Device predictive maintenance system and method based on large model and agent cooperation

Abstract

The application belongs to the field of equipment operation and maintenance, and relates to a data analysis technology, and is used for solving the problems that the prior art cannot distinguish the necessity of data uploading and dynamically optimize the data uploading basis according to the prediction result, and specifically is an equipment predictive maintenance system and method based on cooperation of a large model and an intelligent agent, comprising successively connected edge monitoring modules, pre-diagnosis analysis modules and uploading supervision modules, the edge monitoring modules, the pre-diagnosis analysis modules and the uploading supervision modules are all in communication connection with a database; through the application, preliminary data screening and analysis are realized at the edge end, the data amount transmitted to the data center is greatly reduced, and the computing pressure of the data center is effectively relieved; at the same time, abnormal parameters are analyzed in time through the pre-diagnosis analysis modules, and the risk that potential fault hidden dangers are ignored is avoided; combined with historical data analysis and risk assessment, problems possibly occurring to equipment can be more accurately predicted.

Description

Technical Field

[0001] This invention belongs to the field of equipment operation and maintenance, and involves data analysis technology, specifically a predictive maintenance system and method for equipment based on large models and intelligent agents. Background Technology

[0002] Predictive maintenance systems are an advanced maintenance strategy that differs from traditional "reactive maintenance" or fixed "periodic maintenance." Instead, they use data analysis and artificial intelligence to predict potential equipment failures in advance, enabling precise maintenance to be performed before failures occur and at the most appropriate time.

[0003] Chinese patent application CN119273326B discloses a method and system for predicting the maintenance time of omnidirectional beacon equipment. This method determines a feature vector matrix based on historical operating data and historical climate data for a preset time period. The feature vector matrix is ​​then input into an omnidirectional beacon equipment maintenance time prediction model to predict the maintenance time of the omnidirectional beacon equipment. However, this method requires uploading all data to a data center for fusion analysis, resulting in excessive data upload and analysis volume. This can easily overload the data transmission and processing ends, leading to system malfunctions. Therefore, edge processing of some data to avoid uploading can alleviate the computational pressure on the data center. However, how to determine the necessity of data upload and dynamically optimize the data upload criteria based on the prediction results to balance data processing efficiency and prediction timeliness is a technical problem that urgently needs to be solved in this field.

[0004] To address the aforementioned technical problems, this application proposes a solution. Summary of the Invention

[0005] The purpose of this invention is to provide a predictive maintenance system and method for equipment based on large model and intelligent agent collaboration, which solves the problem that the existing technology cannot distinguish the necessity of data uploading and dynamically optimize the basis for data uploading based on prediction results; The technical problem to be solved by this invention is: how to provide a predictive maintenance system and method for equipment based on large model and intelligent agent collaboration, which can distinguish the necessity of data uploading and dynamically optimize the basis for data uploading based on prediction results.

[0006] The objective of this invention can be achieved through the following technical solutions: A predictive maintenance system for equipment based on large model and intelligent agent collaboration includes an edge monitoring module, a pre-diagnosis analysis module, and an upload monitoring module connected in sequence. The edge monitoring module, the pre-diagnosis analysis module, and the upload monitoring module are all connected to a database. The edge monitoring module is used to perform edge monitoring and analysis on various operating parameters of the device: edge agents are deployed on edge computing nodes near the device, a monitoring cycle is generated and divided into several monitoring periods, and the monitoring values ​​of the device's operating parameters are obtained at the end of each monitoring period. The operating parameters are marked as normal parameters, uploaded parameters, or fault parameters based on the monitoring values. The device's operating status during the monitoring period is determined based on the marking results of normal parameters, uploaded parameters, or fault parameters. The pre-diagnosis analysis module is used to perform pre-diagnosis analysis on potential equipment faults: the ratio of the monitored value of the uploaded parameter to the monitoring threshold is marked as the monitoring coefficient of the uploaded parameter, the monitoring coefficients of all uploaded parameters are input into the pre-diagnosis model, and the pre-diagnosis model receives the monitoring coefficients of the uploaded parameters and performs pre-diagnosis processing. The upload monitoring module is used to monitor and analyze the upload delay status of device operating parameters.

[0007] Furthermore, the specific process of marking operating parameters using monitored values ​​includes: the monitored value is the maximum value of the operating parameter within the monitoring period; the monitoring threshold and upload threshold of the operating parameter are retrieved from the database, where the upload threshold = t1 × monitoring threshold, and t1 is a proportionality coefficient, and 0.88 < t1 < 1; the monitored value is compared with the monitoring threshold and the upload threshold: if the monitored value is less than or equal to the upload threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to meet the requirements, and the corresponding operating parameter is marked as a normal parameter; if the monitored value is between the upload threshold and the monitoring threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to have potential abnormalities, and the corresponding operating parameter is marked as an upload parameter; if the monitored value is greater than or equal to the monitoring threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to not meet the requirements, and the corresponding operating parameter is marked as a fault parameter.

[0008] Furthermore, the specific process for determining whether the operating status of the equipment meets the requirements during the monitoring period includes: if all operating parameters are marked as normal parameters during the monitoring period, the operating status of the equipment during the monitoring period is determined to meet the requirements; if any operating parameter is marked as a fault parameter during the monitoring period, a fault diagnosis signal is generated and the fault diagnosis signal and the fault parameter are sent to the mobile terminal of the management personnel; otherwise, a pre-diagnosis signal is generated and sent to the pre-diagnosis analysis module.

[0009] Furthermore, the specific process of pre-diagnosis processing includes: retrieving historical operation and maintenance data and detection data of the equipment and screening pre-diagnosis analysis periods; marking the number of times the operating parameter is marked as a fault parameter in all pre-diagnosis analysis periods as the marked value of the operating parameter; marking the ratio of the marked value to the number of pre-diagnosis analysis periods as the risk coefficient of the operating parameter; and determining whether the operating parameter has an operational risk through the risk coefficient.

[0010] Furthermore, the screening process for the pre-diagnostic analysis period includes: arranging the uploaded parameters in descending order of monitoring coefficient values ​​to obtain the uploaded sequence; retrieving the historical data of the executed monitoring periods that are exactly the same as the uploaded sequence and marking them as matching periods; and marking the L1 monitoring periods after the matching periods as pre-diagnostic analysis periods.

[0011] Furthermore, the specific process for determining whether an operating parameter has an operational risk includes: obtaining a risk threshold from a database, comparing the risk coefficient with the risk threshold; if the risk coefficient is less than the risk threshold, the operating parameter is determined not to have an abnormal risk; if the risk coefficient is greater than or equal to the risk threshold, the operating parameter is determined to have an operational risk, the corresponding operating parameter is marked as a risk parameter, a risk handling signal is generated, and the risk handling signal and the risk parameter are sent to the mobile terminal of the management personnel.

[0012] Furthermore, the specific process of the upload monitoring module to monitor and analyze the upload delay status of equipment operating parameters includes: at the end of the monitoring period, retrieving all marked fault parameters, and retrieving the marking status of the corresponding operating parameters in the previous monitoring period of the marked fault parameters: if the corresponding operating parameter was marked as a normal parameter in the previous monitoring period, then the corresponding fault parameter is marked as a delayed parameter; otherwise, it is not marked; the ratio of the number of times the same operating parameter is marked as a delayed parameter to the number of times it is marked as a fault parameter in the monitoring period is marked as the delay coefficient, and the timeliness of the upload of operating parameters in the monitoring period is judged by the delay coefficient.

[0013] Furthermore, the specific process for determining whether the timeliness of uploading operating parameters within the monitoring period meets the requirements includes: obtaining the delay threshold from the database, comparing the delay coefficient of the operating parameters with the delay threshold; if the delay coefficient is less than the delay threshold, it is determined that the timeliness of uploading operating parameters within the monitoring period meets the requirements; if the delay coefficient is greater than or equal to the delay threshold, it is determined that the timeliness of uploading operating parameters within the monitoring period does not meet the requirements, and the upload threshold of the operating parameters is adjusted downward proportionally.

[0014] A predictive maintenance method for equipment based on large models and agent collaboration includes the following steps: Step 1: Perform edge monitoring analysis on various operating parameters of the equipment: Generate a monitoring period and divide the monitoring period into several monitoring periods. At the end of each monitoring period, obtain the monitoring values ​​of the equipment operating parameters, evaluate the equipment operating status based on the monitoring values, and filter and upload the parameters. Step 2: Conduct pre-diagnosis analysis of potential equipment malfunctions: Mark the ratio of the monitored value of the uploaded parameter to the monitoring threshold as the monitoring coefficient of the uploaded parameter, and input the monitoring coefficients of all uploaded parameters into the pre-diagnosis model for pre-diagnosis processing; Step 3: Monitor and analyze the upload delay status of equipment operating parameters: At the end of the monitoring period, obtain the delay coefficient of the operating parameters, and use the delay coefficient to determine whether the timeliness of the upload of operating parameters within the monitoring period meets the requirements.

[0015] The present invention has the following beneficial effects: 1. This application enables preliminary data screening and analysis at the edge, significantly reducing the amount of data transmitted to the data center and effectively alleviating the computing pressure on the data center. Simultaneously, the pre-diagnosis analysis module analyzes abnormal parameters in a timely manner, preventing potential faults from being overlooked. The introduction of the upload monitoring module ensures the timely transmission of critical data, improving the system's real-time performance. This three-tiered collaborative mechanism of edge processing, pre-diagnosis, and upload monitoring reduces cloud load and ensures timely processing of abnormal data, significantly improving the efficiency and accuracy of predictive maintenance. Therefore, this application effectively solves the risk of predictive failure caused by data processing delays in traditional systems, reduces the probability of sudden equipment failures, reduces the consumption of network resources by redundant data transmission, and improves the overall operating efficiency and reliability of the system. 2. This application can perform risk assessment on equipment operating parameters based on historical data and identify potential failure risks in advance; by calculating risk coefficients, the risk level of different operating parameters can be quantified, providing data support for equipment maintenance decisions; this method combines historical data analysis and risk assessment, which can more accurately predict potential equipment problems, thereby achieving preventive maintenance, reducing the probability of unexpected equipment downtime, and improving equipment reliability and service life. 3. This application can filter monitoring periods with similar characteristics from historical data based on the distribution characteristics of the monitoring coefficients of the currently uploaded parameters, and determine the pre-diagnosis analysis period based on this. Therefore, the pre-diagnosis analysis can focus on historical data most similar to the current situation, improving the targeting and accuracy of the pre-diagnosis. By setting an appropriate L1 value, the time range of the pre-diagnosis analysis can be flexibly adjusted, ensuring both the comprehensiveness of the analysis and avoiding unnecessary waste of computational resources. 4. This application can effectively monitor the upload delay status of equipment operating parameters, promptly detect and handle problems of untimely uploads; by introducing the concepts of delay coefficient and delay threshold, the timeliness of parameter uploads can be quantitatively evaluated, and the upload threshold can be dynamically adjusted based on the evaluation results; this method can ensure the timely upload of important data while avoiding the transmission of too much irrelevant data, thereby improving the system's data processing efficiency while ensuring prediction accuracy; in addition, by dynamically adjusting the upload threshold, the system can adaptively optimize the data upload strategy, further improving the effectiveness and reliability of predictive maintenance of equipment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a system block diagram of Embodiment 1 of the present invention; Figure 2 This is a flowchart of the method in Embodiment 2 of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Traditional predictive maintenance systems rely on centralized data centers to process all monitoring data, leading to excessive loads on both the data transmission and processing ends. Since edge nodes lack data filtering and preliminary diagnostic capabilities, all operating parameters must be uploaded to the cloud in real-time for fusion analysis, causing a surge in network bandwidth usage and a significant increase in data processing latency. This makes it difficult to meet the real-time-sensitive equipment fault prediction needs of industrial scenarios. For example, in the vibration and temperature monitoring scenario of industrial centrifuges, edge computing nodes collect more than 200 sets of multi-dimensional sensor data per second. Uploading all parameters such as the overall vibration RMS value, bearing temperature, and winding temperature not only congests network transmission channels but also concentrates cloud computing resources on low-value data processing. The upload and diagnostic response time for critical abnormal parameters is squeezed out by unnecessary data, directly impacting the efficiency of early fault identification. If these problems are not addressed, the system will face the long-term risk of predictive failure due to data processing delays, the inability of abnormal parameters to trigger early warning mechanisms in a timely manner, an increased probability of sudden equipment failures, and the continuous consumption of network resources by redundant data transmission, reducing the overall system efficiency and reliability.

[0020] Faced with the aforementioned challenges, this application first considers deploying edge computing nodes near the device to alleviate data processing pressure. However, it needs to address how to filter critical data and dynamically adjust the upload strategy. Traditional solutions upload all data, leading to resource waste. This application recognizes the need to differentiate between normal, pending upload, and fault parameters at the edge, uploading only necessary data. Further analysis reveals that simply reducing the upload volume may overlook potential risks. Therefore, a pre-diagnosis module is introduced to perform preliminary analysis of abnormal parameters, avoiding the omission of critical information. Simultaneously, considering that upload delays may affect the timeliness of early warnings, an independent module is needed to monitor upload timeliness, forming a three-tiered collaborative mechanism of edge processing, pre-diagnosis, and upload monitoring. This reduces cloud load while ensuring timely processing of abnormal data.

[0021] Example 1: As Figure 1 As shown, the predictive maintenance system for equipment based on large model and intelligent agent collaboration includes an edge monitoring module, a pre-diagnosis analysis module, and an upload monitoring module connected in sequence. The edge monitoring module, the pre-diagnosis analysis module, and the upload monitoring module are all connected to the database.

[0022] The edge monitoring module is used to perform edge monitoring analysis on various operating parameters of the equipment. These parameters include unidirectional monitoring parameters such as overall vibration RMS value, bearing temperature, winding temperature, lubricating oil temperature, and filter inlet / outlet pressure difference. Edge agents are deployed on edge computing nodes near the equipment to generate a monitoring cycle, which is then divided into several monitoring periods. At the end of each monitoring period, the monitored values ​​of the equipment's operating parameters are obtained. These monitored values ​​represent the maximum values ​​of the operating parameters within the monitoring period. The monitoring threshold and upload threshold for the operating parameters are retrieved from the database. The upload threshold = t1 × monitoring threshold, where t1 is a proportionality coefficient, and 0.88 < t1 < 1. The monitored values ​​are compared with the monitoring threshold and upload threshold. If the monitored value is less than or equal to the upload threshold, it is determined that the operating parameter's value within the monitoring period is below the maximum value. If the value distribution meets the requirements, the corresponding operating parameter is marked as a normal parameter; if the monitored value is between the upload threshold and the monitoring threshold, it is determined that the value distribution of the operating parameter during the monitoring period has potential abnormalities, and the corresponding operating parameter is marked as an upload parameter; if the monitored value is greater than or equal to the monitoring threshold, it is determined that the value distribution of the operating parameter during the monitoring period does not meet the requirements, and the corresponding operating parameter is marked as a fault parameter; if all operating parameters are marked as normal parameters during the monitoring period, it is determined that the operating status of the equipment during the monitoring period meets the requirements; if any operating parameter is marked as a fault parameter during the monitoring period, a fault diagnosis signal is generated and the fault diagnosis signal and the fault parameter are sent to the mobile terminal of the management personnel; otherwise, a pre-diagnosis signal is generated and sent to the pre-diagnosis analysis module.

[0023] Specifically, at the end of the monitoring period, the system automatically extracts the maximum value of each operating parameter during that period as the monitoring value and compares it with the preset monitoring threshold and the dynamically calculated upload threshold. When the monitoring value exceeds the upload threshold but does not reach the monitoring threshold, the parameter is marked as an upload parameter, triggering the subsequent pre-diagnostic analysis process without immediately uploading all data. By setting the proportional coefficient t1, a dynamic balance can be achieved between data processing efficiency and the timeliness of fault early warning. For example, in lubricating oil temperature monitoring, if the monitoring threshold is 120℃ and t1 is 0.92, then the upload threshold is 110.4℃. When the monitoring value is 115℃, the system only uploads the necessary data to the pre-diagnostic module for analysis, which avoids data center overload and ensures that potential faults are identified in a timely manner.

[0024] The system determines whether the equipment's operating status meets requirements by checking if all marked operating parameters are within the normal range. Fault diagnosis signals are generated by identifying any operating parameter as a faulty parameter, and the fault diagnosis signal and the faulty parameter are transmitted to the mobile terminal via a wireless communication protocol. Pre-diagnosis signals are generated after excluding the first two conditions, and the pre-diagnosis signals are transmitted to the pre-diagnosis analysis module via an internal communication link. The generation logic for fault diagnosis signals and pre-diagnosis signals uses mutually exclusive condition judgments to ensure that only one processing method is executed within the same monitoring period. For example, if the monitored value of an operating parameter exceeds the monitoring threshold within a certain monitoring period, the parameter is marked as a faulty parameter, and the system immediately generates a fault diagnosis signal and sends it to the management terminal.

[0025] Specifically, at the end of the monitoring period, the system iterates through the marking results of all operating parameters. If all parameters are normal, the system determines that the equipment is operating normally and no further action is required. If at least one fault parameter exists, the system generates a fault diagnosis signal containing fault parameter information, pushes it to the management terminal via the mobile network, and triggers the emergency maintenance process. If neither all parameters are normal nor are there any fault parameters, the system generates a pre-diagnosis signal and transmits it to the pre-diagnosis analysis module, initiating the hidden danger analysis process. This judgment logic achieves differentiated responses under different scenarios through hierarchical condition judgment, ensuring that fault parameters are handled first, and potential anomalies are further evaluated through the pre-diagnosis analysis module. For example, when the monitored value of an operating parameter is between the upload threshold and the monitoring threshold during a certain monitoring period, the parameter is marked as an upload parameter, the system generates a pre-diagnosis signal, and initiates subsequent analysis to avoid overlooking potential risks.

[0026] As a preferred embodiment, the solution of this application is specifically implemented as follows: In the monitoring system of an industrial centrifuge unit, an edge monitoring module is installed on an edge computing device near the centrifuge unit. The module is set to monitor a 24-hour period, divided into 144 10-minute monitoring intervals. At the end of each monitoring interval, the module acquires monitoring values ​​for parameters such as overall vibration RMS, bearing temperature, winding temperature, lubricating oil temperature, and filter inlet / outlet pressure difference.

[0027] For example, for the bearing temperature parameter, the normal operating temperature threshold is set to 60℃. If the bearing temperature obtained at the end of a monitoring period is 55℃, which is less than the upload threshold of 58℃ (assuming the upload threshold is 97% of the normal upper limit), then the parameter is marked as a normal parameter. If the temperature is 59℃, which is between the upload threshold and the monitoring threshold (60℃), then it is marked as an uploaded parameter. If the temperature reaches or exceeds 60℃, then it is marked as a fault parameter.

[0028] The pre-diagnosis analysis module is used to perform pre-diagnosis analysis on potential equipment faults. It marks the ratio of the monitored value of an uploaded parameter to the monitoring threshold as the monitoring coefficient of the uploaded parameter. All monitoring coefficients of uploaded parameters are input into the pre-diagnosis model. The pre-diagnosis model receives the monitoring coefficients of the uploaded parameters and performs pre-diagnosis processing: The uploaded parameters are arranged in descending order of monitoring coefficient values ​​to obtain an uploaded sequence. Historical data is retrieved, and the executed monitoring periods that are identical to the uploaded sequence are marked as matching periods. The L1 monitoring periods following the matching periods are marked as pre-diagnosis analysis periods. The number of times an operating parameter is marked as a fault parameter within all pre-diagnosis analysis periods is marked as the marked value of the operating parameter. The ratio of the marked value to the number of pre-diagnosis analysis periods is marked as the risk coefficient of the operating parameter. A risk threshold is obtained from the database, and the risk coefficient is compared with the risk threshold: if the risk coefficient is less than the risk threshold, the operating parameter is determined to have no abnormal risk; if the risk coefficient is greater than or equal to the risk threshold, the operating parameter is determined to have operational risk, the corresponding operating parameter is marked as a risk parameter, a risk processing signal is generated, and the risk processing signal and the risk parameter are sent to the mobile terminal of the management personnel.

[0029] The historical operation and maintenance data and testing data include the labeling results of the equipment's operating parameters within the historical monitoring period. The pre-diagnosis analysis period is determined by matching the currently uploaded parameter sequence with the executed monitoring periods in the historical data. The label value reflects the frequency of failure of the operating parameter under similar historical operating conditions, and the risk coefficient quantifies the long-term risk probability of the operating parameter by the ratio of the label value to the number of pre-diagnosis analysis periods. The risk threshold serves as the judgment benchmark, and the operating risk status of the parameter is distinguished by comparing the risk coefficient and the risk threshold.

[0030] Specifically, during the pre-diagnosis analysis period selection process, uploaded parameters are arranged in descending order of monitoring coefficients to form an uploaded sequence. Executed monitoring periods in historical data that perfectly match the uploaded sequence are identified as matching periods. The L1 monitoring periods following the matching periods are extracted as pre-diagnosis analysis periods. The cumulative number of times an operating parameter is marked as a fault parameter within the pre-diagnosis analysis period is the marker value. The ratio of the marker value to the number of periods generates a risk coefficient. If the risk coefficient exceeds a risk threshold, the operating parameter is marked as a risk parameter, triggering a risk handling signal. This process assesses the risk probability of the current parameter by analyzing the fault frequency under similar operating conditions in historical data, and combines this with the risk threshold to quantify the abnormal risk, avoiding misjudgments caused by fluctuations in single monitoring data, while reducing the possibility of long-term hidden dangers going undetected.

[0031] The generation of the uploaded sequence is achieved by sorting the monitoring coefficient values ​​to ensure that parameters with high monitoring coefficients occupy a priority position in the sequence; the selection criteria for the matching time period are the monitoring time periods in historical data that are completely consistent with the current uploaded sequence, ensuring the similarity between the historical context and the current state; the scope of the L1 monitoring time periods is controlled by the system's preset parameters to form a fixed-length analysis window.

[0032] Specifically, during the screening process for the pre-diagnostic analysis period, all uploaded parameters within the current monitoring period are first sorted from highest to lowest according to their monitoring coefficients, forming an upload sequence that includes parameter type and arrangement order. The system traverses the historical database, selecting executed monitoring periods with identical upload sequences as matching periods. Subsequently, the L1 consecutive monitoring periods following each matching period are designated as pre-diagnostic analysis periods. For example, when L1 is set to 5, historical data from the five monitoring periods following each matching period will be included in the analysis. Screening historical periods through matching sequences ensures that the selected data is highly consistent with the current abnormal parameter combination pattern, avoiding the introduction of irrelevant historical data that could interfere with risk coefficient calculation. The designation of subsequent periods after the matching period captures the evolution trend of equipment status after the occurrence of abnormal parameter combinations, providing dynamic data support for risk assessment.

[0033] The upload monitoring module is used to monitor and analyze the upload delay status of equipment operating parameters. At the end of the monitoring period, it retrieves all marked fault parameters and the marking status of the corresponding operating parameters in the previous monitoring period. If the corresponding operating parameter was marked as a normal parameter in the previous monitoring period, it marks the corresponding fault parameter as a delayed parameter; otherwise, it does not mark it. The ratio of the number of times the same operating parameter is marked as a delayed parameter to the number of times it is marked as a fault parameter in the monitoring period is marked as the delay coefficient. The delay threshold is obtained from the database, and the delay coefficient of the operating parameter is compared with the delay threshold. If the delay coefficient is less than the delay threshold, it is determined that the upload timeliness of the operating parameter in the monitoring period meets the requirements; if the delay coefficient is greater than or equal to the delay threshold, it is determined that the upload timeliness of the operating parameter in the monitoring period does not meet the requirements, and the upload threshold of the operating parameter is adjusted downward proportionally.

[0034] The marking conditions for delayed parameters are based on whether the marked status of the fault parameter was a normal parameter in the previous monitoring period. If the parameter was a normal parameter in the previous period but suddenly becomes a fault parameter in the current period, it indicates that the parameter was not marked in time before reaching the fault threshold, resulting in an upload delay. The delay coefficient is calculated as the ratio of the number of delays to the number of faults, reflecting the probability that the same parameter was not identified in time before a fault occurred within the monitoring period. The delay threshold is preset through the database to quantify the tolerance for upload timeliness. The proportional reduction of the upload threshold makes the marking conditions for uploaded parameters more sensitive by reducing the ratio coefficient between the upload threshold and the monitoring threshold, thereby reducing the risk of delays in subsequent monitoring periods.

[0035] Specifically, at the end of the monitoring period, all marked fault parameters are retrieved and their marking status from the previous monitoring period is traced back one by one. If a fault parameter was still a normal parameter in the previous period, it indicates that the parameter suddenly changed from a normal state to a fault state in a short period of time, but was not marked as an upload parameter during this process, indicating a delay in timely upload, and is therefore marked as a delayed parameter. The ratio of the number of delays to the number of faults for the same operating parameter within the monitoring period is calculated as the delay coefficient, which reflects the frequency with which the parameter was not identified in time before the fault occurred. By comparing the delay coefficient with a preset delay threshold, if the delay coefficient exceeds the threshold, it is determined that the upload timeliness is insufficient. At this time, the upload threshold for the parameter is proportionally reduced, for example, the proportional coefficient t1 is adjusted from 0.9 to 0.85, making the marking conditions for uploaded parameters more stringent in subsequent monitoring periods, thereby prompting abnormal risks to be identified and uploaded earlier and reducing delays.

[0036] Example 2: Figure 2 As shown, the predictive maintenance method for equipment based on large model and agent collaboration includes the following steps: Step 1: Perform edge monitoring analysis on various operating parameters of the equipment: Generate a monitoring period and divide the monitoring period into several monitoring periods. At the end of each monitoring period, obtain the monitoring values ​​of the equipment operating parameters, evaluate the equipment operating status based on the monitoring values, and filter and upload the parameters. Step 2: Conduct pre-diagnosis analysis of potential equipment malfunctions: Mark the ratio of the monitored value of the uploaded parameter to the monitoring threshold as the monitoring coefficient of the uploaded parameter, and input the monitoring coefficients of all uploaded parameters into the pre-diagnosis model for pre-diagnosis processing; Step 3: Monitor and analyze the upload delay status of equipment operating parameters: At the end of the monitoring period, obtain the delay coefficient of the operating parameters, and use the delay coefficient to determine whether the timeliness of the upload of operating parameters within the monitoring period meets the requirements.

[0037] A predictive maintenance system and method for equipment based on large model and intelligent agent collaboration is proposed. During operation, the edge monitoring module is deployed on edge computing nodes near the equipment to monitor and analyze the equipment's operating parameters. Specifically, the edge monitoring module generates a monitoring cycle and divides it into several monitoring periods. At the end of each period, it acquires the parameter monitoring values ​​and marks the parameters as normal, uploaded, or faulty based on the monitoring values. These marking results determine whether the equipment's operating status meets the requirements. The pre-diagnosis analysis module receives the monitoring coefficients of the uploaded parameters and performs pre-diagnosis analysis on potential equipment faults. The upload monitoring module monitors and analyzes the upload delay status of the parameters.

[0038] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0039] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0040] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A predictive maintenance system for equipment based on large-scale model and agent collaboration, characterized in that, It includes an edge monitoring module, a pre-diagnosis analysis module, and an upload monitoring module connected in sequence, all of which are communicatively connected to a database. The edge monitoring module is used to perform edge monitoring analysis on various operating parameters of the device: the edge intelligence agent is deployed on the edge computing node near the device, a monitoring cycle is generated and the monitoring cycle is divided into several monitoring periods. At the end of the monitoring period, the monitoring value of the device operating parameters is obtained, and the operating parameters are marked as normal parameters, uploaded parameters or fault parameters based on the monitoring value. The system uses the marking results of normal parameters, uploaded parameters, or fault parameters to determine whether the operating status of the equipment meets the requirements during the monitoring period. The pre-diagnosis analysis module is used to perform pre-diagnosis analysis on potential equipment faults: the ratio of the monitored value of the uploaded parameter to the monitoring threshold is marked as the monitoring coefficient of the uploaded parameter, the monitoring coefficients of all uploaded parameters are input into the pre-diagnosis model, and the pre-diagnosis model receives the monitoring coefficients of the uploaded parameters and performs pre-diagnosis processing. The upload monitoring module is used to monitor and analyze the upload delay status of equipment operating parameters; The specific process of the upload monitoring module to monitor and analyze the upload delay status of equipment operating parameters includes: at the end of the monitoring period, retrieving all marked fault parameters, and retrieving the marking status of the corresponding operating parameters in the previous monitoring period of the marked fault parameters: if the corresponding operating parameter was marked as a normal parameter in the previous monitoring period, then the corresponding fault parameter is marked as a delayed parameter; otherwise, it is not marked; the ratio of the number of times the same operating parameter is marked as a delayed parameter to the number of times it is marked as a fault parameter in the monitoring period is marked as the delay coefficient, and the timeliness of the upload of operating parameters in the monitoring period is judged by the delay coefficient; The specific process for determining whether the timeliness of uploading operating parameters within the monitoring period meets the requirements includes: obtaining the delay threshold from the database, comparing the delay coefficient of the operating parameters with the delay threshold; if the delay coefficient is less than the delay threshold, it is determined that the timeliness of uploading operating parameters within the monitoring period meets the requirements; if the delay coefficient is greater than or equal to the delay threshold, it is determined that the timeliness of uploading operating parameters within the monitoring period does not meet the requirements, and the upload threshold of the operating parameters is adjusted downward proportionally.

2. The predictive maintenance system for equipment based on large model and agent collaboration according to claim 1, characterized in that, The specific process of marking operating parameters based on monitoring values ​​includes: the monitoring value being the maximum value of the operating parameter within the monitoring period; retrieving the monitoring threshold and upload threshold of the operating parameter from the database, where the upload threshold = t1 × monitoring threshold, and t1 is a proportionality coefficient, and 0.88 < t1 < 1; comparing the monitoring value with the monitoring threshold and upload threshold: if the monitoring value is less than or equal to the upload threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to meet the requirements, and the corresponding operating parameter is marked as a normal parameter; if the monitoring value is between the upload threshold and the monitoring threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to have potential abnormalities, and the corresponding operating parameter is marked as an upload parameter; if the monitoring value is greater than or equal to the monitoring threshold, the numerical distribution of the operating parameter within the monitoring period is deemed to not meet the requirements, and the corresponding operating parameter is marked as a fault parameter.

3. The predictive maintenance system for equipment based on large model and agent collaboration according to claim 2, characterized in that, The specific process for determining whether the operating status of the equipment meets the requirements during the monitoring period includes: if all operating parameters are marked as normal parameters during the monitoring period, the operating status of the equipment during the monitoring period is determined to meet the requirements; if any operating parameter is marked as a fault parameter during the monitoring period, a fault diagnosis signal is generated and the fault diagnosis signal and the fault parameter are sent to the mobile terminal of the management personnel; otherwise, a pre-diagnosis signal is generated and sent to the pre-diagnosis analysis module.

4. The equipment predictive maintenance system based on large model and agent collaboration according to claim 3, characterized in that, The specific process of pre-diagnosis processing includes: retrieving historical operation and maintenance data and detection data of the equipment and screening pre-diagnosis analysis periods; marking the number of times the operating parameter is marked as a fault parameter in all pre-diagnosis analysis periods as the marked value of the operating parameter; marking the ratio of the marked value to the number of pre-diagnosis analysis periods as the risk coefficient of the operating parameter; and determining whether the operating parameter has an operational risk through the risk coefficient.

5. The equipment predictive maintenance system based on large model and agent collaboration according to claim 4, characterized in that, The screening process for the pre-diagnostic analysis period includes: arranging the uploaded parameters in descending order of monitoring coefficient values ​​to obtain the uploaded sequence; retrieving the historical data of the executed monitoring periods that are exactly the same as the uploaded sequence and marking them as matching periods; and marking the L1 monitoring periods after the matching periods as pre-diagnostic analysis periods.

6. The equipment predictive maintenance system based on large model and agent collaboration according to claim 5, characterized in that, The specific process for determining whether an operating parameter has an operational risk includes: obtaining a risk threshold from the database and comparing the risk coefficient with the risk threshold; if the risk coefficient is less than the risk threshold, the operating parameter is determined not to have an abnormal risk; if the risk coefficient is greater than or equal to the risk threshold, the operating parameter is determined to have an operational risk, the corresponding operating parameter is marked as a risk parameter, a risk handling signal is generated, and the risk handling signal and the risk parameter are sent to the mobile terminal of the management personnel.

Images