Spare parts recommendation method and electronic device

Abstract

The application discloses a spare part recommendation method and electronic equipment, and relates to the technical field of equipment management, comprising receiving device alarm information collected and analyzed by a first channel and component health score collected and analyzed by a second channel; updating component state information of at least one component according to the device alarm information and the component health score, wherein the device comprises at least one component; calculating a component confidence score corresponding to the component state information according to the updated component state information, and determining a faulty component; determining a candidate spare part list corresponding to the faulty component according to a knowledge graph; calculating a multi-dimensional utility score of each candidate spare part in the candidate spare part list, and taking the candidate spare part with the highest multi-dimensional utility score as a target spare part, so as to solve the technical problems of the related art, such as the separation of monitoring and spare part systems, the lack of intelligent analysis linkage mechanism, and the low operation and maintenance efficiency, and achieve the technical effects of accurate fault prediction and automatic operation and maintenance decision.

Description

Technical Field

[0001] This invention relates to the field of equipment management technology, and in particular to a spare parts recommendation method and electronic equipment. Background Technology

[0002] As enterprises deepen their digital transformation and data centers continue to expand, the fragmentation of equipment monitoring and spare parts management systems leads to low operational efficiency. In related technologies, equipment alarm information and health status data are collected through independent channels, and fault diagnosis usually relies on a single alarm channel or performance threshold, lacking a fusion analysis mechanism. Fault diagnosis depends on manual experience, and maintenance personnel also need to manually determine the fault type, identify replacement parts, and search for available spare parts in the spare parts system after an alarm occurs. The process is cumbersome, the response is slow, and there is a lack of intelligent spare parts recommendation and automatic association capabilities. This can easily lead to decision-making errors, spare parts mismatch, or untimely preparation, resulting in low overall fault response efficiency and failing to meet the requirements of modern data centers for automated, intelligent, and timely operation and maintenance.

[0003] Therefore, in view of the shortcomings of existing technical solutions, the present invention provides a spare parts recommendation method. Summary of the Invention

[0004] This application provides a spare parts recommendation method and electronic equipment to at least solve the problems in related technologies, such as the disconnect between monitoring and spare parts systems and the lack of intelligent analysis and linkage mechanisms, which lead to low operation and maintenance efficiency.

[0005] This application provides a spare parts recommendation method, which includes: receiving equipment alarm information collected and parsed by a first channel and component health scores collected and analyzed by a second channel; updating the component status information of at least one component based on the equipment alarm information and component health scores, wherein the equipment includes at least one component; calculating the component confidence score corresponding to the updated component status information to determine the faulty component based on the updated component status information; determining a candidate spare parts list corresponding to the faulty component based on a knowledge graph; calculating the multidimensional utility score of each candidate spare parts in the candidate spare parts list, and selecting the candidate spare parts with the highest multidimensional utility score as the target spare parts.

[0006] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for implementing the steps of any of the above-described spare parts recommended methods when executing the computer program.

[0007] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described spare parts recommendation methods.

[0008] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described spare parts recommendation methods.

[0009] This application achieves a comprehensive and dynamic understanding of equipment health status by receiving and analyzing equipment alarm information collected and parsed through a first channel and component health scores collected and analyzed through a second channel. Based on the equipment alarm information and component health scores, the application updates the component status information of at least one component, where the equipment includes at least one component. Based on the updated component status information, it calculates the component confidence score corresponding to the component status information to identify the faulty component. Based on a knowledge graph, it determines a list of candidate spare parts corresponding to the faulty component. It calculates the multidimensional utility score of each candidate spare part in the candidate spare part list and selects the candidate spare part with the highest multidimensional utility score as the target spare part. Therefore, by fusing information from two channels to construct unified component status information, it combines instantaneous fault alarms with gradual health trend prediction, achieving a three-dimensional and dynamic understanding of equipment health status. This significantly improves the accuracy of fault judgment, reduces false alarms and business interruptions. Driven by a multi-objective optimization algorithm, it automatically selects the comprehensive optimal solution, promoting the transformation of operation and maintenance models from relying on manual experience to data-driven intelligent transformation. Attached Figure Description

[0010] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A flowchart illustrating a spare parts recommendation method provided in an embodiment of this application;

[0012] Figure 2 A system schematic diagram of a spare parts recommendation method provided in an embodiment of this application;

[0013] Figure 3 A schematic diagram of the intelligent decision engine process for a spare parts recommendation method provided in this application embodiment;

[0014] Figure 4 A schematic diagram of the screening and recommendation process for a spare parts recommendation method provided in this application embodiment;

[0015] Figure 5 A structural block diagram of a spare parts recommendation device provided in an embodiment of this application;

[0016] Figure 6 This is an internal structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0018] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0019] It should be noted that the terms "S1," "S2," etc., are used only for descriptive purposes and do not specifically refer to the order or sequence, nor are they intended to limit this application. They are merely for the convenience of describing the method of this application and should not be construed as indicating the sequential order of the steps. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0020] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The embodiments of this application provide a spare parts recommendation method, and the method is described in detail in conjunction with the execution flow of the spare parts recommendation method.

[0022] S101: Receive device alarm information collected and parsed by the first channel and component health score collected and analyzed by the second channel.

[0023] The device alarm information may include alarm code, alarm name, alarm level, alarm source device IP, alarm type, alarm location, etc.

[0024] Among them, the closer the component health score is to 1, the healthier the component is; the closer it is to 0, the higher the risk of component failure.

[0025] Among them, the device alarm information can be trap messages automatically reported by the device, and the component health score can be data obtained through monitoring.

[0026] The received device alarm information and component health scores may include information about some components and / or some devices.

[0027] S102: Update the component status information of at least one component based on the device alarm information and component health score, wherein the device includes at least one component.

[0028] Here, the component status information includes raw data and calculated indicator data. Specifically, the raw data may include the latest health score and the most recent trap event, while the indicator data may include the rate of health decline, operational stability, number of abnormal fluctuations, trap event frequency, historical baseline deviation, and the degree to which the current status of the component affects the health of its upstream and downstream related components.

[0029] Component status information can also be represented through the component status view.

[0030] S103: Based on the updated component status information, calculate the component confidence score corresponding to the component status information and determine the faulty component.

[0031] The higher the confidence score of a component, the more certain it is that the component needs to be replaced.

[0032] Specifically, when the confidence score of a component is higher than a preset threshold, the component is determined to be a faulty component and needs to be replaced.

[0033] S104: Based on the knowledge graph, determine the list of candidate spare parts corresponding to the faulty component.

[0034] Here, knowledge graphs are used to store and manage all devices, components, and their relationships. The goal is to integrate structured data that was originally scattered across different systems into a unified, semantically rich relational network, providing reasonable knowledge support for intelligent recommendations.

[0035] Here, candidate spare parts are backup parts prepared in advance to deal with component failures or maintenance needs.

[0036] A faulty component may include one or more candidate spare parts.

[0037] S105: Calculate the multidimensional utility score of each candidate spare part in the candidate spare parts list, and select the candidate spare part with the highest multidimensional utility score as the target spare part.

[0038] Among them, the multidimensional utility score is a comprehensive evaluation index that quantifies an object's overall performance or value across multiple dimensions through a single score.

[0039] Specifically, the candidate spare part with the highest multidimensional utility score is selected as the final recommendation result, and the status of the spare part is changed to pending shipment. The system then automatically generates a replacement work order in the work order management system.

[0040] It should be noted that this application constructs unified component status information through dual-channel information fusion, combining instantaneous fault alarms with gradual health trend prediction to achieve a three-dimensional and dynamic understanding of equipment health status, significantly improving the accuracy of fault judgment and reducing false alarms and business interruptions; driven by multi-objective optimization algorithms, it automatically selects the comprehensive optimal solution, promoting the transformation of operation and maintenance mode from relying on manual experience to data-driven intelligent transformation.

[0041] In some specific implementations, the device alarm information collected and parsed by the first channel and the component health score collected and analyzed by the second channel are received, including:

[0042] Event information actively reported by the first channel monitoring device;

[0043] The event information is analyzed through the first channel, key information is extracted and structured, and device alarm information is generated.

[0044] Here, the event information actively reported by the device can be a Trap message.

[0045] Specifically, the system receives raw Trap packets through the listening port, obtains the device IP address and port information, parses the protocol stack through the decoder, identifies the SNMP version and message structure, extracts message header information, parses PDU basic information, performs deep parsing of variable binding, and maps the extracted scattered information to a unified event object.

[0046] In this way, the transformation from raw network protocol messages to standardized alarm events is realized. Through real-time decoding, intelligent mapping and unified formatting, binary Trap data from heterogeneous devices is transformed into device alarm information rich in business semantics, which significantly improves the real-time performance, standardization and observability of alarm processing.

[0047] In some specific implementations, receiving device alarm information collected and parsed by the first channel and component health scores collected and analyzed by the second channel also includes:

[0048] At least one performance index data of each component in the device is collected by polling at a fixed period through the second channel;

[0049] Preprocess at least one performance index data to construct the feature vector corresponding to each component;

[0050] The feature vectors are input into a prediction model based on machine learning algorithms to obtain the health scores of each component.

[0051] Performance metrics data may include the number of bad sectors on the hard drive, the number of remapped sectors, and the number of memory errors. For example, for a hard drive component, the following performance metrics data may be collected: remapped sector count, read / write error rate, seek error rate, power-on time, hard drive temperature, and start-stop count; for a memory component, the following performance metrics data may be collected: correctable error count and uncorrectable error count.

[0052] In one embodiment, the frequency at which performance metric data is collected may differ for components with different historical health scores.

[0053] Specifically, based on the real-time collected performance index data of the components, a feature vector x(c) is constructed and input into the trained prediction model M. The health score H(c) can then be calculated using the defined calculation expression.

[0054] The calculation expression can be:

[0055] ;

[0056] Where H(t) is the component health score, x(t) represents the feature vector reflecting the status of the equipment component collected at time t, and M represents the machine learning model, whose function is to map the feature vector into an abstract raw score value. Let represent an activation function that maps the original output of the prediction model M to the probability interval (0, 1). Its expression is: .

[0057] Here, the machine learning algorithm in the prediction model based on machine learning algorithms can be SVM (Support Vector Machine), Random Forest algorithm, LightGBM algorithm, or XGBoost algorithm, etc.

[0058] Specifically, when the machine learning algorithm is LightGBM, the training process of the prediction model based on the machine learning algorithm may include: extracting performance index data of equipment components at N historical time points from the historical database. For each time point t, the performance index data is preprocessed (e.g., calculating differences, logarithmic transformation, etc.), and the processed performance index data is arranged in a predefined order to form a feature vector x(t). The label value y(t) is assigned according to the rule that if the component fails within a preset time window T after time t, the value is 0, otherwise it is 1. This construction yields the training set {X_train, Y_train}. The prediction model is constructed using the LightGBM algorithm, and the model is trained using the training set {X_train, Y_train} to obtain the prediction model M.

[0059] In this way, early warning and refined prediction of component health status can be achieved, improving the level of intelligence in fault prediction, which helps to optimize maintenance strategies, reduce operation and maintenance costs, and improve system reliability and security.

[0060] In some specific implementations, the component status information of at least one component is updated based on device alarm information and component health scores, including:

[0061] The target component is determined based on the device identifier in the device alarm information and the component identifier in the component health score;

[0062] Based on device alarm information and component health scores, update the component status information corresponding to the target component.

[0063] Specifically, updating the status information of the target component can be done by querying and updating the status information of the target component that currently exists, or by creating the status information of the target component and filling in the data if no status information is found after querying.

[0064] Specifically, based on the device identifier (such as device IP) and component identifier in the message, the status information of the device component is found or created. This information is used to dynamically store the component's latest health score and its historical sequence, the most recently received Trap events and their timestamps. Based on the status information, heterogeneous information about the same entity that arrives at different times is associated through the stored data and internal calculation logic to form unified component status information.

[0065] In one embodiment, the storage level of data is determined based on the data importance of device alarm information and component health scores. If the data is stored in the hot data layer, it is written to a memory cache, a fast index is created, a time-to-live (TTL) is set, and the recent access timestamp and access counter are updated. If the data is stored in the warm data layer, a suitable time-series database is selected, and the data is stored in time-partitioned format to optimize query performance, create a composite index, and implement data compression to reduce storage space. If the data is stored in the cold data layer, it is converted to a columnar storage format, a high-ratio compression algorithm is applied, metadata index files are added to support fast location, and the data is uploaded to cost-optimized object storage.

[0066] The hot data layer is used to store recently active devices, high-frequency access data, and key business components; the warm data layer is used to store recent data, historical data with medium access frequency, and data that needs to be queried quickly; and the cold data layer is used to store historical archive data, low-frequency access data, and data that requires long-term storage.

[0067] The quick index includes Device ID → Component ID → Data Record.

[0068] The composite index includes time + device ID + component ID.

[0069] Specifically, an LRU (Least Recently Used) eviction policy is implemented in the hot data layer; memory usage is monitored and automatic cleanup is triggered; and important data is periodically synchronized to the warm data layer.

[0070] Specifically, based on time, capacity, and business conditions, data that meets the migration criteria is identified, data is extracted and transformed, a target directory structure is created in the cold data layer, data files are uploaded to object storage, the integrity and readability of uploaded files are verified, the global data index and metadata database are updated, migrated data is deleted from the warm data layer, and migration audit logs and performance metrics are recorded.

[0071] In this way, the tiered storage processing mechanism ensures the optimal storage strategy for data throughout its lifecycle, balancing performance, cost, and accessibility requirements, while reducing operational complexity through automated migration processes.

[0072] In this way, scattered equipment status information can be dynamically aggregated and correlated to achieve unified and continuous perception of the health status of equipment components.

[0073] In some specific implementations, based on device alarm information and component health scores, the component status information corresponding to the target component is updated, including:

[0074] Based on equipment alarm information and component health scores, determine the business importance, alarm severity, and topology impact;

[0075] The update strategy is determined based on the importance of the business, the severity of the alarm, and the topology impact. The update strategy includes real-time updates and batch updates.

[0076] In response to the real-time update policy, the alarm event list in the status information is updated based on the device alarm information;

[0077] Update the health score sequence in the status information based on the health score;

[0078] Update the component's health trend based on its health score;

[0079] In response to the update strategy being batch updates, multiple alarm messages and health scores within a preset time period are aggregated.

[0080] Here, real-time updates are used for critical data, which are processed and persisted immediately; batch updates are used for regular data, which are processed after accumulating to a certain quantity or time window.

[0081] Specifically, check the importance of the business and ensure that critical business components are updated in real time; assess the severity of alarms and handle severe alarms immediately; consider the impact on the topology and prioritize the handling of components with downstream dependencies.

[0082] In one embodiment, in addition to the rate of decline in health, other metrics can be calculated in the component status information, such as the moving average of health, variance, and the correlation between health and Trap events.

[0083] In one embodiment, Trap messages and health scores may be out of sync. To associate Trap events and health scores, a time window (e.g., 5 minutes) can be defined. Only Trap messages and health scores within this time window will be associated with the same component status information. If a Trap message occurs within the last 5 minutes, the status corresponding to that Trap message is used; if there are multiple Trap messages, the most severe status is taken.

[0084] In this way, combining instantaneous, clear fault alerts with gradual, predictive health decline trends overcomes the limitations of a single data source.

[0085] In some specific implementations, based on the updated component status information, a component confidence score corresponding to the component status information is calculated to determine the faulty component. The calculation formula includes:

[0086] ;

[0087] Where C is the component confidence score; W is the weighting function for device alarm information, assigned a value according to the alarm level; and H is the latest health score. The rate at which the health score decreases within a preset time window; The maximum descent rate threshold is a preset value used for standardization; α, β, and γ are weighting coefficients that satisfy α+β+γ=1.

[0088] Where W can be the standardized data.

[0089] Specifically, because alarms, health status, and trends complement each other, even if one indicator is noisy, other indicators can offset its impact, reducing the false alarm rate; the introduction of the health decline rate helps to distinguish between temporary fluctuations and real faults, improving the accuracy of detection.

[0090] In this way, by integrating multi-dimensional data, the limitations of a single indicator can be reduced, and the comprehensiveness and accuracy of fault detection can be improved.

[0091] In some specific implementations, before determining the candidate spare parts list corresponding to the faulty component based on the knowledge graph, the method includes:

[0092] Acquire multi-source data;

[0093] Based on multi-source data, determine the relationships between nodes, including equipment model nodes, component model nodes, and spare part instance nodes. The relationships between nodes include compatibility relationships and instance relationships.

[0094] A knowledge graph is constructed based on the relationships between nodes.

[0095] Here, multi-source data can include equipment management and monitoring data, spare parts management data, and manufacturer compatibility documents, etc.

[0096] Among them, equipment management and monitoring data can be obtained from equipment monitoring systems or databases, and may include equipment manufacturers, models, specifications, component information (manufacturer, model, specifications, serial number, etc.); spare parts management data can be obtained from spare parts inventory management systems, and may include whole machine information (manufacturer, model, specifications) and spare parts information (manufacturer, model, specifications, etc.); manufacturer compatibility documents can be obtained from documents provided by equipment manufacturers, and may include compatibility rules between equipment and components (such as supported component types, capacity limits, etc.).

[0097] In one embodiment, data processing is performed on multi-source data, such as data cleaning, data transformation, and relationship establishment.

[0098] Among them, the device model node can represent an abstract whole machine device model (such as a server model), and the attribute information can include manufacturer, model, device type, etc.; the component model node can represent an abstract component specification (such as a hard drive model), and the attribute information can include model, manufacturer, component type and specification information, etc.; the specific spare part node can represent a specific physical spare part in the warehouse with a unique number, and the attributes can include spare part ID, warehouse entry time, mean time between failures, inventory status, etc.

[0099] Here, compatibility relationships refer to the compatibility between device model nodes and component model nodes. Specifically, a pointer from a device model node to a component model node indicates that a certain device model can be equipped with a certain component model.

[0100] Here, the instance relationship represents the instantiation association between the component model node and the spare part instance node. Specifically, a pointer from a specific spare part node to a component model node indicates that this specific spare part is an instance of that component model.

[0101] Specifically, data is extracted from multi-source data to generate node and relationship data, which are then imported into a graph database for storage to obtain a knowledge graph.

[0102] In one embodiment, the knowledge graph is updated periodically or in real time to reflect changes in devices, spare parts, or compatibility rules (e.g., by monitoring data source changes or by timed batch processing).

[0103] In this way, by interconnecting the scattered equipment, spare parts, and compatibility rule data to form a unified network, the originally isolated static data is transformed into a semantic network capable of intelligent reasoning. This enables efficient and accurate answers to complex queries, achieving a qualitative leap from data storage to precise spare parts recommendation and decision support, thereby improving operational efficiency and accuracy.

[0104] In some specific implementations, a list of candidate spare parts corresponding to the faulty component is determined based on a knowledge graph, including:

[0105] Determine the equipment model node based on the equipment model corresponding to the faulty component;

[0106] Based on compatibility relationships, determine the component model nodes that are compatible with the device model node;

[0107] Based on the instance relationships, determine the spare part instance node corresponding to the component model;

[0108] Based on the filtering criteria, the spare parts instance nodes are filtered to determine the candidate spare parts list.

[0109] The filtering criteria may include at least one of inventory availability, cost threshold, or priority.

[0110] Specifically, use a graph query language (such as Cypher for Neo4j) to write the query logic.

[0111] In some specific implementations, the multidimensional utility score of each candidate spare part in the candidate spare parts list is calculated, including:

[0112] Calculate the reliability utility of the candidate spare parts based on their mean unobstructed time.

[0113] Calculate the inventory leveling utility of the candidate spare parts based on the remaining inventory quantity of the corresponding model.

[0114] Calculate the lifecycle utility of the candidate spare parts based on their entry time, current time, and lifecycle.

[0115] Calculate the multidimensional utility score of candidate spare parts based on reliability utility, inventory leveling utility, and lifecycle utility.

[0116] Here, reliability utility characterizes the probability of a candidate spare part operating without failure during the planned mission period.

[0117] Specifically, reliability utility can be calculated using the following formula:

[0118] ;

[0119] Among them, U rel For reliability purposes, MTBF is the mean time between failures for candidate spare parts, and t can be the desired stable operating time of the equipment, which can be set to 1 year.

[0120] Here, the inventory equilibrium utility characterizes the impact of recommending the candidate spare part on the health of the inventory structure.

[0121] Specifically, the inventory equilibrium utility can be calculated using the following formula:

[0122] ;

[0123] Among them, U inv For inventory equilibrium utility, Q left To determine the remaining inventory quantity of the corresponding model after recommending the candidate spare part, Q avg Q represents the average inventory level for all models. max and Q min The maximum and minimum inventory levels for all models.

[0124] Here, lifecycle utility characterizes the ability to control the risk of spare parts inventory aging.

[0125] Specifically, life-cycle utility can be calculated using the following formula:

[0126] ;

[0127] Among them, U life For lifecycle utility, t now It is the current time, t in It is the time when spare parts are received into the warehouse, T max Maximum recommended inventory lifecycle.

[0128] Where, if t now -t in >T max Then set U life =0 (indicates that the spare parts are expired).

[0129] Specifically, the multidimensional utility score of candidate spare parts is calculated by weighting reliability utility, inventory leveling utility, and lifecycle utility, as shown in the following formula:

[0130] ;

[0131] Among them, w rel w inv w life These are pre-configured weighting coefficients, and the sum of the weighting coefficients is 1.

[0132] In this way, multiple potentially conflicting decision dimensions are integrated into a unified comprehensive score, thereby avoiding overall decision imbalance caused by unilaterally pursuing a certain indicator, and enabling more scientific and comprehensive optimization choices to be made in complex trade-offs.

[0133] In one embodiment, Figure 2 This is a system schematic diagram in an embodiment of this application, such as... Figure 2 As shown, the system in this application includes: a Trap message listening and parsing channel, a predictive analysis channel, an intelligent decision engine, a component knowledge graph module, and a multi-objective optimization recommendation module.

[0134] Specifically, the Trap message listening and parsing channel is used to respond instantly to event messages proactively reported by each monitored device. Its implementation is as follows: A daemon process is deployed on the monitoring server to continuously listen to the port specified by the device management system and capture event messages proactively reported by the devices; the received Trap message packets are parsed, key information is extracted from the received messages and structured to generate alarm event objects in a unified format, including: alarm code, alarm name, alarm level, alarm source device IP, alarm type, alarm location, etc.; the alarm events are then published for consumption by the intelligent decision engine.

[0135] Specifically, the predictive analytics channel defines a unified quantitative standard for the health status of equipment components through the construction and calculation of a health scoring model. This standard represents the predicted probability that the component will maintain normal operation within a specific time window T in the future. This includes: data collection of component monitoring indicators and training and calculation of the health scoring model.

[0136] Specifically, the intelligent decision engine is responsible for receiving and fusing dual-channel information and constructing a device status view (i.e., component status information) based on this, thereby assessing the component status and making a replacement decision. A flowchart of the intelligent decision engine is shown below. Figure 3As shown, the intelligent decision engine's process includes: listening to and receiving alarm events (i.e., device alarm information) from the Trap channel and health score messages (i.e., component health scores) from the predictive analytics channel; finding or creating a state context object (i.e., state information) for the device component based on the device identifier (e.g., device IP) and component identifier in the message. This object is used to dynamically store the component's latest HS score and its historical sequence, the most recently received Trap events and their timestamps; based on the context object, associating heterogeneous information about the same entity arriving at different times through its stored data and internal calculation logic to form a unified state view; calculating a comprehensive confidence score C based on the component's state view to quantify the certainty of the current judgment; comparing the comprehensive confidence score C with a preset threshold, and sending a replacement instruction (i.e., faulty component) to the knowledge graph when it is higher than the preset threshold, otherwise continuing to observe.

[0137] Specifically, the component knowledge graph module aims to integrate structured data that was originally scattered across different systems into a unified, semantically rich relational network, providing reasonable knowledge support for intelligent recommendations. This includes: multi-source data preparation; knowledge graph model definition; data extraction from multi-source data to generate node and relational data, which is then imported into a graph database for storage; and receiving replacement instructions from the intelligent decision engine, outputting a list of available spare parts compatible with the device, which is then filtered by multi-objective optimization recommendations.

[0138] Specifically, the filtering and recommendation process of the multi-objective optimization recommendation module is illustrated in the diagram below. Figure 4 As shown, the screening and recommendation process includes: receiving a list of compatible parts output from the part knowledge graph module as a candidate list; traversing the candidate list and calculating the multi-dimensional utility score for each part; calculating the comprehensive utility score and sorting them; calculating the comprehensive utility score by weighted summing of the above multiple utility scores; selecting the spare part with the highest comprehensive utility score as the final recommendation result, and changing the status of the spare part to "pending dispatch" status. The system then automatically generates a replacement work order in the work order management system, completing this recommendation process.

[0139] It should be understood that, although Figures 1-4 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figures 1-4At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0141] An embodiment of this application also provides a spare parts recommendation device, comprising: a first processing module 501, configured to receive and parse device alarm information collected by a first channel and component health scores collected and analyzed by a second channel; a second processing module 502, configured to update component status information of at least one component based on the device alarm information and component health scores, wherein the device includes at least one component; a third processing module 503, configured to calculate a component confidence score corresponding to the updated component status information and determine a faulty component based on the updated component status information; a fourth processing module 504, configured to determine a candidate spare parts list corresponding to the faulty component based on a knowledge graph; and a fifth processing module 505, configured to calculate a multidimensional utility score for each candidate spare parts in the candidate spare parts list and select the candidate spare parts with the highest multidimensional utility score as the target spare parts.

[0142] As a preferred implementation, in this embodiment of the application, the first processing module 501 is specifically used for: actively reporting event information through the first channel monitoring device;

[0143] The event information is analyzed through the first channel, key information is extracted and structured, and device alarm information is generated.

[0144] As a preferred implementation, in this embodiment of the application, the first processing module 501 is further configured to: collect at least one performance index data of each component in the device by polling at a fixed period through the second channel;

[0145] Preprocess at least one performance index data to construct the feature vector corresponding to each component;

[0146] The feature vectors are input into a prediction model based on machine learning algorithms to obtain the health scores of each component.

[0147] In a preferred embodiment of this application, the second processing module 502 is specifically used to: determine the target component based on the device identifier in the device alarm information and the component identifier in the component health score;

[0148] Based on device alarm information and component health scores, update the component status information corresponding to the target component.

[0149] As a preferred implementation, in this embodiment of the application, the second processing module 502 is further configured to: determine the business importance, alarm severity and topology impact based on the device alarm information and component health score;

[0150] The update strategy is determined based on the importance of the business, the severity of the alarm, and the topology impact. The update strategy includes real-time updates and batch updates.

[0151] In response to the real-time update policy, the alarm event list in the status information is updated based on the device alarm information;

[0152] Update the health score sequence in the status information based on the health score;

[0153] Update the component's health trend based on its health score;

[0154] In response to the update strategy being batch updates, multiple alarm messages and health scores within a preset time period are aggregated.

[0155] In a preferred embodiment of this application, the calculation formula of the third processing module 503 includes:

[0156] ;

[0157] Where C is the component confidence score; W is the weighting function for device alarm information, assigned a value according to the alarm level; and H is the latest health score. The rate at which the health score decreases within a preset time window; The maximum descent rate threshold is a preset value used for standardization; α, β, and γ are weighting coefficients that satisfy α+β+γ=1.

[0158] In a preferred embodiment of this application, the device further includes a construction module, which is specifically used for: acquiring multi-source data;

[0159] Based on multi-source data, determine the relationships between nodes, including equipment model nodes, component model nodes, and spare part instance nodes. The relationships between nodes include compatibility relationships and instance relationships.

[0160] A knowledge graph is constructed based on the relationships between nodes.

[0161] In a preferred embodiment of this application, the fourth processing module 504 is specifically used to: determine the equipment model node according to the equipment model corresponding to the faulty component;

[0162] Based on compatibility relationships, determine the component model nodes that are compatible with the device model node;

[0163] Based on the instance relationships, determine the spare part instance node corresponding to the component model;

[0164] Based on the filtering criteria, the spare parts instance nodes are filtered to determine the candidate spare parts list.

[0165] In a preferred embodiment of this application, the fifth processing module 505 is specifically used to: calculate the reliability utility of the candidate spare parts based on the average unobstructed time of the candidate spare parts;

[0166] Calculate the inventory leveling utility of the candidate spare parts based on the remaining inventory quantity of the corresponding model.

[0167] Calculate the lifecycle utility of the candidate spare parts based on their entry time, current time, and lifecycle.

[0168] Calculate the multidimensional utility score of candidate spare parts based on reliability utility, inventory leveling utility, and lifecycle utility.

[0169] For a description of the features in the embodiment corresponding to the spare parts recommendation device, please refer to the relevant description of the embodiment corresponding to the spare parts recommendation method, which will not be repeated here.

[0170] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above-described spare parts recommended method embodiments.

[0171] This electronic device can be a server, and its internal structure diagram can be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and the database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores spare parts recommendation data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a spare parts recommendation method.

[0172] Embodiments of this application also provide a computer-readable storage medium storing a computer program configured to execute the steps in any of the above-described spare parts recommendation method embodiments when run.

[0173] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0174] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described spare parts recommendation method embodiments.

[0175] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the above-described spare parts recommended method embodiments.

[0176] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0177] The foregoing has provided a detailed description of a spare parts recommendation method, electronic device, storage medium, and computer program product provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to aid in understanding the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A spare parts recommendation method, characterized in that, The method includes: Receive device alarm information collected and analyzed by the first channel and component health scores collected and analyzed by the second channel; Based on the device alarm information and the component health score, update the component status information of at least one component, wherein the device includes at least one component; Based on the updated component status information, calculate the component confidence score corresponding to the component status information to determine the faulty component; Based on the knowledge graph, a list of candidate spare parts corresponding to the faulty component is determined; Calculate the multidimensional utility score of each candidate spare part in the candidate spare parts list, and select the candidate spare part with the highest multidimensional utility score as the target spare part; The step of calculating the component confidence score corresponding to the updated component status information and determining the faulty component based on the updated component status information includes the following calculation formula: ； Where C is the component confidence score; W is the weighting function for device alarm information, assigned a value according to the alarm level; and H is the latest health score. The rate at which the health score decreases within a preset time window; The maximum descent rate threshold is a preset value used for standardization; α, β, and γ are weighting coefficients that satisfy α + β + γ = 1. Before determining the candidate spare parts list corresponding to the faulty component based on the knowledge graph, the method includes: Acquire multi-source data; Based on the multi-source data, the relationships between nodes are determined, wherein the nodes include equipment model nodes, component model nodes, and spare part instance nodes, and the relationships between the nodes include compatibility relationships and instance relationships. Construct a knowledge graph based on the nodes and the relationships between them; The step of determining the candidate spare parts list corresponding to the faulty component based on the knowledge graph includes: Based on the equipment model corresponding to the faulty component, determine the equipment model node; Based on compatibility relationships, determine the component model nodes that are compatible with the device model node; Based on the instance relationship, determine the spare part instance node corresponding to the component model; The spare parts instance nodes are filtered according to the filtering conditions to determine the candidate spare parts list; The calculation of the multidimensional utility score of each candidate spare part in the candidate spare parts list includes: The reliability utility of the candidate spare parts is calculated based on the average unobstructed time of the candidate spare parts. Calculate the inventory leveling utility of the candidate spare parts based on the remaining inventory quantity of the corresponding model of the candidate spare parts; Calculate the lifecycle utility of the candidate spare parts based on their entry time, current time, and lifecycle. The multidimensional utility score of the candidate spare parts is calculated based on the reliability utility, the inventory balancing utility, and the lifecycle utility.

2. The spare parts recommendation method according to claim 1, characterized in that, The process of receiving and parsing device alarm information from the first channel and component health scores from the second channel includes: The first channel is used to listen for event information actively reported by the device; The event information is parsed through the first channel, key information is extracted and structured, and the device alarm information is generated.

3. The spare parts recommendation method according to claim 1, characterized in that, The process of receiving and parsing device alarm information from the first channel and component health scores from the second channel also includes: At least one performance index data of each component in the device is collected by polling at a fixed period through the second channel; The at least one performance index data is preprocessed to construct the feature vector corresponding to each component; The feature vectors are input into a prediction model based on a machine learning algorithm to obtain the health scores of each component.

4. The spare parts recommendation method according to claim 1, characterized in that, The step of updating the component status information of at least one component based on the device alarm information and the component health score includes: The target component is determined based on the device identifier in the device alarm information and the component identifier in the component health score; Based on the device alarm information and the component health score, update the component status information corresponding to the target component.

5. The spare parts recommendation method according to claim 4, characterized in that, The step of updating the component status information corresponding to the target component based on the device alarm information and the component health score includes: Based on the device alarm information and the component health score, determine the business importance, alarm severity, and topology impact; Based on the importance of the business, the severity of the alarm, and the topology impact, an update strategy is determined, which includes real-time updates and batch updates. In response to the update strategy being real-time updates, the alarm event list in the status information is updated according to the device alarm information; Update the health score sequence in the status information based on the health score; Update the health trend of the component based on the health score; In response to the update strategy being a batch update, multiple alarm messages and health scores within a preset time period are updated and aggregated.

6. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the spare parts recommendation method as described in any one of claims 1 to 5 when executing the computer program.

Images