Distributed industrial equipment monitoring and management system and method based on AIoT

Abstract

The invention relates to the technical field of equipment monitoring and management, and particularly discloses an AI oT-based distributed industrial equipment monitoring and management system and method, and the system comprises a distributed equipment monitoring module which collects operation parameters, performance indexes and environment data in real time through a plurality of types of sensors disposed on industrial equipment, a dynamic hierarchical edge architecture is utilized to process data and evaluate a static health state, and task intelligent layering and cloud collaboration are supported. And the multi-modal perception analysis module generates an equipment association graph based on the collaborative task, extracts multi-modal features by using a cross-node collaborative algorithm, and realizes dynamic health assessment by combining a CNN-LSTM fusion model and equipment implicit dependency relationship analysis. And the distributed equipment management module fuses equipment dynamic and static health data, cooperates with a task association graph and a fault tracing model, and realizes efficient resource configuration and fault prevention. The system guarantees intelligent monitoring and management of the whole life cycle of industrial equipment through a cooperation mechanism of edge intelligence and cloud deep analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of device monitoring and management, and particularly to a distributed industrial device monitoring and management system and method based on AIoT. Background Art

[0002] Currently, in the era of the rapid development of Industry 4.0 and intelligent manufacturing, distributed industrial devices are widely used in various industrial fields. These devices are widely distributed, of various types and interrelated, and their stable operation is crucial for ensuring the efficiency, quality and safety of industrial production. Therefore, a complete set of distributed industrial device monitoring and management systems has become an urgent need for the development of modern industry. Traditional industrial device monitoring and management mainly rely on manual inspections and simple sensor monitoring. Manual inspections not only consume a large amount of manpower and time, but also, due to human subjective factors and limited observation ranges, it is difficult to comprehensively and real-time grasp the operating conditions of the devices. Although simple sensor monitoring can obtain some device operating parameters, its ability to analyze the complex correlation relationships between devices and multi-modal data is limited. With the rapid development of technologies such as the Internet of Things (IoT) and artificial intelligence (AI), a distributed industrial device monitoring and management system based on AIoT (the integration of artificial intelligence and the Internet of Things) has emerged. This system densely distributes various sensors on distributed industrial devices to collect rich operating parameters, performance indicators and environmental data in real time, and uses AI technology to deeply analyze these data. It can achieve precise perception of the device operating status, early warning of faults and intelligent scheduling management of devices, which helps to improve the automation and intelligent level of industrial production, reduce production costs, improve production efficiency and product quality, and is of great significance for promoting the transformation of industry to intelligence and has broad application prospects in future industrial development.

[0003] However, the existing distributed industrial device monitoring and management systems based on AIoT (the integration of artificial intelligence and the Internet of Things) cannot fully recognize the mutual influence of devices in collaborative tasks. When conducting device management and fault troubleshooting, key factors are easily overlooked, affecting production efficiency and device maintenance effects. Traditional methods are difficult to perform efficient feature extraction and fusion analysis. Relying only on a single data type or simple data processing methods, it is impossible to deeply understand the device operating status, which limits the ability to detect and accurately diagnose device faults at an early stage. In terms of device scheduling management, it is difficult to make optimal decisions based on the real-time status of devices, resulting in unreasonable device resource allocation, inability to adapt to complex and changing industrial production environments, and affecting overall production efficiency.

[0004] Therefore, the present invention proposes a distributed industrial device monitoring and management system and method based on AIoT. Summary of the Invention

[0005] The present invention provides a distributed industrial equipment monitoring and management system and method based on AIoT, including: The distributed equipment monitoring module, with the help of various sensors densely arranged on numerous distributed industrial equipment, collects operation parameters, performance indicators, and environmental data in real time, and performs processing and health assessment through a dynamically hierarchical edge intelligent processing architecture to obtain the static equipment health status of each device, which provides detailed and immediate data support for the basic health of the equipment, enabling managers to have a clear understanding of the current basic state of the equipment. The multimodal perception analysis module generates an equipment association map based on the collaborative tasks of all distributed industrial equipment, extracts and analyzes multimodal features using a cross-node collaborative algorithm, mines the implicit dependency relationships between devices based on the equipment association map, and then deeply analyzes the multimodal features by combining these implicit dependency relationships with a heterogeneous data fusion model integrating CNN-LSTM, so as to obtain the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task. This comprehensive and in-depth analysis method not only considers the characteristics of the equipment itself but also takes into account the complex relationships between devices in collaborative tasks, greatly improving the accuracy and comprehensiveness of the equipment health status assessment. Finally, the distributed equipment management module comprehensively considers the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task and the static equipment health status of all distributed industrial equipment to perform scheduling management on all distributed industrial equipment, enabling more scientific and reasonable equipment scheduling, improving the overall operation efficiency of the equipment, reducing the equipment failure risk, ensuring the stability and high efficiency of the industrial production process, and helping industrial enterprises improve production efficiency and management level.

[0006] The present invention provides a distributed industrial equipment monitoring and management system based on AIoT, including:

[0007] A distributed equipment monitoring module, configured to collect the operation parameters, performance indicators, and environmental data of all distributed industrial equipment in real time based on various sensors densely arranged on numerous distributed industrial equipment, and perform processing and health assessment based on a dynamically hierarchical edge intelligent processing architecture to obtain the static equipment health status of each distributed industrial equipment;

[0008] A multimodal perception analysis module, configured to generate an equipment association map based on the collaborative tasks of all distributed industrial equipment, extract and analyze multimodal features based on a cross-node collaborative algorithm, mine the implicit dependency relationships between devices based on the equipment association map, and analyze the multimodal features based on the heterogeneous data fusion model integrating CNN-LSTM and the implicit dependency relationships between devices to obtain the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task;

[0009] A distributed equipment management module, configured to perform scheduling management on all distributed industrial equipment based on the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task and the static equipment health status of all distributed industrial equipment.

[0010] Preferably, the distributed device monitoring module includes:

[0011] The multimodal data perception sub-module is used to collect the operation parameters, performance indicators and environmental data of distributed industrial devices in real time based on various sensors densely arranged on numerous distributed industrial devices;

[0012] The edge gateway initial deployment sub-module is used to deploy transmission and protocol conversion tasks for edge gateway nodes based on the real-time traffic and data protocol types of the data streams of the operation parameters, performance indicators and environmental data of distributed industrial devices, and obtain the first real-time deployment status of the edge gateway;

[0013] The edge intelligent processing architecture hierarchical sub-module is used to generate health assessment task threads based on the health assessment tasks of all distributed industrial devices, and perform intelligent hierarchical processing on the health assessment task threads based on the first real-time deployment status of the edge gateway to obtain the dynamic hierarchical status of the edge gateway and regional edge servers;

[0014] The health assessment sub-module is used to execute the health assessment tasks of all distributed industrial devices based on the dynamic hierarchical status of the edge gateway and regional edge servers, and obtain the static device health status of each distributed industrial device.

[0015] Preferably, the edge intelligent processing architecture hierarchical sub-module includes:

[0016] The task decomposition unit is used to decompose the health assessment tasks of all distributed industrial devices based on preset conditions to generate health assessment task threads;

[0017] The task complexity calculation unit is used to calculate the complexity of each sub-task based on the multi-dimensional eigenvalue of each sub-task in the health assessment task thread;

[0018] The task real-time requirement standard parsing unit is used to determine the task real-time requirements of each sub-task in the health assessment task thread and the processing delay threshold of the regional edge server;

[0019] The edge gateway node status determination unit is used to determine the current load, maximum load, remaining memory and total memory of each edge gateway node in the edge gateway based on the first real-time deployment status of the edge gateway;

[0020] The intelligent hierarchical unit is used to perform intelligent hierarchical processing on the health assessment task thread based on the current load, maximum load, remaining memory, total memory of each edge gateway node in the edge gateway, the total task real-time requirement of the health assessment task thread and the processing delay threshold of the regional edge server, and obtain the dynamic hierarchical status of the edge gateway and regional edge servers.

[0021] Preferably, the intelligent hierarchical unit includes:

[0022] An execution priority calculation sub-unit, configured to generate the task execution priority of each edge network node based on the current load, maximum load, remaining memory, total memory, total task real-time requirement of the health assessment task thread, regional edge server processing delay threshold, and hierarchical decision formula of each edge network node in the edge gateway;

[0023] A task calculation density determination sub-unit, configured to regard the ratio of the total task calculation amount of the health assessment task thread to the task real-time requirement of each sub-task as the task calculation density of each sub-task;

[0024] A hierarchical coefficient determination sub-unit, configured to regard the product of the task calculation density of each sub-task in the health assessment task thread and the data localization rate of the health assessment task thread, divided by the task execution priority of each edge network node, as the hierarchical coefficient of each edge network node relative to each sub-task in the health assessment task thread;

[0025] An intelligent hierarchical sub-unit, configured to, when the hierarchical coefficient of a single edge node relative to a single sub-task in the health assessment task scenario is less than a preset threshold, layer the corresponding sub-task to the corresponding edge node for execution; otherwise, layer the corresponding sub-task to the regional edge server for execution until the dynamic hierarchical status of the edge gateway and the regional edge server is obtained.

[0026] Preferably, the health assessment sub-module includes:

[0027] A cloud model call unit, configured to call the execution model adopted in each sub-task link of the health assessment task thread of the distributed industrial device in the cloud database respectively;

[0028] A model deployment and task execution unit, configured to, based on the dynamic hierarchical status of the edge gateway and the regional edge server, deploy the execution model adopted in each sub-task link of the health assessment task thread of the distributed industrial device to the corresponding nodes in the control edge gateway and the regional edge server respectively for task processing until the static device health status of each distributed industrial device is obtained.

[0029] Preferably, the multi-modal perception analysis module includes:

[0030] A device association graph generation sub-module, configured to generate a device association graph based on the collaborative tasks of all distributed industrial devices;

[0031] A multi-modal feature extraction sub-module, configured to extract and analyze multi-modal features from the operation parameters, performance indicators, and environmental data of all distributed industrial devices based on the cross-node collaboration algorithm;

[0032] The implicit dependency relationship mining sub-module is used to determine the data dependency relationships between devices based on the device association graph, and mine the implicit dependency relationships between devices based on all the data dependency relationships between devices in the device association graph;

[0033] The dynamic health status assessment sub-module is used to analyze multi-modal features based on the heterogeneous data fusion model integrating CNN-LSTM and the implicit dependency relationships between devices, and obtain the dynamic device health status of each distributed industrial device participating in each collaborative task.

[0034] Preferably, the device association graph generation sub-module includes:

[0035] The task structure unit is used to deconstruct each collaborative task to obtain all sub-link tasks in the collaborative task, and screen out all executable devices for each sub-link task among all distributed industrial devices;

[0036] The preferred device unit for sub-links is used to determine the preferred execution device for each sub-link task of each collaborative task based on all executable devices for each sub-link task in each collaborative task and the static device health status of each distributed industrial device;

[0037] The device association graph generation unit is used to generate the device association graph for each collaborative task based on the preferred execution devices for each sub-link task of each collaborative task.

[0038] Preferably, the distributed device management module includes:

[0039] The fault tracing graph generation sub-module is used to generate the fault tracing data flow graph for each collaborative task based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices;

[0040] The device scheduling management sub-module is used to perform scheduling management on all distributed industrial devices based on the dynamic health assessment values of all distributed industrial devices participating in each collaborative task, all explicit fault items, all implicit fault items and their corresponding incidence rates, the static health assessment values of all distributed industrial devices, all explicit fault items, all implicit fault items and their corresponding incidence rates, the fault tracing data flow graph of each collaborative task, and the device scheduling optimization model.

[0041] Preferably, the fault tracing graph generation sub-module includes:

[0042] The dynamic device health status analysis unit is used to extract the dynamic health assessment value, all explicit fault items, all implicit fault items and their corresponding incidence rates of each distributed industrial device from the dynamic device health status of each distributed industrial device participating in each collaborative task;

[0043] A static device health status analysis unit, which is used to extract the static health assessment value, all explicit fault items, all implicit fault items and corresponding incidence rates of each distributed industrial device from the static device health status of all distributed industrial devices;

[0044] A dynamic and static alignment comparison unit, which is used to align and compare all explicit fault items, all implicit fault items and corresponding incidence rates of each distributed industrial device participating in each collaborative task with all explicit fault items, all implicit fault items and corresponding incidence rates of all distributed industrial devices, and determine all explicit fault difference items, all implicit fault difference items and all difference incidence rates of each distributed industrial device participating in each collaborative task;

[0045] A fault data flow tracing unit, which is used to trace the source of all explicit fault difference items, all implicit fault difference items of each distributed industrial device participating in each collaborative task and all implicit fault items generating all difference incidence rates based on the device association graph, and obtain the fault tracing data flow graph of each collaborative task.

[0046] The present invention provides a method for monitoring and managing distributed industrial devices based on AIoT, which is applied to any of the above-mentioned distributed industrial device monitoring and management systems based on AIoT, and includes:

[0047] S1: Based on various sensors densely arranged on many distributed industrial devices, the operation parameters, performance indicators and environmental data of all distributed industrial devices are collected in real time, and processed and health evaluated based on a dynamically hierarchical edge intelligent processing architecture to obtain the static device health status of each distributed industrial device;

[0048] S2: Generate a device association graph based on the collaborative tasks of all distributed industrial devices, extract and analyze multi-modal features based on a cross-node collaborative algorithm, mine the implicit dependency relationships between devices based on the device association graph, and analyze multi-modal features based on a heterogeneous data fusion model integrating CNN-LSTM and the implicit dependency relationships between devices to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task;

[0049] S3: Schedule and manage all distributed industrial devices based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices.

[0050] The beneficial effects of the present invention compared with the prior art are as follows: The distributed device monitoring module, with the help of various sensors densely distributed on numerous distributed industrial devices, collects real-time operating parameters, performance indicators, and environmental data, and processes and conducts health assessments through a dynamically layered edge intelligent processing architecture to obtain the static device health status of each device, which provides detailed and immediate data support for the basic health of the device, enabling managers to have a clear understanding of the current basic state of the device. The multimodal perception analysis module generates a device association graph based on the collaborative tasks of all distributed industrial devices, extracts and analyzes multimodal features using cross-node collaborative algorithms, simultaneously mines the implicit dependencies between devices based on the graph, and then deeply analyzes the multimodal features by means of a heterogeneous data fusion model integrating CNN-LSTM in combination with these implicit dependencies, so as to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task. This comprehensive and in-depth analysis method not only considers the characteristics of the device itself but also takes into account the complex relationships between devices in collaborative tasks, greatly improving the accuracy and comprehensiveness of the assessment of the device health status. Finally, the distributed device management module comprehensively considers the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices to schedule and manage all distributed industrial devices, enabling more scientific and reasonable device scheduling, improving the overall operating efficiency of the devices, reducing the risk of device failures, ensuring the stability and efficiency of the industrial production process, and helping industrial enterprises improve production efficiency and management level.

[0051] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in this application document.

[0052] The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the accompanying drawings:

[0054] Figure 1 is the architecture diagram of the distributed industrial device monitoring and management system based on AIoT in the embodiment of the present invention;

[0055] Figure 2 is the schematic diagram of the intramodules of the distributed device monitoring module in the embodiment of the present invention;

[0056] Figure 3Schematic diagram of an intron module of a multimodal perception analysis module in an embodiment of the present invention;

[0057] Figure 4 Schematic diagram of an intron module of a distributed device management module in an embodiment of the present invention;

[0058] Figure 5 This is a flow chart of the AIoT-based distributed industrial equipment monitoring and management method in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0060] Example 1:

[0061] refer to Figure 1 , the present invention provides a distributed industrial equipment monitoring and management system based on AIoT, including:

[0062] The distributed equipment monitoring module collects operating parameters, performance indicators, and environmental data of all distributed industrial equipment in real time based on various sensors densely distributed across numerous distributed industrial equipment. This module then processes and assesses health based on a dynamically layered edge intelligent processing architecture to obtain the static health status of each distributed industrial equipment.

[0063] The multimodal perception and analysis module is used to generate a device association graph based on the collaborative tasks of all distributed industrial equipment, extract and analyze multimodal features based on a cross-node collaborative algorithm, mine implicit dependencies between devices based on the device association graph, and analyze multimodal features based on a heterogeneous data fusion model integrating CNN-LSTM and implicit dependencies between devices to obtain the dynamic device health status of each distributed industrial equipment participating in each collaborative task.

[0064] The distributed equipment management module is used to schedule and manage all distributed industrial equipment based on the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task and the static equipment health status of all distributed industrial equipment.

[0065] Distributed Device Monitoring Module: With the help of various sensors deployed in large numbers on numerous distributed industrial devices, such as temperature sensors, pressure sensors, etc., it collects in real time the operating parameters (such as equipment rotation speed, voltage, etc.), performance indicators (such as production efficiency, yield rate, etc.) and environmental data (such as workshop temperature, humidity, etc.) of all devices. Then, it processes these data using a dynamic hierarchical edge intelligent processing architecture. This architecture can intelligently allocate tasks to different levels such as edge gateways or regional edge servers according to different task requirements and device states, and on this basis, conduct a health assessment of the devices, and finally obtain the static device health status of each distributed industrial device, that is, the health degree of the device under the current relatively stable state (when not participating in collaborative tasks). For example, by analyzing the average operating parameters of the device over a period of time, combined with historical data and standard parameters, it is judged whether the device is in a normal and healthy state.

[0066] Multi-modal Perception Analysis Module: Based on the collaborative tasks jointly participated in by all distributed industrial devices, this module generates a device association graph, which shows the mutual relationships of devices in the collaborative tasks, such as which devices cooperate closely to complete specific subtasks. Through cross-node collaborative algorithms, it extracts and analyzes multi-modal features from various modal data such as operating parameters, performance indicators, and environmental data collected from devices, such as extracting features from different aspects such as the operating sound and vibration frequency of the device. At the same time, based on the device association graph, it mines the implicit dependency relationships between devices. For example, it is found that some devices seemingly operate independently, but there are hidden data associations under specific working conditions. Then, using a heterogeneous data fusion model that combines CNN-LSTM (CNN is good at processing spatial data such as images, and LSTM is suitable for processing time series data, and the combination of the two can better analyze complex multi-modal data), combined with the mined implicit dependency relationships, it deeply analyzes the multi-modal features, so as to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task, that is, the real-time health change of the device during the execution of the collaborative task. For example, in a certain collaborative production task, it monitors in real time the impact of the cooperation between devices on the health status of each device.

[0067] Distributed Device Management Module: It comprehensively considers the dynamic device health status of each distributed industrial device participating in each collaborative task, as well as the static device health status of all distributed industrial devices, to conduct scheduling management of all distributed industrial devices. For example, according to the dynamic health status of the device, it is known that a certain device has too high a load and its health status has declined in the current collaborative task. Combining its static health status, it decides whether to adjust the task allocation or perform maintenance on the device to achieve a reasonable configuration of device resources, ensure the efficient and stable operation of industrial production, and avoid production stagnation caused by device failures.

[0068] Example 2:

[0069] Based on Example 1, with reference to Figure 2 , the distributed device monitoring module includes:

[0070] The multimodal data perception sub-module is used to collect the operation parameters, performance indicators, and environmental data of distributed industrial devices in real time based on various sensors densely distributed on numerous distributed industrial devices; this module is responsible for using a large number of various sensors distributed on numerous distributed industrial devices to collect various types of data in real time. These data include the operation parameters of the device itself, such as the rotation speed, power, and operation duration of the device, which reflect the operation status of the device; performance indicators, such as the production accuracy and output efficiency of the device, which are used to measure the work effectiveness of the device; and the environmental data where the device is located, such as environmental temperature, humidity, dust concentration, etc. Environmental factors will affect the operation of the device. For example, the ambient temperature around the device is collected in real time through a temperature sensor, and the current parameters during the operation of the device are obtained through a current sensor, etc. These multimodal data provide a basis for the subsequent comprehensive analysis of the device.

[0071] The edge gateway initial deployment sub-module is used to deploy the transmission and protocol conversion tasks for edge gateway nodes based on the real-time traffic and data protocol type of the data stream of the operation parameters, performance indicators, and environmental data of distributed industrial devices, and obtain the first real-time deployment status of the edge gateway; this module deploys tasks for edge gateway nodes based on the real-time traffic of the data stream formed by the operation parameters, performance indicators, and environmental data collected from distributed industrial devices, as well as the data protocol type. Different types of data may have different protocol formats, and the real-time traffic will also change at any time. This module allocates transmission and protocol conversion tasks for each edge gateway node according to these factors. For example, if the data transmitted by a certain sensor uses a specific industrial protocol and the traffic is large, the edge gateway initial deployment sub-module will arrange corresponding nodes for protocol conversion so that it can be processed by subsequent systems and ensure the high efficiency of data transmission. After completing the task deployment, this module will obtain the first real-time deployment status of the edge gateway, that is, the instant status information of the task allocation and operation of each node of the edge gateway at this time.

[0072] The edge intelligent processing architecture hierarchical sub-module is used to generate health assessment task threads based on the health assessment tasks of all distributed industrial devices, and perform intelligent hierarchical classification on the health assessment task threads based on the first real-time deployment status of the edge gateway to obtain the dynamic hierarchical status of the edge gateway and the regional edge server; it disassembles these tasks based on the health assessment tasks of all distributed industrial devices and generates health assessment task threads. These task threads contain a series of operation steps required for health assessment of each device. Then, according to the first real-time deployment status of the edge gateway, that is, the current task allocation and operation of each node of the edge gateway, intelligent hierarchical classification is performed on the health assessment task threads. This means that it will consider factors such as the processing capacity and load of the edge gateway nodes, and reasonably allocate different tasks to different levels of the edge gateway and the regional edge server. For example, for some simple tasks with high real-time requirements and small computational complexity, they may be allocated to the edge gateway nodes for direct processing; while for complex tasks with large computational complexity, they are allocated to the regional edge server for processing. Through this intelligent hierarchical classification method, the dynamic hierarchical status of the edge gateway and the regional edge server is finally obtained. This status reflects the task allocation on different-level devices and will change in real time with the device status and task requirements.

[0073] The health assessment sub-module is used to execute the health assessment tasks of all distributed industrial devices based on the dynamic hierarchical status of the edge gateway and the regional edge server to obtain the static device health status of each distributed industrial device. This module executes the health assessment tasks of all distributed industrial devices based on the dynamic hierarchical status of the edge gateway and the regional edge server. That is to say, according to the task allocation determined by the previous intelligent hierarchical classification, the corresponding parts in the health assessment task threads are respectively run on the corresponding nodes of the edge gateway and the regional edge server. For example, the basic data preprocessing tasks with high real-time requirements are processed on the edge gateway nodes, and complex data mining and analysis tasks are performed on the regional edge server. Through these operations, the static device health status of each distributed industrial device is finally obtained, that is, the health degree of each device is evaluated from a relatively stable perspective, providing basic data support for device management.

[0074] Embodiment 3:

[0075] Based on Embodiment 2, the edge intelligent processing architecture hierarchical sub-module includes:

[0076] The task decomposition unit is used to perform task decomposition on the health assessment tasks of all distributed industrial devices based on preset conditions to generate health assessment task threads, where the constraint conditions are:

[0077]

[0078] In the formula, T iThe computational load of the i-th sub-task in the health assessment task thread, with the unit of FLOPS, P j is the processing capacity value of the j-th edge gateway node in the edge gateway, with the unit of FLOPS, max() is to take the maximum value, min() is to take the minimum value, and δ is the task balancing coefficient (dimensionless, recommended value is 0.8 - 1.2);

[0079] This unit decomposes the health assessment tasks of all distributed industrial devices based on preset conditions. Its constraint conditions are reflected by formulas, aiming to reasonably allocate the task computational load to each edge gateway node. Among them, the computational load T i of the i-th sub-task in the health assessment task thread is j associated with the processing capacity value P

[0080] of the j-th edge gateway node in the edge gateway. By taking the maximum value, minimum value, and task balancing coefficient, it ensures the balanced distribution of the sub-task computational load among edge gateway nodes. For example, assuming there are multiple edge gateway nodes with different processing capacity values, the task decomposition unit, according to this formula, allocates sub-tasks with larger computational loads to nodes with stronger processing capabilities, and at the same time, through the task balancing coefficient, avoids overloading a certain node, ensuring the efficient operation of the overall system. The generated health assessment task thread refines complex health assessment tasks into multiple manageable sub-tasks.

[0081]

[0082] In the formula, C i is the complexity of the i-th sub-task, f ip is the eigenvalue of the p-th dimension of the i-th sub-task, q is the total number of dimensions of the multi-dimensional eigenvalues of a single sub-task, ln is the natural logarithm with the base of the natural constant e and the value of e is 2.71828;

[0083] This unit is responsible for calculating the complexity of each sub-task. For example, a sub-task may be measured from multiple dimensions such as the number of vibration signal frequency components, current waveform related parameters, and process parameters. By accumulating the eigenvalues of these dimensions, the complexity of the sub-task is obtained. By quantifying the sub-task complexity, it helps to make reasonable allocations according to the node processing capabilities and task characteristics in the subsequent process.

[0084] The task real-time requirement standard parsing unit is used to determine the task real-time requirements of each subtask in the health assessment task thread and the processing delay threshold of the regional edge server; for each subtask in the health assessment task thread, its task real-time requirement is clarified, that is, the subtask must be completed within a specified time to meet the timeliness requirements of the system for device health assessment. At the same time, the processing delay threshold of the regional edge server is determined, which is the maximum delay time that the regional edge server can accept when processing tasks. For example, some subtasks for real-time monitoring of device operating status have high real-time requirements and must complete calculations and feedback within a short time; while the processing delay threshold of the regional edge server limits the longest time for it to process tasks, ensuring the timeliness of the entire health assessment process.

[0085] The edge gateway node status determination unit is used to determine the current load, maximum load, remaining memory, and total memory of each edge gateway node in the edge gateway based on the first real-time deployment status of the edge gateway; the current load reflects the amount of tasks that the node is currently processing; the maximum load represents the maximum task processing capacity that the node can bear; the remaining memory and total memory information are related to the data storage and operation space when the node processes tasks. For example, if an edge gateway node has a high current load, is close to the maximum load, and has less remaining memory, then careful consideration is needed when allocating tasks to avoid the node affecting the task processing efficiency or causing system failures due to overload.

[0086] The intelligent stratification unit is used to perform intelligent stratification on the health assessment task thread based on the current load, maximum load, remaining memory, total memory of each edge gateway node in the edge gateway, the total task real-time requirements of the health assessment task thread, and the processing delay threshold of the regional edge server, and obtain the dynamic stratification status of the edge gateway and the regional edge server. This unit performs intelligent stratification on the health assessment task thread based on the current load, maximum load, remaining memory, total memory of the edge gateway node, as well as the total task real-time requirements of the health assessment task thread and the processing delay threshold of the regional edge server. It will comprehensively consider these factors and reasonably allocate subtasks with different complexities and real-time requirements to the edge gateway and the regional edge server, so as to obtain the dynamic stratification status of the edge gateway and the regional edge server. For example, for subtasks with high real-time requirements, low complexity, and small data volume, they may be preferentially allocated to edge gateway nodes with low current load and sufficient remaining memory for processing; while for subtasks with high complexity, large amount of calculations but relatively low real-time requirements, they may be allocated to the regional edge server for processing. This dynamic stratification can flexibly adjust task allocation according to device status and task requirements, optimize system performance, and ensure the efficient and accurate execution of the distributed industrial device health assessment task.

[0087] Embodiment 4:

[0088] Based on Embodiment 3, the intelligent hierarchical unit includes:

[0089] An execution priority calculation subunit, configured to calculate the task execution priority of each edge network node in the edge gateway based on the current load, maximum load, remaining memory, total memory, total task real-time requirement of the health assessment task thread, area edge server processing delay threshold, and hierarchical decision formula:

[0090]

[0091] In the formula, θ j is the task execution priority of the j-th edge network node in the edge gateway, α is the weight value of the node load, L j is the current load (CPU utilization percentage) of the j-th edge network node in the edge gateway, L jmax is the maximum load of the j-th edge network node in the edge gateway, β is the weight value of the node memory, M j is the remaining memory of the j-th edge network node in the edge gateway, M jmax is the total memory of the j-th edge network node in the edge gateway, γ is the weight value of the task real-time requirement, D req is the total task real-time requirement of the health assessment task thread, in milliseconds, D th is the area edge server processing delay threshold, in milliseconds, where the values of α, β, and γ are set by domain experts according to historical data and business characteristics. For example, they can be taken as 0.3, 0.3, and 0.4 respectively;

[0092] This subunit calculates the task execution priority of each edge network node by comprehensively considering multiple factors through a complex hierarchical decision formula. In the formula, this part reflects the impact of the current load on the task execution priority through the ratio relationship between the current load and the maximum load, combined with the weight α. The lower the load, the relatively higher the priority.

[0093] In addition, this part in the formula reflects the role of the node memory situation on the priority. The more remaining memory, the higher the possible priority.

[0094] Meanwhile, this part reflects the impact of the task real-time requirement and the area edge server processing delay threshold on the task execution priority of the node. For example, if an edge network node has a low current load, a large amount of remaining memory, and a high task real-time requirement, then its task execution priority will be relatively high. Through this formula, a comprehensive task execution priority can be calculated for each edge network node, providing a basis for subsequent task layering.

[0095] A task calculation density determination subunit is used to regard the ratio of the total task calculation amount of the health assessment task thread to the task real-time requirement of each subtask as the task calculation density of each subtask; this subunit divides the total task calculation amount of the health assessment task thread by the task real-time requirement of each subtask to obtain the task calculation density of each subtask. The task calculation density can be understood as the amount of calculation that a subtask needs to complete per unit time. For example, if the real-time requirement of a subtask is 100 milliseconds and the total task calculation amount is equivalent to 1000 operations, then the task calculation density is 10 operations / millisecond. This indicator can help evaluate the calculation pressure of each subtask and has important reference value for subsequent determination of task allocation at different levels. Subtasks with high calculation pressure may be more suitable to be allocated to the edge servers with stronger processing capabilities, while subtasks with low calculation pressure can be considered to be allocated to the edge network gateway nodes.

[0096] A layering coefficient determination subunit is used to regard the product of the task calculation density of each subtask in the health assessment task thread and the data localization rate of the health assessment task thread, divided by the task execution priority of each edge network gateway node, as the layering coefficient of each edge network gateway node relative to each subtask in the health assessment task thread; this subunit multiplies the task calculation density of each subtask by the data localization rate of the health assessment task thread and then divides by the task execution priority of each edge network gateway node to obtain the layering coefficient of each edge network gateway node relative to each subtask in the health assessment task thread. The data localization rate represents the proportion of task data processed locally (at the location of the edge network gateway node). A higher data localization rate means that the task can be completed more locally, reducing data transmission overhead. The layering coefficient comprehensively considers the calculation characteristics of the task itself (task calculation density), the data processing location characteristics (data localization rate), and the execution priority of the edge network gateway node, and is used to measure the suitability of allocating a certain subtask to a specific edge network gateway node. For example, if a subtask has a low task calculation density, a high data localization rate, and a high execution priority of a certain edge network gateway node for this subtask, then the corresponding layering coefficient will be relatively high, indicating that this subtask is more suitable to be allocated to this edge network gateway node.

[0097] An intelligent hierarchical subunit is used to layer the corresponding subtask into the corresponding edge node for execution when there is a single edge node with a layering coefficient less than the preset threshold for a single subtask in the health assessment task site. Otherwise, the corresponding subtask is layered to the regional edge server for execution until the dynamic layering status of the edge gateway and the regional edge server is obtained. This subunit determines the execution location of the subtask based on the layering coefficient. When there is a single edge node with a layering coefficient less than the preset threshold for a single subtask in the health assessment task, it means that it is relatively appropriate to allocate this subtask to this edge node, and the corresponding subtask is layered into the corresponding edge node for execution; otherwise, if the layering coefficient is greater than or equal to the preset threshold, it indicates that this subtask may be more suitable for execution on the regional edge server, and it is layered to the regional edge server. By making such judgments and allocations for each subtask, the distribution of tasks between the edge gateway and the regional edge server is continuously adjusted until the dynamic layering status of the edge gateway and the regional edge server is obtained. This dynamic layering status will change in real time according to factors such as device operation conditions and task characteristics to achieve the optimal utilization of system resources and the efficient execution of tasks. For example, in the industrial production process, as the device status and task requirements change, the allocation of tasks between the edge gateway and the regional edge server will also be adjusted accordingly to ensure that the entire distributed industrial device monitoring and management system always maintains a good operating state.

[0098] Embodiment 5:

[0099] Based on Embodiment 2, the health assessment sub-module includes:

[0100] A cloud model calling unit is used to call the execution models adopted in each subtask link of the health assessment task thread of the distributed industrial device in the cloud database respectively; when conducting a health assessment of the distributed industrial device, different subtask links may require different execution models. The responsibility of the cloud model calling unit is to retrieve from the cloud database the execution models applicable to each subtask link in the health assessment task thread. The cloud database stores a large number of trained and verified models, which are optimized for different types of industrial device health assessment tasks. For example, for the subtask of evaluating whether the device operation parameters are abnormal, an anomaly detection model based on machine learning may be called;

[0101] The operation parameters can be divided into time-series parameters (such as temperature, pressure, rotational speed), frequency-domain parameters (such as vibration frequency), digital input / output parameters (such as valve status), and composite parameters (such as energy efficiency ratio), etc.;

[0102] The judgment task of whether the above operation parameters are abnormal can be achieved through an anomaly detection model obtained by pre-learning a large number of normal operation parameters and a large number of abnormal operation parameters;

[0103] For the subtask of analyzing the trend of device performance indicators, a time series prediction model may be called.

[0104] The trends of device performance indicators can be, for example, the change trend of the data processing speed of the device (such as the attenuation speed or aging speed of the processing speed), the change trend of the communication speed, etc. (the attenuation speed or aging speed of the communication speed, etc.);

[0105] The long short-term memory recurrent neural network can be used to learn the sequence data of the device's data processing speed and the change trend of the artificially calculated data processing speed, and obtain a model that can determine the change trend of the device's data processing speed;

[0106] Similarly, the long short-term memory recurrent neural network can be used to learn the sequence data of the device's communication speed and the change trend calculated manually, and obtain a model that can determine the change trend of the device's communication speed;

[0107] By calling these professional models, it provides strong support for accurately evaluating the health status of distributed industrial devices.

[0108] The model deployment and task execution unit is used to deploy the execution models adopted in each subtask link of the health assessment task thread of the distributed industrial device to the corresponding nodes in the control edge gateway and the regional edge server for task processing respectively based on the dynamic hierarchical status of the edge gateway and the regional edge server, until the static device health status of each distributed industrial device is obtained. This unit works according to the dynamic hierarchical status of the edge gateway and the regional edge server. As mentioned above, the edge intelligent processing architecture hierarchical sub-module has intelligently stratified the health assessment task thread and determined which subtasks are executed at the corresponding nodes of the edge gateway and which are executed at the regional edge server. The model deployment and task execution unit will, according to this stratification result, deploy the execution models adopted in each subtask link called from the cloud to the corresponding nodes in the edge gateway and the regional edge server respectively.

[0109] For example, if the execution model of a certain subtask is lightweight and suitable for quickly processing data locally, and according to the dynamic stratification, this subtask is assigned to the edge gateway node, then this unit will deploy this model to the corresponding edge gateway node for task processing; for complex models with large amounts of calculation, if their corresponding subtasks are assigned to the regional edge server, the models will be deployed to the regional edge server for execution. Through this precise model deployment, each subtask can be efficiently executed on the corresponding computing resources, and finally the static device health status of each distributed industrial device is obtained, providing data support for comprehensively understanding the basic health status of the device.

[0110] Example 6:

[0111] Based on Embodiment 1, the multimodal perception analysis module refers to Figure 3 , and includes:

[0112] The device association graph generation sub-module is used to generate a device association graph based on the collaborative tasks of all distributed industrial devices; in the industrial production scenario, numerous distributed industrial devices will jointly participate in various collaborative tasks. This sub-module generates a device association graph based on these collaborative tasks. It analyzes the collaboration relationships between devices in each collaborative task, such as which devices directly interact with data and which devices execute task steps sequentially, and presents the connections between devices in a graphical manner. For example, in the collaborative task of automobile manufacturing, stamping equipment, welding equipment, painting equipment, etc. have a clear sequence and data interaction in the entire production process. The device association graph generation sub-module will clearly present these relationships, with nodes representing devices and edges representing the associations between devices, thus generating a device association graph, laying a foundation for further analyzing the relationships between devices.

[0113] The multimodal feature extraction sub-module is used to extract and analyze multimodal features from the operating parameters, performance indicators, and environmental data of all distributed industrial devices based on the cross-node collaboration algorithm; distributed industrial devices generate multimodal data such as operating parameters, performance indicators, and environmental data. This sub-module uses the cross-node collaboration algorithm to extract and analyze multimodal features from these complex data. The cross-node collaboration algorithm can coordinate the data of different device nodes and comprehensively consider the characteristics of various data types. For example, extract features related to thermal stability from the operating parameters such as the temperature and pressure of the device; extract features reflecting the working efficiency of the device from the performance indicators such as production efficiency and yield rate; extract environmental-related features that may affect the device operation from the environmental data such as environmental humidity and dust concentration. In this way, comprehensively mine the useful information in the data and provide rich data features for accurately evaluating the dynamic health status of the device.

[0114] The implicit dependency mining sub-module is used to determine the data dependencies between devices based on the device association graph, and mine the implicit dependencies between devices based on all the data dependencies between devices in the device association graph; first, determine the data dependencies between devices based on the device association graph, that is, clarify which device's data output is the prerequisite for the data input of other devices, or which device's data change will affect the operation of other devices. On this basis, further mine the implicit dependencies between devices. These implicit dependencies are not intuitively visible and may only appear under specific working conditions, long-term operation or specific production processes. For example, in chemical production, the slight vibration change of a certain device seems to have nothing to do with another device on the surface, but after long-term operation, it is found that this vibration change will gradually affect the material transfer efficiency of another device. By deeply analyzing the data dependencies in the device association graph, this implicit dependency is mined, which helps to more comprehensively understand the internal connections between devices.

[0115] The dynamic health status assessment sub-module is used to analyze multi-modal features based on the heterogeneous data fusion model integrating CNN-LSTM and the implicit dependencies between devices, and obtain the dynamic device health status of each distributed industrial device participating in each collaborative task. This sub-module uses the heterogeneous data fusion model integrating CNN-LSTM, combines the mined implicit dependencies between devices, and deeply analyzes the multi-modal features, so as to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task. CNN (Convolutional Neural Network) is good at processing data with spatial structures, such as the distribution of device operation parameters in time series; LSTM (Long Short-Term Memory Network) has excellent performance in dealing with long-term dependencies in time series data. The combination of the two can better process the complex spatio-temporal features in multi-modal data. For example, when analyzing the vibration data (time series) of a device and the surface temperature distribution image (spatial data) of the device, the fusion model can effectively extract the features therein. Combining with the implicit dependencies between devices and considering the influence of other related devices on the health status of the target device, finally accurately evaluate the dynamic health status of each device during the execution of the collaborative task and timely discover potential health problems of the device.

[0116] Example 7:

[0117] Based on Example 6, the device association graph generation sub-module includes:

[0118] The task structure unit is used to deconstruct each collaborative task to obtain all sub - task links in the collaborative task, and screen out all executable devices for each sub - task link among all distributed industrial devices; in a complex industrial production environment, each collaborative task contains multiple specific sub - task links. The role of the task structure unit is to disassemble each collaborative task in detail and clearly sort out all the sub - task links therein. For example, in a collaborative task of assembling electronic products, it may be subdivided into sub - task links such as component handling, welding, and inspection.

[0119] After completing the disassembly of the sub - task links, this unit will screen out all devices that can execute each sub - task link among all distributed industrial devices. This means that factors such as the functions and performance of the devices need to be comprehensively considered to determine which devices have the ability to execute specific sub - task links. For example, for the sub - task link of welding, screen out all distributed industrial devices that have welding functions and can meet the welding process requirements (such as welding accuracy, speed, etc.). Through this step, a device candidate pool is provided for determining the optimal execution device for each sub - task link in the subsequent process.

[0120] The preferred sub - task device unit is used to determine the preferred execution device for each sub - task link of each collaborative task based on all executable devices for each sub - task link in each collaborative task and the static device health status of each distributed industrial device; this unit determines the preferred execution device based on all executable devices screened for each sub - task link and the static device health status of each distributed industrial device. The static device health status reflects the health degree of the device in a relatively stable state, and it is an important basis for evaluating whether the device can execute tasks efficiently and reliably.

[0121] When determining the preferred execution device, devices with good health status will be preferred because healthy devices are more likely to ensure the quality and efficiency of task execution and reduce production interruptions caused by device failures. For example, if multiple devices can execute the component handling task, but the static health assessment of one device shows that its various indicators are excellent and it runs stably, then this device is more likely to be determined as the preferred execution device for this sub - task link. This selection strategy based on device health status helps to improve the stability and reliability of the execution of the entire collaborative task.

[0122] The device association graph generation unit is used to generate the device association graph of each collaborative task based on the preferred execution devices of each sub - task link of each collaborative task. This unit generates the device association graph based on the preferred execution devices determined for each sub - task link of each collaborative task. The graph graphically shows the relationships between different sub - task links and the preferred execution devices in each collaborative task, as well as the associations between devices due to collaborative tasks.

[0123] For example, in the generated device association map, nodes represent the preferred execution devices, and the connections indicate the sequence or data interaction relationship of these devices in collaborative tasks. Taking the collaborative task of automobile manufacturing as an example, the map may show that the stamping equipment first stamps the steel plate into shape and then conveys it to the welding equipment for welding. The welded parts are then transferred to the painting equipment for painting. In this way, the device association situation of the entire collaborative task is clearly presented, providing an intuitive basis for subsequent analysis of the collaborative working mode and potential problems between devices.

[0124] Example 8:

[0125] Based on Example 1, referring to Figure 4 , the distributed device management module includes:

[0126] A fault traceability map generation sub-module, which is used to generate a fault traceability data flow map for each collaborative task based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices; this sub-module comprehensively considers the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices, and thus generates a fault traceability data flow map for each collaborative task. The dynamic device health status reflects the real-time health changes of the device during the execution of the collaborative task, while the static device health status reflects the health degree of the device under a relatively stable state.

[0127] By analyzing these two health statuses, the potential causes and propagation paths of device failures in collaborative tasks can be found. For example, when a certain device shows a performance decline (change in dynamic health status) in a collaborative task, by combining its static health status and the relevant data of other devices, it can be traced whether the failure is caused by the potential problems of the device itself (analyzed from the static health status) or due to the collaborative relationship with other devices (analyzed from the dynamic health status and the overall collaborative task). The fault traceability data flow map shows this information in a graphical way, clearly presenting the fault associations between devices, the sources of fault generation, and the paths of fault propagation, providing an intuitive and crucial basis for quickly locating and solving faults.

[0128] The device scheduling management sub-module is used to perform scheduling management on all distributed industrial devices based on the dynamic health assessment values of all distributed industrial devices participating in each collaborative task, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the static health assessment values of all distributed industrial devices, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the fault tracing data flow map of each collaborative task, and the device scheduling optimization model. When this sub-module performs device scheduling management, the information it relies on is very comprehensive. It includes the dynamic health assessment values of all distributed industrial devices participating in each collaborative task, which intuitively reflect the health status of the devices during the collaborative task; all explicit fault items, that is, the device fault situations that have been clearly manifested; all implicit fault items and their corresponding occurrence rates. These implicit faults may not have been significantly manifested yet, but have a certain probability of occurrence and pose a potential threat to the device operation; the static health assessment values of all distributed industrial devices, as a basic reference for device health; and the fault tracing data flow map of each collaborative task, which provides associated information related to the faults.

[0129] In addition, the device scheduling optimization model is also utilized. This model is obtained by using an artificial neural network to learn from the dynamic health assessment values of all distributed industrial devices participating in a single lineage task, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the static health assessment values of all distributed industrial devices, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the fault tracing data flow map of each collaborative task, and the strategies for manual scheduling management based on the aforementioned data. Therefore, by inputting the dynamic health assessment values of all distributed industrial devices participating in each collaborative task, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the static health assessment values of all distributed industrial devices, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the fault tracing data flow map of each collaborative task into the device scheduling optimization model, a scheduling management strategy for all distributed industrial devices is obtained, and all distributed industrial devices are scheduled and managed according to the scheduling management strategy output by this model.

[0130] Embodiment 9:

[0131] Based on Embodiment 8, the fault tracing map generation sub-module includes:

[0132] Dynamic Equipment Health Status Analysis Unit, which is used to extract the dynamic health assessment value, all overt fault items, all latent fault items and corresponding incidence rates of each distributed industrial equipment from the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task; during the operation of industrial equipment, each distributed industrial equipment will present a dynamic health status when participating in collaborative tasks. The responsibility of this unit is to conduct a detailed analysis of this dynamic equipment health status and extract key information from it. Specifically, it is to obtain the dynamic health assessment value of each distributed industrial equipment in the collaborative task, which comprehensively reflects the real-time health degree of the equipment during task execution. For example, this assessment value is obtained through the monitoring and analysis of the equipment's real-time operating parameters and performance indicators. The higher the value, the better the current health status of the equipment.

[0133] At the same time, this unit is also responsible for finding out all overt fault items, that is, those equipment faults that have been manifested obviously, such as abnormal noises of the equipment, over-temperature alarms, etc. In addition, it will identify all latent fault items and corresponding incidence rates. Latent fault items refer to potential problems that may exist in the equipment but have not been fully manifested. For example, some components show slight wear, but it has not affected the normal operation of the equipment. Through data analysis and model prediction and other means, these latent fault items and their occurrence probabilities can be estimated. By extracting this information, it provides an important basis for comprehensively understanding the health changes and potential fault risks of the equipment in the collaborative task subsequently.

[0134] Static Equipment Health Status Analysis Unit, which is used to extract the static health assessment value, all overt fault items, all latent fault items and corresponding incidence rates of each distributed industrial equipment from the static equipment health status of all distributed industrial equipment; corresponding to the dynamic equipment health status, this unit focuses on analyzing the static equipment health status of all distributed industrial equipment. From the static health status data, the static health assessment value of each distributed industrial equipment is extracted, which is a quantitative index of the overall health level of the equipment in a relatively stable state. For example, this assessment value is obtained based on the data of regular maintenance and detection of the equipment, statistical analysis of long-term operating parameters, etc.

[0135] Just like the analysis of dynamic equipment health status, it will also sort out all overt fault items, latent fault items and corresponding incidence rates. These static fault information reflect the problems and potential risks that may exist in the equipment under normal conditions and provide basic data for the long-term maintenance and management of the equipment. For example, through the analysis of factors such as the equipment's historical fault records and aging degree, the overt and latent fault items and incidence rates in the static situation are determined. Different from dynamic analysis, static analysis focuses more on the inherent and relatively stable health characteristics of the equipment itself.

[0136] A dynamic and static alignment comparison unit is used to align and compare all the explicit fault items, all the implicit fault items and the corresponding occurrence rates of each distributed industrial device participating in each collaborative task with those of all distributed industrial devices, so as to determine all the explicit fault difference items, all the implicit fault difference items and all the difference occurrence rates of each distributed industrial device participating in each collaborative task; this unit compares the information extracted by the dynamic device health status analysis unit and the static device health status analysis unit. Specifically, it aligns and compares the explicit fault items, implicit fault items and their occurrence rates of each distributed industrial device participating in each collaborative task with the corresponding information of all distributed industrial devices. Through this comparison, it can clearly determine the explicit fault difference items and implicit fault difference items of each distributed industrial device participating in each collaborative task, as well as the occurrence rates of these differential faults.

[0137] For example, in a certain collaborative task, a specific implicit fault item occurs in a device during dynamic operation, and its occurrence rate is different from that of the same type of implicit fault items of this device and other devices in the static case. The dynamic and static alignment comparison unit can find out this difference. These difference items are crucial for understanding the special fault situations of devices in collaborative tasks, and may reveal factors such as special working conditions during task execution and mutual influences between devices that lead to fault changes.

[0138] A fault data flow tracing unit is used to trace the origin of all the explicit fault difference items, all the implicit fault difference items and all the implicit fault items that generate all the difference occurrence rates of each distributed industrial device participating in each collaborative task based on the device association graph, and obtain the fault tracing data flow graph of each collaborative task. Based on the differential fault information analyzed by the previous unit, this unit uses the device association graph to conduct fault tracing. The device association graph shows the mutual relationships of devices in collaborative tasks, including information such as data interaction and task sequence. The fault data flow tracing unit uses these relationships to trace the origin of the explicit fault difference items, implicit fault difference items and implicit fault items that cause the difference occurrence rates of each distributed industrial device participating in each collaborative task.

[0139] For example, if an explicit fault difference item occurs in a device during a collaborative task, through the device association graph, it can be traced which upstream device has abnormal data and which collaborative operation link leads to this fault. Finally, through such a tracing process, the fault tracing data flow graph of each collaborative task is generated. This graph intuitively shows the propagation path and generation causes of faults between devices, provides a key visualization tool for the fault diagnosis and maintenance of industrial devices, helps technicians quickly locate the root cause of problems, take effective solutions, and ensure the stable operation of industrial production.

[0140] Example 10:

[0141] The present invention provides a distributed industrial equipment monitoring and management method based on AIoT, which is applied to any one of the distributed industrial equipment monitoring and management systems based on AIoT in Examples 1 to 9. Refer to Figure 5 , including:

[0142] S1: Based on various sensors densely distributed on numerous distributed industrial equipment, real-time collect the operation parameters, performance indicators and environmental data of all distributed industrial equipment, and perform processing and health assessment based on a dynamically hierarchical edge intelligent processing architecture to obtain the static equipment health status of each distributed industrial equipment;

[0143] S2: Generate an equipment association map based on the collaborative tasks of all distributed industrial equipment, extract and analyze multi-modal features based on a cross-node collaborative algorithm, mine the implicit dependence relationships between equipment based on the equipment association map, analyze the multi-modal features based on a heterogeneous data fusion model integrating CNN-LSTM and the implicit dependence relationships between equipment, and obtain the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task;

[0144] S3: Perform scheduling management on all distributed industrial equipment based on the dynamic equipment health status of each distributed industrial equipment participating in each collaborative task and the static equipment health status of all distributed industrial equipment.

[0145] By separately obtaining the static and dynamic health status of the equipment, the above method not only considers the health basis of the equipment under its own stable state, but also takes into account the real-time changes and mutual influences of the equipment in the collaborative tasks, comprehensively and accurately evaluates the equipment health, and greatly improves the accuracy and comprehensiveness of the evaluation. Based on the comprehensive equipment health status information for scheduling management, more scientific and reasonable equipment scheduling can be achieved, the overall operation efficiency of the equipment can be improved, the equipment failure risk can be reduced, the stable and efficient operation of industrial production can be guaranteed, and the production efficiency and management level of industrial enterprises can be enhanced.

[0146] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A distributed industrial equipment monitoring and management system based on AIoT, characterized in that, Including: A distributed device monitoring module, which is used to collect the operation parameters, performance indicators and environmental data of all distributed industrial devices in real time based on various sensors densely distributed on numerous distributed industrial devices, and perform processing and health assessment based on a dynamically hierarchical edge intelligent processing architecture to obtain the static device health status of each distributed industrial device; A multi-modal perception analysis module, which is used to generate a device association map based on the collaborative tasks of all distributed industrial devices, extract and analyze multi-modal features based on a cross-node collaborative algorithm, mine the implicit dependency relationships between devices based on the device association map, and analyze multi-modal features based on a heterogeneous data fusion model integrating CNN-LSTM and the implicit dependency relationships between devices to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task; A distributed device management module, which is used to perform scheduling management on all distributed industrial devices based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices.

2. The distributed industrial equipment monitoring and management system based on AIoT according to claim 1, wherein, The distributed device monitoring module includes: A multi-modal data perception sub-module, which is used to collect the operation parameters, performance indicators and environmental data of distributed industrial devices in real time based on various sensors densely distributed on numerous distributed industrial devices; An edge gateway initial deployment sub-module, which is used to deploy transmission and protocol conversion tasks for edge gateway nodes based on the real-time traffic and data protocol types of the data streams of the operation parameters, performance indicators and environmental data of distributed industrial devices, and obtain the first real-time deployment status of the edge gateway; An edge intelligent processing architecture layering sub-module, which is used to generate health assessment task threads based on the health assessment tasks of all distributed industrial devices, and perform intelligent layering on the health assessment task threads based on the first real-time deployment status of the edge gateway to obtain the dynamic layering status of the edge gateway and regional edge servers; A health assessment sub-module, which is used to execute the health assessment tasks of all distributed industrial devices based on the dynamic layering status of the edge gateway and regional edge servers to obtain the static device health status of each distributed industrial device.

3. The distributed industrial equipment monitoring and management system based on AIoT according to claim 2, characterized in that, The edge intelligent processing architecture layering sub-module includes: A task decomposition unit, which is used to decompose the health assessment tasks of all distributed industrial devices based on preset conditions to generate health assessment task threads; A task complexity calculation unit, which is used to calculate the complexity of each sub-task based on the multi-dimensional eigenvalue of each sub-task in the health assessment task thread; A task real-time requirement standard parsing unit, which is used to determine the task real-time requirements and the processing delay threshold of the regional edge server for each sub-task in the health assessment task thread; An edge gateway node status determination unit, which is used to determine the current load, maximum load, remaining memory and total memory of each edge gateway node in the edge gateway based on the first real-time deployment status of the edge gateway. An intelligent hierarchical unit for intelligently hierarchizing health assessment task threads based on the current load, maximum load, remaining memory, total memory, total task real-time requirement of the health assessment task threads, and the processing delay threshold of the regional edge server for each edge node in the edge gateway, so as to obtain the dynamic hierarchical status of the edge gateway and the regional edge server.

4. The distributed industrial equipment monitoring and management system based on AIoT according to claim 3, characterized in that, The intelligent hierarchical unit includes: An execution priority calculation subunit for generating the task execution priority of each edge node in the edge gateway based on the current load, maximum load, remaining memory, total memory, total task real-time requirement of the health assessment task threads, the processing delay threshold of the regional edge server, and a hierarchical decision formula. A task calculation density determination subunit for taking the ratio of the total task calculation amount of the health assessment task threads to the task real-time requirement of each subtask as the task calculation density of each subtask. A hierarchical coefficient determination subunit for taking the ratio of the product of the task calculation density of each subtask in the health assessment task threads and the data localization rate of the health assessment task threads to the task execution priority of each edge node as the hierarchical coefficient of each edge node relative to each subtask in the health assessment task threads. An intelligent hierarchical subunit for, when the hierarchical coefficient of a single edge node relative to a single subtask in the health assessment task scenario is less than a preset threshold, hierarchizing the corresponding subtask to the corresponding edge node for execution; otherwise, hierarchizing the corresponding subtask to the regional edge server for execution until the dynamic hierarchical status of the edge gateway and the regional edge server is obtained.

5. The distributed industrial equipment monitoring and management system based on AIoT according to claim 2, wherein, The health assessment sub-module includes: A cloud model invocation unit for invoking the execution models used in each subtask link of the health assessment task threads of distributed industrial devices in the cloud database respectively. A model deployment and task execution unit for, based on the dynamic hierarchical status of the edge gateway and the regional edge server, deploying the execution models used in each subtask link of the health assessment task threads of distributed industrial devices to the corresponding nodes in the control edge gateway and the regional edge server respectively for task processing until the static device health status of each distributed industrial device is obtained.

6. The distributed industrial equipment monitoring and management system based on AIoT according to claim 1, characterized in that, The multi-modal perception analysis module includes: A device association graph generation sub-module for generating a device association graph based on the collaborative tasks of all distributed industrial devices. A multi-modal feature extraction sub-module for extracting and analyzing multi-modal features from the operating parameters, performance indicators, and environmental data of all distributed industrial devices based on a cross-node collaboration algorithm. A hidden dependency relationship mining sub-module for determining the data dependency relationships between devices based on the device association graph and mining the hidden dependency relationships between devices based on all the data dependency relationships between devices in the device association graph. A dynamic health status assessment sub-module for analyzing multi-modal features based on a heterogeneous data fusion model integrating CNN-LSTM and the hidden dependency relationships between devices to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task.

7. The AIoT-based distributed industrial equipment monitoring and management system according to claim 6, characterized in that, The device association graph generation sub-module includes: A task structure unit, which is used to deconstruct each collaborative task to obtain all sub - task tasks in the collaborative task, and screen out all executable devices for each sub - task task among all distributed industrial devices; A sub - task device preference unit, which is used to determine the preferred execution device for each sub - task task of each collaborative task based on all executable devices for each sub - task task in each collaborative task and the static device health status of each distributed industrial device; A device association graph generation unit, which is used to generate a device association graph for each collaborative task based on the preferred execution device for each sub - task task of each collaborative task.

8. The distributed industrial equipment monitoring and management system based on AIoT according to claim 1, characterized in that A distributed device management module, including: A fault traceability graph generation sub - module, which is used to generate a fault traceability data flow graph for each collaborative task based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices; A device scheduling management sub - module, which is used to perform scheduling management on all distributed industrial devices based on the dynamic health assessment values of all distributed industrial devices participating in each collaborative task, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the static health assessment values of all distributed industrial devices, all explicit fault items, all implicit fault items and their corresponding occurrence rates, the fault traceability data flow graph of each collaborative task, and the device scheduling optimization model.

9. The AIoT-based distributed industrial equipment monitoring and management system according to claim 8, wherein The fault traceability graph generation sub - module includes: A dynamic device health status analysis unit, which is used to extract the dynamic health assessment value, all explicit fault items, all implicit fault items and their corresponding occurrence rates of each distributed industrial device from the dynamic device health status of each distributed industrial device participating in each collaborative task; A static device health status analysis unit, which is used to extract the static health assessment value, all explicit fault items, all implicit fault items and their corresponding occurrence rates of each distributed industrial device from the static device health status of all distributed industrial devices; A dynamic - static alignment comparison unit, which is used to align and compare all explicit fault items, all implicit fault items and their corresponding occurrence rates of each distributed industrial device participating in each collaborative task with all explicit fault items, all implicit fault items and their corresponding occurrence rates of all distributed industrial devices, and determine all explicit fault difference items, all implicit fault difference items and all difference occurrence rates of each distributed industrial device participating in each collaborative task; A fault data flow traceability unit, which is used to trace the source of all explicit fault difference items, all implicit fault difference items and all implicit fault items generating all difference occurrence rates of each distributed industrial device participating in each collaborative task based on the device association graph, and obtain the fault traceability data flow graph of each collaborative task.

10. A distributed industrial device monitoring and management method based on AIoT, characterized in that, Applied to the AIoT - based distributed industrial device monitoring and management system described in any one of claims 1 to 9, including: S1: Real-time collect the operation parameters, performance indicators of all distributed industrial devices and environmental data based on various sensors densely distributed in many distributed industrial devices, and perform processing and health assessment based on a dynamically hierarchical edge intelligent processing architecture to obtain the static device health status of each distributed industrial device; S2: Generate a device association map based on the collaborative tasks of all distributed industrial devices, extract and analyze multi-modal features based on a cross-node collaborative algorithm, mine the implicit dependency relationships between devices based on the device association map, and analyze multi-modal features based on a heterogeneous data fusion model integrating CNN-LSTM and the implicit dependency relationships between devices to obtain the dynamic device health status of each distributed industrial device participating in each collaborative task; S3: Perform scheduling management on all distributed industrial devices based on the dynamic device health status of each distributed industrial device participating in each collaborative task and the static device health status of all distributed industrial devices.

Images