling system for equipment based on the integration of IoT and AI according to claim 5 is characterized by: The cloud-based intelligent hub layer includes a digital twin module, a multi-agent scheduler module, and an adaptive strategy library module, wherein: The digital twin module integrates real-time device data from the perception layer, data processed by the edge computing layer, and business data from the enterprise information system to build a comprehensive and accurate digital model of the device; Leveraging 3D modeling technology, physics engines, and data-driven approaches, we create a virtual image of the device, encompassing its appearance, structure, performance characteristics, and operational status. This ensures that the digital twin reflects the actual state of the physical device in real time. Synchronize the operating status of physical equipment to the digital twin through a real-time data update mechanism; Use visualization tools and interfaces to present digital twins to users in an intuitive way; The multi-agent scheduler module allocates production tasks to different equipment, production lines and work cells in a reasonable manner according to production plans, order requirements, equipment status and resource availability factors, ensuring balanced and efficient task allocation; Utilize path planning algorithms to plan optimal transportation and movement paths for AGVs and robots; dynamically adjust the production rhythm of production equipment based on the actual operating status of the equipment and the urgency of the production task. Optimize production processes, improve production efficiency, and avoid equipment overload and idleness; Adopting a multi-agent reinforcement learning algorithm, through collaboration and competition among agents, it continuously learns and optimizes the global scheduling strategy to improve the overall performance and adaptability of the system, and monitors and detects resource conflicts and task conflicts that may arise during the production process in real time. The adaptive policy library module stores a large number of verified scheduling policies, fault handling policies, and resource optimization policies; regularly evaluates and updates the policies in the policy library. When encountering specific production scenarios and problems, the module quickly retrieves and calls the corresponding policies from the policy library based on the characteristics and requirements of the scenarios for processing. The called policies are adaptively adjusted based on the actual execution results and environmental changes.

10. The real-time monitoring and intelligent scheduling system for equipment based on the integration of IoT and AI according to claim 6 is characterized by: The application interaction layer includes an intelligent decision cabin module, a cross-system executor module, and a root cause analyzer module, wherein: The intelligent decision-making module visualizes the risk data transmitted from the edge computing layer and the cloud-based intelligent hub layer, generates an intuitive risk field heat map, and displays intelligent scheduling plans in various visual forms such as charts, flow charts, and timelines. Users input different decision-making plans into the intelligent decision-making module based on the current production status and risk information, and the system evaluates the implementation effect of each plan through simulation operations. The cross-system executor module receives scheduling instructions from the intelligent decision cabin module and the cloud-based intelligent central layer, parses the instructions, and extracts key information from the instructions; converts the parsed instructions into instructions of a specific format and protocol that each subsystem can recognize and execute; monitors the execution of the instructions by each subsystem in real time, and feeds back the execution results of each subsystem to the intelligent decision cabin module and the cloud-based intelligent central layer so that the system can evaluate and adjust the entire scheduling process; The root cause analyzer module collects various operating data before and after the equipment failure occurs, and builds a graph model of the equipment and its operating environment based on the graph neural network algorithm, with equipment components, sensors, and production processes as nodes in the graph, and data association relationships and fault propagation relationships as edges. By analyzing the characteristics of the nodes and edges in the graph, the propagation path of the fault is traced to determine which component and link the fault started to spread from. The graph embedding technology and classification algorithm of the graph neural network algorithm are used to evaluate the importance of the nodes in the graph model and identify the initial failure point that caused the fault, that is, the root cause of the fault.

Images