Livestock machinery fault monitoring system based on edge calculation

Abstract

The invention discloses a livestock mechanical fault monitoring system based on edge calculation, and relates to the related technical field of mechanical fault monitoring, and the system comprises a mechanical transmission network building module which is used for building a multi-layer mechanical transmission network; the computing power configuration module is used for carrying out computing power configuration according to the criticality, the transmission dependency degree and the fault propagation influence of the node equipment and building a cloud-side platform structure; the monitoring execution link obtaining module is used for connecting a mechanical equipment sensing network to obtain a monitoring execution link; and the fault abnormal information identification module is used for dynamically distributing edge tasks and identifying fault abnormal information according to the change characteristics of the acquired data. The technical problem of poor mechanical fault monitoring and identification precision caused by insufficient real-time performance, weak fault propagation analysis capability and unreasonable computing power resource allocation of animal husbandry mechanical fault monitoring in the prior art is solved, and the technical effects of efficiently and accurately identifying faults, dynamically optimizing edge computing power and reducing the load of a cloud platform are achieved.

Description

Technical Field

[0001] This application relates to the technical field of mechanical fault monitoring, specifically to a livestock machinery fault monitoring system based on edge computing. Background Technology

[0002] As the livestock industry develops towards large-scale and intelligent operations, the automation level and complexity of livestock machinery and equipment are increasing. These machines typically consist of multiple transmission components and working units. If a critical piece of equipment malfunctions, the failure can spread rapidly through the transmission network, paralyzing the entire production system. Therefore, efficiently and accurately monitoring the operating status of livestock machinery and promptly identifying faults and anomalies has become crucial. Traditional mechanical fault monitoring relies on centralized cloud computing or localized sensor detection, which struggles to meet real-time requirements and lacks global analysis of fault propagation relationships between devices, easily leading to misjudgments or missed detections. Edge computing can significantly reduce data transmission latency and improve real-time processing capabilities. However, how to rationally design an edge computing architecture and dynamically allocate computing resources based on the transmission characteristics and fault propagation patterns of livestock machinery to achieve accurate identification and early warning of mechanical faults remains a pressing challenge.

[0003] Therefore, current technologies suffer from several technical problems, including insufficient real-time monitoring of livestock machinery faults, weak fault propagation analysis capabilities, and unreasonable allocation of computing resources, resulting in poor accuracy in monitoring and identifying mechanical faults. Summary of the Invention

[0004] This application provides a livestock machinery fault monitoring system based on edge computing, which solves the technical problems in the existing technology of insufficient real-time monitoring of livestock machinery faults, weak fault propagation analysis capabilities, and unreasonable allocation of computing resources, resulting in poor accuracy of mechanical fault monitoring and identification. It achieves the technical effects of efficient and accurate fault identification, dynamic optimization of edge computing power, and reduction of cloud platform load.

[0005] This application provides a livestock machinery fault monitoring system based on edge computing. The system includes: a mechanical transmission network establishment module, used to analyze the transmission relationship and working coupling characteristics between livestock machinery equipment and establish a multi-layer mechanical transmission network; a computing power configuration module, used to configure computing power according to the criticality, transmission dependence, and fault propagation influence of each node equipment in the multi-layer mechanical transmission network, deploy edge nodes, and build a cloud-edge platform structure; a monitoring execution link acquisition module, used to connect the mechanical equipment sensing network according to the causal relationship between mechanical equipment sensing data and fault events, and obtain the monitoring execution link between the mechanical equipment sensing network and the multi-layer mechanical transmission network; and a fault anomaly information identification module, used to dynamically allocate edge tasks according to the changing characteristics of the collected data of the mechanical equipment sensing network and the computing power of the monitoring execution link, and identify fault anomaly information according to the processing results of the edge tasks through the cloud-edge platform structure, based on the multi-layer mechanical transmission network.

[0006] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: identifying the functional levels of the main drive equipment, linkage equipment, and auxiliary equipment based on the transmission relationship and working coupling characteristics; analyzing the direct transmission relationship and logical dependency relationship between each equipment based on the functional level; calculating the key metric of each equipment in the transmission network based on the direct transmission relationship and logical dependency relationship between each equipment; and establishing the multi-layer mechanical transmission network based on the equipment correlation of the functional level, direct transmission relationship, logical dependency relationship, and key metric.

[0007] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: building a hierarchical structure according to the functional hierarchy, with each device in the functional hierarchy as a node; generating a device structure mapping relationship based on the direct transmission relationship and logical dependency relationship, and establishing node connection edges; and establishing the multi-layer mechanical transmission network based on the nodes, node connection edges, and edge attributes, using the key metric as edge attributes.

[0008] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: determining the radar map of the sensing device based on the sensing data attributes, sensing data characteristics, and sensing range of the sensing device; analyzing the influencing factors and causal relationships of the fault events based on historical sample records of livestock machinery fault events, and establishing a fault event heat map by analyzing the influencing factors and causal relationships with the fault events as the center; and obtaining the mapping relationship between the sensing device and the fault events by overlaying and mapping the fault event heat map with the radar map of the sensing device, thereby establishing the machinery sensing network.

[0009] In a possible implementation, the livestock machinery fault monitoring system based on edge computing further performs the following processing: assigning influence weights to different application factors in the fault event heatmap to obtain a multidimensional causal intensity vector; assigning a sensing intensity and sensitivity index to each sensing data point in the radar map of the sensing device to obtain a sensing capability vector; performing similarity matching between the multidimensional causal intensity vector and the sensing capability vector, filtering associated sensing devices based on the matching degree to obtain the mapping relationship between the sensing devices and the fault events; and constructing a matching map between the sensing devices and the fault events based on the mapping relationship between the sensing devices and the fault events to obtain the machinery sensing network, which is used to form causal path connection relationships in the machinery sensing network.

[0010] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: using graph neural networks and causal graph reasoning, a causal chain between changes in the characteristics of the sensing signal and the type of equipment fault is established; with the sensing equipment as the input node and the fault event as the target node, a directed weighted graph is constructed, wherein the edge weight represents the intensity of causal influence, and the mechanical equipment sensing network is obtained.

[0011] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: obtaining data feature changes through real-time data collection from the mechanical equipment sensing network; for sensing nodes with high causal weight in the mapping relationship between sensing devices and fault events, prioritizing the triggering of corresponding edge nodes for local analysis; expanding edge tasks according to the influence range of the transmission path involved in the multi-layer mechanical transmission network, matching computing power with task-activated edge nodes, and allocating tasks according to the monitoring execution link when the task load of the edge node meets the allocation conditions.

[0012] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: when the task load of the edge node exceeds the allocation conditions, computing power collaborative matching is performed in the cloud-edge platform structure based on the data feature changes to obtain collaborative nodes; and task allocation is performed based on the task load status of the collaborative nodes.

[0013] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: based on the processing results of matching edge nodes or cloud platforms, it integrates the correlation characteristics of fault events and propagates them in the multi-layer mechanical transmission network to obtain fault propagation information; it determines the mechanical fault level according to the key metrics of the propagating mechanical equipment; and it generates structured fault anomaly information, including fault location, fault type, fault level, occurrence time, and response measures, based on the fault propagation information and its mechanical fault level.

[0014] In a possible implementation, the livestock machinery fault monitoring system based on edge computing also performs the following processing: based on the cloud-edge platform structure, it analyzes the remaining computing power of each edge node and the computational transmission relationship with the target task edge node; according to the data feature changes and key metric indicators of fault events, it optimizes the solution with the remaining computing power of each edge node and the computational transmission relationship with the target task edge node, with the goal of maximizing the response speed, to obtain the collaborative node, which is either an edge node or a cloud platform.

[0015] This application proposes a livestock machinery fault monitoring system based on edge computing, comprising: a mechanical transmission network establishment module for building a multi-layered mechanical transmission network; a computing power configuration module for allocating computing power based on the criticality of node equipment, transmission dependence, and fault propagation impact, thus constructing a cloud-edge platform structure; a monitoring execution link acquisition module for connecting to the mechanical equipment sensing network to obtain the monitoring execution link; and a fault anomaly information identification module for dynamically allocating edge tasks and identifying fault anomaly information based on the changing characteristics of collected data. This system addresses the technical problems in existing livestock machinery fault monitoring technologies, such as insufficient real-time performance, weak fault propagation analysis capabilities, and unreasonable allocation of computing resources, leading to poor fault detection accuracy. It achieves the technical effects of efficient and accurate fault identification, dynamic optimization of edge computing power, and reduced cloud platform load. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 A schematic diagram of the structure of a livestock machinery fault monitoring system based on edge computing provided in this application embodiment.

[0018] Figure 2 This is a schematic diagram illustrating the execution process of the mechanical transmission network establishment module in the livestock machinery fault monitoring system based on edge computing provided in this application embodiment.

[0019] Figure labeling: Mechanical transmission network establishment module 10, computing power configuration module 20, monitoring execution link acquisition module 30, fault and abnormal information identification module 40. Detailed Implementation

[0020] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides an edge computing-based fault monitoring system for livestock machinery, such as... Figure 1 As shown, the system includes: The mechanical transmission network establishment module 10 is used to analyze the transmission relationship and working coupling characteristics between livestock machinery and equipment, and to establish a multi-layer mechanical transmission network.

[0024] Preferably, in livestock production systems, mechanical equipment such as feed mixers, milking machines, and ventilation fans are typically interconnected through mechanical transmissions or electrical controls, such as gears, belts, chains, and hydraulic systems, forming a collaborative working network. Analyzing the transmission relationships and operational coupling characteristics between livestock machinery refers to analyzing the power transmission, motion dependence, and fault mutual influence mechanisms among the mechanical equipment. A multi-layered mechanical transmission network is a structured model of these relationships to more accurately monitor and predict mechanical faults. Specifically, the transmission relationships of livestock machinery may include: power transmission paths, such as an electric motor driving a reducer, which in turn drives the mixing shaft, forming a motor-reducer-mixing shaft transmission chain; motion dependence, where the operation of some equipment depends on the output of other equipment, such as a vacuum pump failure in a milking machine causing the entire milking system to fail; and fault coupling effects, where an abnormality in one piece of equipment, such as bearing wear, may affect other equipment through the transmission chain, such as causing belt slippage or motor overload. The characteristics of operational coupling are manifested in several ways: mechanical coupling, such as physical connections like gear meshing and belt drives; functional coupling, such as the coordination between the feed conveyor system and the mixer (if the conveyor belt jams, the mixer may run idle or be overloaded); and temporal coupling, where some equipment needs to be started and stopped in a specific order, such as starting the fan before starting the heating system, otherwise malfunctions may occur. Based on the transmission relationships and operational coupling characteristics, modeling is performed from different dimensions, including physical layer hardware connections, functional layer task dependencies, and fault propagation layer impact paths. This describes the mechanical / electrical connections between equipment, the collaborative relationships between equipment in the process flow, and how faults spread from one piece of equipment to other equipment. Finally, a multi-layered mechanical transmission network is established to ensure that fault propagation paths are analyzed from a global perspective and the source of faults is quickly located, thereby achieving more intelligent and reliable operation and maintenance management of livestock machinery.

[0025] Furthermore, such as Figure 2 As shown, the specific configuration of the mechanical transmission network establishment module 10 further includes: identifying the functional levels of the main drive device, linkage device, and auxiliary device based on the transmission relationship and working coupling characteristics; analyzing the direct transmission relationship and logical dependency relationship between each device based on the functional level; calculating the key metric index of each device in the transmission network based on the direct transmission relationship and logical dependency relationship between each device; and establishing the multi-layer mechanical transmission network based on the device correlation of the functional level, direct transmission relationship, logical dependency relationship, and key metric index.

[0026] Preferably, based on the transmission relationship and working coupling characteristics, livestock machinery equipment is divided into three categories according to functional importance: main drive equipment, linkage equipment, and auxiliary equipment. Among them, the main drive equipment is the core power source, providing the main power of the system, such as the main motor of the feed mixer and the vacuum pump motor of the milking system. If it fails, the entire system will be paralyzed. The linkage equipment is the transmission / execution unit, which depends on the main drive equipment for operation and transmits power or performs specific operations, such as the pulley driven by the main motor, the reducer of the milking machine, the drive shaft, and the mixing paddle. The auxiliary equipment is a support unit that does not directly affect the core transmission, but ensures the stable operation of the system, such as the lubricating oil pump of the reducer and the cooling fan of the motor. Analyze the direct transmission relationships and logical dependencies between devices based on functional hierarchy to clarify fault propagation paths and distinguish between hardware and logical faults. Specifically, direct transmission relationships refer to physical power transmission paths, usually achieved through mechanical connections such as gears and belts, such as motor → coupling → reducer → mixing shaft. Logical dependencies refer to functional collaboration or control timing dependencies between devices, which may not have direct physical connections. For example, feed conveyor belts need to run before the mixer starts, otherwise they will be blocked. Temperature control systems rely on environmental sensor data, which logically affects the start and stop of ventilation equipment.

[0027] Preferably, based on the direct transmission relationships and logical dependencies between devices, key metrics for each device in the transmission network are calculated. These metrics are then used to quantitatively assess the importance of each device within the network. Key metrics include criticality, transmission dependency, and fault propagation impact. Criticality refers to the degree of impact of a device failure on the overall system, with the main drive device having the highest criticality. Transmission dependency refers to the dependence of other devices on that device; for example, a reducer depends on a motor (dependency of 1.0), while a lubrication pump might have a dependency of 0.3. Fault propagation impact is the number of devices that may experience cascading failures due to a device failure; for example, a motor failure could cause the reducer, agitator shaft, and belt to all stop, resulting in an impact of 3. Finally, by integrating functional levels, direct transmission relationships, logical dependencies, and the device correlation of key metrics, a three-layer network model is constructed, including a physical transmission layer, a functional dependency layer, and a fault propagation layer. This results in a multi-layered mechanical transmission network, ensuring accurate location of the root cause of the failure and enabling early intervention to predict potential risks.

[0028] Furthermore, the specific configuration of the mechanical transmission network establishment module 10 also includes: building a hierarchical structure according to the functional hierarchy, taking each device in the functional hierarchy as a node; generating a device structure mapping relationship based on the direct transmission relationship and logical dependency relationship, and establishing node connection edges; and establishing the multi-layer mechanical transmission network based on the nodes, node connection edges, and edge attributes, using the key metric index as edge attributes.

[0029] Preferably, a hierarchical structure is built according to the functional hierarchy of main drive equipment, linkage equipment, and auxiliary equipment, with the main drive equipment at the highest level and the linkage equipment and auxiliary equipment at the lowest level. Livestock machinery equipment is divided into three types of nodes according to functional hierarchy: main drive nodes, linkage nodes, and auxiliary nodes. Then, based on direct transmission relationships and logical dependencies, a mapping relationship of the equipment structure is generated, and node connection edges are established. This includes using physical power transmission such as gear meshing and belt drives as direct transmission edges, represented by solid lines, and using functional or control dependencies such as sensor → PLC → actuator as logical dependency edges, represented by dashed lines. The process involves: first, assigning values ​​to key metrics as edge attributes, including criticality (the weight of the impact of a fault on the system, e.g., 0.9); transmission dependence (the degree to which a child node depends on its parent node, 1.0 for complete dependence); and fault propagation probability (the probability that a parent node's failure will lead to a child node's failure, e.g., 0.8), to quantify fault risk and guide edge computing resource allocation; and finally, integrating nodes, node connections, and edge attributes through a graph theory model to establish a multi-layered mechanical transmission network, visually demonstrating the fault propagation chain and ensuring that edge nodes prioritize processing highly critical mechanical equipment monitoring data.

[0030] The computing power configuration module 20 is used to configure computing power based on the criticality, transmission dependence, and fault propagation impact of each node device in the multi-layer mechanical transmission network, deploy edge nodes, and build a cloud-edge platform structure.

[0031] Preferably, computing power is configured based on the criticality, transmission dependence, and fault propagation impact of each node device in the multi-layer mechanical transmission network. This involves dynamically allocating edge computing resources. Specifically, for highly critical nodes such as main motors and reducers, computing power priority is highest, and dedicated edge computing nodes, such as industrial-grade edge servers, are deployed to support real-time high-frequency data acquisition and rapid fault diagnosis. For example, a main motor node needs to process 1000 sampling points per second, allocating 2 CPU cores and 4GB of memory. For moderately critical nodes such as drive shafts and auxiliary pumps, computing power priority is moderate, and shared edge nodes, such as embedded gateways, are configured to perform periodic monitoring and simple threshold alarms. For low-criticality nodes such as environmental sensors, computing power priority is lowest, with lightweight processing handled by the cloud and the edge serving only as a data cache. Devices with strong transmission dependence are deployed centrally; for example, if device A's dependence on device B is >0.8, their edge nodes are deployed in the same location to reduce communication latency. Devices with weak transmission dependence are processed in a distributed manner; logically dependent devices connect to the edge gateway via low-power wireless protocols, reducing cabling costs. For nodes with high fault propagation impact, dual edge nodes are deployed to avoid single points of failure, and complete historical data of critical equipment are stored synchronously at the edge and in the cloud. For nodes with low propagation impact, a low-cost single-node architecture is adopted, relying on cloud-based disaster recovery. Then, edge nodes are deployed, including primary, secondary, and tertiary edge nodes, covering the main drive equipment, linkage equipment, and auxiliary equipment, respectively. Finally, a cloud-edge platform structure is built based on the multi-layer mechanical transmission network. The cloud platform is used for long-term trend analysis and cross-device fault correlation analysis. Primary and secondary edge nodes serve as the real-time layer to perform equipment status detection and local fault determination, while tertiary edge nodes serve as the preprocessing layer to filter and compress monitoring data to reduce uplink bandwidth consumption. An elastic resource pool is used to dynamically adjust the computing power of edge nodes according to network load, ensuring the utilization rate of edge computing resources, the real-time performance of fault monitoring, and reliability, thereby achieving precise and intelligent operation and maintenance of complex livestock machinery equipment.

[0032] The monitoring execution link acquisition module 30 is used to connect the mechanical equipment sensing network according to the causal relationship between the mechanical equipment sensing data and the fault event, and obtain the monitoring execution link between the mechanical equipment sensing network and the multi-layer mechanical transmission network.

[0033] Preferably, a sensing network is used to acquire real-time sensing data on the operating status of the mechanical equipment. This sensing network consists of multiple sensors, controllers, and data acquisition components of different types installed on the livestock machinery. These may include physical state sensors, such as accelerometers and infrared thermometers, to sense data such as vibration, temperature, noise, and current; performance parameter sensors, such as encoders and pressure transmitters, to sense pressure, flow rate, speed, and torque; and control signals such as PLC commands, relay status, and frequency converter outputs. Through analysis of the livestock machinery and its mechanisms, causal relationships are derived based on the mechanical transmission principle, establishing a causal relationship of abnormal sensing data → equipment status degradation → fault event occurrence. Examples of direct causal relationships include excessive bearing vibration → bearing wear → drive shaft seizure; and indirect causal relationships include decreased lubrication pump pressure → increased reducer temperature → gear pitting.

[0034] Furthermore, the specific configuration of the monitoring execution link acquisition module 30 also includes: determining the radar map of the sensing device based on the sensing data attributes, sensing data characteristics, and sensing range of the sensing device; analyzing the influencing factors and causal relationships of the failure events based on historical sample records of livestock machinery failure events, performing influencing factor and causal relationship analysis with the failure events as the center, and establishing a failure event heat map; and using the failure event heat map and the radar map of the sensing device to perform overlay mapping to obtain the mapping relationship between the sensing device and the failure events, and establishing the machinery equipment sensing network.

[0035] Preferably, the sensing data of the mechanical equipment is analyzed to obtain the sensing data attributes, sensing data characteristics, and sensing range of the sensing equipment. Sensing data attributes refer to the data types collected by the sensors, such as vibration, temperature, and current, to determine the type of physical phenomenon being identified. Sensing data characteristics include sampling rate, accuracy, and range, to determine the quality of the sensing data. The sensing range represents the equipment components or spatial area that the sensor can cover. Then, a radar chart of the sensing equipment is established to comprehensively evaluate the monitoring capabilities of a single sensing device. Historical sample records of livestock machinery failure events are obtained from all historical operating status data of the livestock machinery, and the influencing factors and causal relationships of these failure events are analyzed. Influencing factors are the direct causes of the failure events, and causal relationships are the propagation paths of the failures. Then, the influencing factors and causal relationships are analyzed centered on the failure events, i.e., the influencing factors and causal relationships are analyzed, the frequency of each factor occurring in historical failures is statistically analyzed, the reliability of the causal relationship is represented by probability, and the correlation strength is represented by color intensity, thereby constructing a heat map of failure events.

[0036] Preferably, the fault event heatmap and the radar map of the sensing device are overlaid and mapped. The sensor capabilities and fault requirements are matched according to the preset mapping rules. The preset mapping rules include attribute matching, feature matching and range matching. Attribute matching means that the physical quantities to be monitored in the fault heatmap must be included in the data types of the sensor radar map. Feature matching means that the sensor sampling rate / accuracy must meet the fault detection requirements. Range matching means that the sensor must cover the location where the fault occurs. Finally, the mapping relationship between the sensing device and the fault event is obtained as the overlay mapping result. Based on the overlay mapping result, a mechanical equipment sensing network is established, including a sensing layer, i.e., physical sensor nodes, an association layer, i.e., sensor-fault binding relationship, and an optimization layer for identifying areas with high heatmap values ​​but low radar map values ​​for supplementary monitoring. This combines data-driven fault analysis with the quantitative integration of equipment sensing capabilities.

[0037] Furthermore, the specific configuration of the monitoring execution link acquisition module 30 also includes: assigning influence weights to different application factors in the fault event heatmap to obtain a multidimensional causal intensity vector; assigning a sensing intensity and sensitivity index to each sensing data in the radar chart of the sensing device to obtain a sensing capability vector; performing similarity matching between the multidimensional causal intensity vector and the sensing capability vector, filtering associated sensing devices based on the matching degree to obtain the mapping relationship between the sensing devices and the fault events; and constructing a matching map of the sensing devices and fault events based on the mapping relationship between the sensing devices and the fault events to obtain the mechanical equipment sensing network, which is used to form the causal path connection relationship in the mechanical equipment sensing network.

[0038] Preferably, the influencing factors and causal relationships of the failure event are transformed into quantifiable mathematical vectors to characterize the triggering conditions and propagation intensity of the failure. Specifically, different application factors in the failure event heatmap are assigned influence weights, that is, the contribution of each factor to the failure is analyzed based on historical data. The influencing factors of bearing wear failure may include insufficient lubrication weight of 0.6, overload weight of 0.3, and foreign object intrusion weight of 0.1. The influence weights are then combined with the causal path to form a multidimensional causal intensity vector. For example, the causal intensity vector of insufficient lubrication → temperature rise → bearing wear is represented as [0.6, 0.8, 0.9], which correspond to the weight of insufficient lubrication, the probability of insufficient lubrication leading to temperature rise, and the probability of temperature rise leading to bearing wear, respectively. Through vectorization, the severity and probability of different failure paths can be clearly quantified.

[0039] Preferably, each sensing data point in the radar chart of the sensing device is assigned a sensing intensity and a sensitivity index. This means that the sensor's monitoring capability is converted into a numerical vector to match the fault requirements. The sensing intensity represents the sensor's ability to detect a certain type of physical quantity. For example, the sensing intensity of a vibration sensor may include acceleration detection capability 0.9, frequency range coverage 0.8, and noise immunity 0.7. The sensitivity index represents the sensor's ability to capture minute anomalies. For example, the sensitivity of a temperature sensor may include resolution ±0.1℃, response time <1 second, and environmental adaptability 0.8. The final combination forms a sensing capability vector, which represents the sensor's capabilities and assesses its suitability for monitoring specific faults.

[0040] Preferably, the multidimensional causal intensity vector and the sensing capability vector are similarly matched. That is, cosine similarity or Euclidean distance is used to calculate the similarity between the fault causal vector and the sensor capability vector, quantify the matching degree of the two vectors and select the most matching sensing device. A high matching degree indicates that the sensor can effectively cover the monitoring needs of the fault, while a low matching degree indicates that the sensor capability is insufficient and the device needs to be supplemented or replaced. Then, based on the matching degree, related sensing devices are selected to obtain the mapping relationship between sensing devices and fault events. Then, based on the mapping relationship between sensing devices and fault events, a matching graph of sensing devices and fault events is constructed. Specifically, the matching results are visualized to form a networked graph, clarifying the connection relationship between sensing devices and fault events. The graph nodes include sensing device nodes and fault event nodes, and the graph edges represent the connection between sensors and fault events and are labeled with the matching degree. When a new fault mode is added, the matching degree is recalculated and the matching graph is updated. When the sensor performance is upgraded, its capability vector is adjusted and rebound. Finally, through the matching graph, a complete mechanical equipment sensing network is constructed to form the causal path connection relationship in the mechanical equipment sensing network, thereby determining the monitoring logic of livestock machinery equipment faults and ensuring accurate monitoring of each fault event.

[0041] Furthermore, the specific configuration of the monitoring execution link acquisition module 30 also includes using graph neural networks and causal graph reasoning to establish a causal chain between changes in the characteristics of the sensing signal and the type of equipment fault. With the sensing device as the input node and the fault event as the target node, a directed weighted graph is constructed, wherein the edge weight represents the intensity of causal influence, thereby obtaining the mechanical equipment sensing network.

[0042] Preferably, a causal chain between changes in perceived signal characteristics and equipment fault types is established through graph neural networks and causal graph reasoning. The complex relationship between sensor signals and equipment faults is then modeled as a directed weighted graph, where the input nodes are the sensing equipment and its real-time signal characteristics, the target nodes are possible fault events, the directed edges represent the causal impact path of signal characteristic changes on the fault, and the edge weights represent the quantified causal impact strength. Specifically, features are extracted from the original sensor data to obtain fault-related key indicators, i.e., perceived signal characteristics, which may include vibration signals, temperature signals, and current signals. A time-series causal algorithm is used to extract the potential causal relationship between signals and faults from historical data. The causal chain is determined by mapping and matching the equipment fault type based on changes in perceived signal characteristics. Finally, a mechanical equipment sensing network is established by combining graph neural networks and causal chains to achieve closed-loop intelligent monitoring of perception-reasoning-decision, ensuring accurate location of fault sources in livestock machinery and equipment, and predicting the fault occurrence time based on the strength of the causal chain.

[0043] Preferably, a mechanical equipment sensing network is connected to establish a closed-loop analysis path from real-time sensing data to transmission fault diagnosis. Combining the physical relationships of the mechanical transmission network, the root cause of the fault is accurately located. Specifically, sensors in the mechanical equipment sensing network are associated with specific equipment nodes in the multi-layer mechanical transmission network, and the analysis dimensions corresponding to the sensor data are determined. Then, based on the topological relationship between the mechanical equipment sensing network and the multi-layer mechanical transmission network, a fault diagnosis logic chain, i.e., a monitoring execution link, is automatically generated. This includes physical layer sensors directly connecting to the physical connection edges of the transmission network, and functional layer sensors connecting to logically dependent edges. Then, the causal reasoning results in the mechanical equipment sensing network are superimposed on the fault propagation layer of the multi-layer mechanical transmission network to form a joint analysis path. The fault propagation probability of the transmission network edges is updated based on real-time sensing data. When a new sensor is added, the monitoring execution link is automatically expanded. For example, after adding an oil sensor, a lubrication status judgment branch is added. If the conclusions of multiple sensors are contradictory, such as normal vibration but high temperature, the data of high-critical nodes in the multi-layer transmission network are given priority, thereby ensuring the accuracy and real-time nature of fault location and reducing the false alarm rate of livestock machinery equipment monitoring faults.

[0044] The fault and anomaly information identification module 40 is used to dynamically allocate edge tasks based on the data change characteristics collected by the mechanical equipment sensing network and the computing power matching of the monitoring execution link. Based on the processing results of the edge tasks through the cloud-edge platform structure, fault and anomaly information is identified according to the multi-layer mechanical transmission network.

[0045] Furthermore, the specific configuration of the fault and anomaly information identification module 40 also includes: obtaining data feature changes through real-time data collection from the mechanical equipment sensing network; for sensing nodes with high causal weight in the mapping relationship between the sensing devices and fault events, prioritizing the triggering of corresponding edge nodes for local analysis; expanding edge tasks according to the influence range of the transmission path involved in the multi-layer mechanical transmission network, matching computing power with task-activated edge nodes, and allocating tasks according to the monitoring execution link when the task load of the edge node meets the allocation conditions.

[0046] Preferably, through the collaborative analysis of the mechanical equipment sensing network and the multi-layer mechanical transmission network, intelligent task scheduling driven by data features, prioritizing key nodes, and dynamically matching computing power is achieved. Specifically, real-time feature calculations are performed on the real-time collected data of the mechanical equipment sensing network to obtain time-domain features such as mean, peak value, and fluctuation amplitude, such as a sudden increase of 20% in vibration amplitude; frequency-domain features such as dominant frequency components and harmonic energy, such as an increase in the proportion of frequency energy characteristic of gear faults; and trend features such as slope and cumulative amount, such as a 5°C increase in bearing temperature per hour. Then, according to the predefined causal weights of sensing devices and fault events, the calculation priority is dynamically allocated, and the corresponding edge nodes are triggered to perform local analysis first. That is, nodes with high causal weights immediately trigger full analysis of edge nodes, while nodes with low causal weights are processed at lower frequencies or their analysis is postponed. For example, when both abnormal vibration and slight temperature rise are detected at the same time, vibration data is processed first.

[0047] Preferably, based on a multi-layer mechanical transmission network, the scope of equipment potentially affected by anomalies is predicted, i.e., the impact range of the transmission path is assessed. Primary impacts are directly connected equipment, and secondary impacts are indirectly related equipment. For example, abnormal motor vibration requires simultaneous inspection of the coupling alignment and reducer gear meshing. Then, edge tasks are expanded and matched with the task-activated edge nodes for computing power matching. When the task load of an edge node meets the allocation conditions, tasks are allocated according to the monitoring execution link. Specifically, for urgent diagnostic tasks, computing resources are preemptively allocated, such as immediately interrupting low-priority tasks and using the nearest idle edge node for computation; for routine monitoring, computing resources are allocated in a round-robin fashion, with edge nodes with lighter loads performing computation; for predictive analysis, computing resources are allocated in batches, with computation performed in the cloud or by clusters of idle edge nodes. Simultaneously, a load balancing mechanism is used to monitor the CPU / memory utilization of each edge node in real time. When the load exceeds a threshold, horizontal expansion is performed, such as migrating some tasks to adjacent nodes, or vertical degradation is performed, i.e., reducing the sampling rate of non-critical tasks.

[0048] Furthermore, the specific configuration of the fault and anomaly information identification module 40 also includes: when the task load of the edge node exceeds the allocation conditions, performing computing power collaborative matching in the cloud-edge platform structure based on the data feature changes to obtain collaborative nodes; and allocating tasks based on the task load status of the collaborative nodes.

[0049] Preferably, when the task load of an edge node exceeds the allocation conditions—that is, due to task overload, such as high CPU utilization or memory usage, or long task queue backlog, preventing timely processing of critical data—computing power collaborative matching is performed within the cloud-edge platform structure based on changes in data characteristics. This involves selecting collaborative nodes through a three-layer matching mechanism: prioritizing nearby edge nodes (selecting adjacent edge nodes within the same local area network with a load of <60%); supplementing computing in the cloud (activating cloud virtualization resources when all edge nodes are overloaded); and a hybrid computing mode (edge ​​nodes handle real-time feature extraction). Finally, tasks are allocated based on the task load status of the collaborative nodes. Specifically, for high real-time tasks, such as rapid vibration anomaly diagnosis, nearby edge nodes are prioritized; for computationally intensive tasks, such as bearing remaining life prediction, cloud GPU clusters are used; and for continuous monitoring tasks, such as temperature trend analysis, load is evenly distributed across multiple edge nodes. Furthermore, the allocation strategy is adjusted in real-time based on network conditions, including automatically reducing the cloud allocation ratio when network bandwidth decreases and immediately reclaiming tasks when nodes regain idleness.

[0050] Furthermore, the specific configuration of the fault and anomaly information identification module 40 also includes: based on the cloud-edge platform structure, analyzing the remaining computing power of each edge node and the computational transmission relationship with the target task edge node; according to the data feature changes and key metric response requirements of the fault event, optimizing the solution with the remaining computing power of each edge node and the computational transmission relationship with the target task edge node as the goal, and obtaining the collaborative node, which is an edge node or a cloud platform.

[0051] Preferably, the remaining computing power of each edge node is analyzed based on the cloud-edge platform structure. Specifically, the available computing power of the edge node CPU is 1 - (current utilization rate / number of cores), and a value >30% can accept new tasks; the available memory space is the total memory - the memory already occupied, and a value >500MB can process vibration data; the task queue depth is the number of tasks to be processed × the average processing time, and a delay <5ms is considered idle; at the same time, the operation and transmission relationship with the edge node of the target task is determined, that is, a topology graph between nodes is constructed, the real-time delay of the data transmission path is recorded, and a dynamic capability vector is generated for each node. The key metric for fault events, response requirement, refers to the urgency of the fault event. Then, based on changes in data characteristics, the response requirement of the key metric, the remaining computing power of each edge node, and the computational transmission relationship between the target task's edge nodes, the computational requirements of the task are quantified. High real-time requirements, such as vibration and shock detection, require a response time of <50ms; complex analysis requirements, such as fault prediction, require at least two GPU cores. Next, optimization is performed with the goal of maximizing response speed. This is achieved through multi-objective optimization algorithms, such as constrained particle swarm optimization, to select collaborative nodes from candidate nodes, including optimal edge nodes, prioritizing neighboring nodes with a transmission latency of <15ms and remaining computing power >40%, or cloud platforms. When edge resources are insufficient, cloud virtualization resources are activated and data transmission volume is automatically compressed. This ensures real-time processing of critical faults and enables elastic utilization of the entire cloud-edge platform resources.

[0052] Furthermore, the specific configuration of the fault and anomaly information identification module 40 also includes: based on the processing results of matching edge nodes or cloud platforms, integrating the correlation features of fault events, propagating them in the multi-layer mechanical transmission network to obtain fault propagation information; determining the mechanical fault level according to the key metrics of the propagating mechanical equipment; and generating structured fault and anomaly information, including fault location, fault type, fault level, occurrence time, and response measures, based on the fault propagation information and its mechanical fault level.

[0053] Preferably, based on the topology of a multi-layer mechanical transmission network, the initial fault signal is propagated bidirectionally along the physical transmission layer and the functional dependency layer, dynamically simulating the impact range. Affected nodes are marked using a graph traversal algorithm, and the propagation intensity is calculated using causal weights. For example, a motor bearing fault leads to a reducer input shaft, which then propagates to the agitator blades. The reducer input shaft represents a first-level propagation with an intensity of 0.8, while the agitator blades represent a second-level propagation with an intensity of 0.5. Then, the fault level is quantified and determined based on the key metrics and propagation depth of the propagating mechanical equipment, including Level I (urgent), affecting the main drive equipment and propagating through more than three layers, such as a motor fault causing a complete line shutdown; Level II (important), affecting local functional units, such as a lubrication pump failure affecting only a single gearbox; and Level III (general), auxiliary equipment malfunctions with no immediate risk of propagation, such as a decrease in cooling fan speed. Finally, based on the fault propagation information and its mechanical fault level, structured fault anomaly information is generated, including fault location, fault type, fault level, occurrence time, and response measures. For example, in the case of a gearbox input shaft bearing failure, the fault type is rolling bearing outer ring peeling, the fault level is Level II (important), affecting the gearbox and coupling, the occurrence time is 2025-03-20 14:25:36, and the response measures are to reduce the motor speed to 70% of the rated speed or replace the bearing and check the lubrication pipeline within 72 hours, thereby achieving closed-loop management from anomaly detection to decision support.

[0054] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0055] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A livestock machinery fault monitoring system based on edge computing, characterized in that, include: The mechanical transmission network establishment module is used to analyze the transmission relationship and working coupling characteristics between livestock machinery and equipment, and to establish a multi-layer mechanical transmission network. The computing power configuration module is used to configure computing power based on the criticality, transmission dependence, and fault propagation impact of each node device in the multi-layer mechanical transmission network, deploy edge nodes, and build a cloud-edge platform structure. The monitoring execution link acquisition module is used to connect the mechanical equipment sensing network according to the causal relationship between the mechanical equipment sensing data and the fault event, and obtain the monitoring execution link between the mechanical equipment sensing network and the multi-layer mechanical transmission network. The fault and anomaly information identification module is used to dynamically allocate edge tasks based on the changing characteristics of the data collected by the mechanical equipment sensing network and the computing power matching of the monitoring execution link. Based on the processing results of the edge tasks through the cloud-edge platform structure, fault and anomaly information is identified according to the multi-layer mechanical transmission network.

2. The livestock machinery fault monitoring system based on edge computing according to claim 1, characterized in that, The steps performed by the mechanical transmission network establishment module include: Based on the transmission relationship and working coupling characteristics, the functional levels of the main drive equipment, linkage equipment, and auxiliary equipment are identified. Based on the functional hierarchy, analyze the direct transmission relationships and logical dependencies between the devices; Based on the direct transmission relationships and logical dependencies between the devices, calculate the key metrics for each device in the transmission network; Based on the aforementioned functional hierarchy, direct transmission relationship, logical dependency relationship, and equipment correlation of key metrics, the multi-layer mechanical transmission network is established.

3. The livestock machinery fault monitoring system based on edge computing according to claim 2, characterized in that, The steps performed by the mechanical transmission network establishment module include: Build a hierarchical structure according to the functional hierarchy, and treat each device in the functional hierarchy as a node; Based on the direct transmission relationship and logical dependency relationship, generate the equipment structure mapping relationship and establish node connection edges; Based on the key metrics as edge attributes, the multi-layer mechanical transmission network is established based on the nodes, node connection edges, and edge attributes.

4. The livestock machinery fault monitoring system based on edge computing according to claim 1, characterized in that, The monitoring execution link acquisition module further includes a mechanical equipment sensing network establishment module, the steps of which are: The radar chart of the sensing device is determined based on the sensing data attributes, sensing data characteristics, and sensing range of the sensing device. Based on historical sample records of livestock machinery failure events, the influencing factors and causal relationships of the failure events are analyzed. Taking the failure events as the center, the influencing factors and causal relationships are analyzed, and a heat map of failure events is established. By overlaying and mapping the heat map of the fault event with the radar map of the sensing device, the mapping relationship between the sensing device and the fault event is obtained, and the mechanical equipment sensing network is established.

5. The livestock machinery fault monitoring system based on edge computing according to claim 4, characterized in that, The steps performed by the mechanical equipment sensing network establishment module also include: By assigning influence weights to different application factors in the heatmap of the fault event, a multidimensional causal intensity vector is obtained. Each sensing data point in the radar chart of the sensing device is assigned a sensing intensity and sensitivity index to obtain a sensing capability vector. The multidimensional causal strength vector and the perception capability vector are similarly matched, and associated perception devices are filtered based on the matching degree to obtain the mapping relationship between the perception devices and the fault events. Based on the mapping relationship between the sensing devices and fault events, a matching map between the sensing devices and fault events is constructed to obtain the mechanical equipment sensing network, which is used to form causal path connection relationships in the mechanical equipment sensing network.

6. The livestock machinery fault monitoring system based on edge computing according to claim 5, characterized in that, The steps performed by the mechanical equipment sensing network establishment module also include: By using graph neural networks and causal graph reasoning, a causal chain between changes in the characteristics of sensing signals and equipment fault types is established. With sensing devices as input nodes and fault events as target nodes, a directed weighted graph is constructed, where the edge weights represent the intensity of causal influence, thus obtaining the mechanical equipment sensing network.

7. The livestock machinery fault monitoring system based on edge computing according to claim 5, characterized in that, The steps performed by the fault and anomaly information identification module include: By collecting real-time data through the sensing network of the mechanical equipment, changes in data characteristics are obtained; For sensing nodes with high causal weight in the mapping relationship between the sensing devices and fault events, the corresponding edge nodes are preferentially triggered for local analysis. Based on the influence range of the transmission path involved in the multi-layer mechanical transmission network, edge tasks are extended and matched with the computing power of the task-activated edge nodes. When the task load of the edge node meets the allocation conditions, the task is allocated according to the monitoring and execution link.

8. The livestock machinery fault monitoring system based on edge computing according to claim 7, characterized in that, The steps performed by the fault and anomaly information identification module also include: When the task load of the edge node exceeds the allocation conditions, computing power collaborative matching is performed in the cloud-edge platform structure based on the data feature changes to obtain collaborative nodes. Task allocation is performed based on the task load status of the collaborative nodes.

9. The livestock machinery fault monitoring system based on edge computing according to claim 8, characterized in that, The steps performed by the fault and anomaly information identification module include: Based on the processing results of matching edge nodes or cloud platforms, the correlation characteristics of fault events are integrated and propagated in the multi-layer mechanical transmission network to obtain fault propagation information. The mechanical failure level is determined based on the key metrics of the transmission equipment. Based on the fault propagation information and its mechanical fault level, structured fault anomaly information is generated, including fault location, fault type, fault level, occurrence time, and response measures.

10. The livestock machinery fault monitoring system based on edge computing according to claim 8, characterized in that, The steps performed by the fault and anomaly information identification module include: Based on the cloud-edge platform structure, the remaining computing power of each edge node and the computational transmission relationship with the target task edge node are analyzed. Based on the key metrics of the data feature changes and fault events, and considering the remaining computing power of each edge node and the computational transmission relationship with the target task edge node, the solution is optimized with the goal of maximizing the response speed to obtain the collaborative node, which is either an edge node or a cloud platform.

Images