After-sales work order automatic classification and dispatch optimization method based on internet of things analysis

Abstract

The application discloses an after-sales work order automatic classification and dispatching optimization method based on Internet of Things analysis, comprising the following steps: constructing a dynamic digital twin model of an after-sales work order service, performing collaborative anomaly analysis on a device group based on device state data in the model, and automatically generating a warning work order when more than a preset threshold of devices in the group are identified to have collaborative anomalies, and marking a predictive spare part demand list for the warning work order; in response to a received user repair work order or warning work order, performing multi-resource collaborative optimization dispatching decision based on dynamic constraint parameters provided by the dynamic digital twin model, taking the predictive spare part demand list as input; and collecting feedback data to drive model updating after service is performed according to the dispatching scheme. The application constructs a dynamic digital twin model to comprehensively represent device, engineer and supply chain states, performs multi-resource collaborative optimization dispatching decision, ensures synchronous matching of engineers and spare part resources, and reduces on-site waiting time.

Description

Technical Field

[0001] This invention relates to the field of work order management technology, and in particular to an optimized method for automatic classification and dispatch of after-sales work orders based on Internet of Things (IoT) analysis. Background Technology

[0002] Traditional after-sales work order management systems typically handle equipment repair requests and dispatch manually or semi-automatically. This approach often suffers from the following drawbacks: First, repair requests require manual reading of the repair description and querying of equipment history to determine the fault type and required skills, resulting in long processing times and a high risk of errors. Engineers' repair experience is often limited to individual levels and cannot be systematically extracted and reused, leading to inconsistent repair quality. Second, engineer dispatch often only considers geographical location or simple skill matching, lacking a comprehensive assessment of spare parts inventory, logistics status, and equipment health. This can result in engineers arriving on-site unable to perform timely repairs due to insufficient spare parts. Furthermore, existing systems are mostly reactive, generating work orders only after equipment failure occurs, failing to provide early warnings based on the operational status of equipment groups. This leads to delays in spare parts preparation, impacting work order service progress and reducing the efficiency of after-sales service management. Summary of the Invention

[0003] To address at least one of the aforementioned technical problems, this invention provides a method for automatically classifying and optimizing after-sales work orders based on Internet of Things (IoT) analysis.

[0004] In a first aspect, the present invention provides a method for automatic classification and dispatch optimization of after-sales work orders based on Internet of Things (IoT) analysis, the method comprising:

[0005] Construct a dynamic digital twin model for after-sales work order services. The dynamic digital twin model is used to represent the status of equipment, service resources, and material supply chain.

[0006] Based on the equipment status data in the dynamic digital twin model, a collaborative anomaly analysis is performed on the equipment group. When a collaborative anomaly is detected in the equipment group that exceeds a preset threshold, an early warning work order is automatically generated, and a predictive spare parts demand list is marked on the early warning work order.

[0007] In response to the received user repair work order or the warning work order, based on the dynamic constraint parameters provided by the dynamic digital twin model, the predictive spare parts demand list is used as input to perform multi-resource collaborative optimization dispatch decision-making;

[0008] The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list.

[0009] After the service is executed according to the dispatch plan, feedback data is collected and fed back to the dynamic digital twin model to drive model updates.

[0010] Preferably, the dynamic digital twin model of the after-sales work order service includes:

[0011] The device digital twin sub-model is constructed based on the target device's real-time IoT time-series data, historical maintenance data, and static device archives. It is used to characterize the device's real-time health status, performance degradation trend, and historical fault map.

[0012] The engineer digital twin sub-model is built based on the engineer's real-time location, skill tags, historical work order execution data, task queue, and implicit experience features extracted from maintenance process data. It is used to dynamically evaluate the engineer's availability, skill proficiency, and task load.

[0013] The supply chain digital twin sub-model is built based on dynamic inventory data of spare parts in warehouses at all levels, in-transit logistics information and supplier delivery cycle data, and is used to characterize the real-time availability and replenishment expectations of spare parts.

[0014] Preferably, the method for extracting implicit experiential features from the engineer's digital twin model includes:

[0015] Analyze the differences between the final solutions adopted by engineers in their historical maintenance records and the standard solutions in the knowledge base;

[0016] Analyze the correspondence between the fault characteristic multimedia data uploaded by engineers through the terminal during the maintenance process and the associated root causes of the fault;

[0017] Analyze the completeness and stability curves of IoT performance indicators after the equipment has been repaired by engineers;

[0018] Based on all the analysis results, in addition to the basic skill tags for engineers, implicit expertise tags for specific equipment models and specific failure modes are generated for engineers through dynamic weighting.

[0019] Preferably, before responding to the received user repair work order or the warning work order, the method further includes an automated work order classification step:

[0020] For user repair work orders, based on the repair text description and the real-time abnormal data stream provided by the digital twin sub-model of the equipment, a multimodal fusion classification model is used to output classification labels containing the root cause of the fault, the required skills, and key spare parts.

[0021] For early warning work orders, classification labels are generated based on the results of collaborative anomaly analysis.

[0022] Preferably, the multi-resource collaborative optimization of order dispatch decision-making specifically includes:

[0023] To minimize the overall system response cost and risk, a multi-objective optimization model is constructed and solved.

[0024] The decision variables of the optimization model include the matching relationship between work orders and engineers, and the supply path of required spare parts; the constraints of the optimization model are derived from the dynamic parameters of the dynamic digital twin model.

[0025] Solve the multi-objective optimization model and output a dispatch plan that includes a specified engineer, a recommended spare parts retrieval plan, and a suggested execution sequence.

[0026] Preferably, the overall system response cost includes:

[0027] Explicit costs: Engineer travel costs, time costs, spare parts logistics costs;

[0028] Hidden risk costs include: downtime losses predicted by the equipment's digital twin model, backlog of work orders due to task delays, and maintenance failure risk costs due to skill mismatch.

[0029] Preferably, the dynamic parameters include:

[0030] The system includes the engineer's real-time location and future availability window, the predicted rate of equipment failure deterioration, the real-time inventory and lock-up status of spare parts, and the dynamic skill matching degree of the engineer for the root cause of the current work order failure.

[0031] Secondly, the present invention also provides an automatic classification and dispatch optimization system for after-sales work orders based on Internet of Things (IoT) analysis, the system comprising:

[0032] The model building unit is used to build a dynamic digital twin model of after-sales work order service. The dynamic digital twin model is used to represent the equipment status, service resource status and material supply chain status.

[0033] An anomaly analysis unit is used to perform collaborative anomaly analysis on equipment groups based on equipment status data in the dynamic digital twin model. When a collaborative anomaly is detected in equipment within the group that exceeds a preset threshold, an early warning work order is automatically generated, and a predictive spare parts demand list is marked on the early warning work order.

[0034] The work order response unit is used to respond to the received user repair work order or the warning work order, and to make multi-resource collaborative optimization dispatch decisions based on the dynamic constraint parameters provided by the dynamic digital twin model and the predictive spare parts demand list as input.

[0035] The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list.

[0036] The model update unit is used to collect feedback data after performing services according to the dispatch scheme, and to feed the feedback data back to the dynamic digital twin model to drive model updates.

[0037] Thirdly, the present invention also provides an electronic device including a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device performs the method as described in the first aspect above and any possible implementation thereof.

[0038] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor of an electronic device, cause the processor to perform a method as described in the first aspect above and any possible implementation thereof.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] 1) By constructing a dynamic digital twin model to comprehensively represent the status of equipment, engineers, and the supply chain, unified management of multi-dimensional data is achieved; based on equipment group collaborative anomaly analysis, early warning work orders are automatically generated and predictive spare parts requirements are marked, realizing a shift from passive response to proactive early warning; through multi-resource collaborative optimization of dispatch decisions, synchronous matching of engineers and spare parts resources is ensured, reducing on-site waiting time; a feedback data feedback mechanism enables continuous model optimization and improves the system's adaptability. The equipment digital twin sub-model, based on real-time IoT data, historical maintenance data, and static archives, achieves accurate assessment of equipment health status and fault prediction; the engineer digital twin sub-model integrates real-time location, skill tags, historical work orders, and implicit experience features to achieve a dynamic profile of engineer capabilities; the supply chain digital twin sub-model integrates multi-level inventory, in-transit logistics, and supplier cycles to achieve real-time perception of spare parts availability, providing comprehensive data support for dispatch decisions.

[0041] 2) Through automated work order classification, a multimodal fusion classification model is used to process user repair work orders, processing the repair text and equipment anomaly data, automatically outputting the root cause of the fault, required skills, and key spare parts tags. For early warning work orders, classification tags are directly generated based on collaborative anomaly analysis results, achieving refined work order classification and providing structured input for subsequent dispatch decisions, improving classification accuracy and processing efficiency. By analyzing solution differences, the correspondence between fault characteristic multimedia data and root causes, and performance recovery integrity and stability curves, the implicit experience characteristics of engineers are systematically extracted. Implicit expertise tags for specific equipment models and fault modes are generated through dynamic weighting, enabling the accumulation and reuse of experience knowledge, improving the matching accuracy between engineers and tasks, and reducing the risk of repair failure.

[0042] 3) By constructing a multi-objective optimization model, aiming to minimize the overall system response cost and risk, the model comprehensively considers multi-dimensional decision variables such as the matching relationship between work orders and engineers, and spare parts supply paths. The optimization model integrates dynamic parameters from a dynamic digital twin model, outputting a dispatch plan that includes a designated engineer, spare parts retrieval plan, and execution sequence, achieving multi-resource collaborative optimization scheduling. The overall system response cost includes not only explicit costs such as engineer travel, time, and spare parts logistics, but also implicit costs such as downtime losses predicted by the equipment digital twin sub-model, work order backlog caused by task delays, and maintenance failure risks caused by skill mismatches, achieving a comprehensive trade-off between cost and risk and improving the scientific nature of decision-making. Dynamic parameters include the engineer's real-time location and available time window, the predicted speed of equipment failure deterioration, the real-time inventory and lockout status of spare parts, and the dynamic skill matching degree of the engineer for the root cause of the current work order failure, providing real-time and accurate constraints for the multi-objective optimization model to ensure the executability and optimality of the dispatch plan.

[0043] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0045] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0046] Figure 1 A flowchart illustrating an optimized method for automatic classification and dispatch of after-sales work orders based on Internet of Things (IoT) analysis, provided as an embodiment of the present invention.

[0047] Figure 2This is a schematic diagram of the structure of an after-sales work order automatic classification and dispatch optimization system based on Internet of Things analysis, provided in an embodiment of the present invention. Detailed Implementation

[0048] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0049] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0050] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0051] Please see Figure 1 , Figure 1 This is a flowchart illustrating an optimized method for automatic classification and dispatch of after-sales work orders based on Internet of Things (IoT) analysis, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0052] S10. Construct a dynamic digital twin model for after-sales work order services. The dynamic digital twin model is used to represent the equipment status, service resource status, and material supply chain status.

[0053] S20. Based on the equipment status data in the dynamic digital twin model, perform collaborative anomaly analysis on the equipment group. When a collaborative anomaly is detected in the equipment group that exceeds a preset threshold, an early warning work order is automatically generated, and a predictive spare parts demand list is marked on the early warning work order.

[0054] S30. In response to the received user repair work order or the early warning work order, based on the dynamic constraint parameters provided by the dynamic digital twin model, and taking the predictive spare parts demand list as input, perform multi-resource collaborative optimization dispatch decision.

[0055] The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list.

[0056] S40. After performing the service according to the dispatch plan, collect feedback data and feed the feedback data back to the dynamic digital twin model to drive model updates.

[0057] In traditional after-sales work order management systems, equipment status data, engineer resource information, and spare parts inventory data are typically scattered across different business systems. Due to this lack of information sharing, work order dispatch cannot comprehensively consider multiple factors such as equipment health status, engineer skill matching, and spare parts availability. Furthermore, after-sales work order service often relies on manual or semi-automated dispatch methods, leading to inefficiency and unsatisfactory service quality. Therefore, in step S10, a dynamic digital twin model is first constructed, holographically mapping equipment, engineers, and the spare parts supply chain from the physical world into a virtual space. This model unifies the real-time status of equipment, engineer skill profiles, and spare parts availability, breaking down data silos and making work order dispatch more rational and efficient.

[0058] In one embodiment, the dynamic digital twin model of the after-sales work order service includes:

[0059] The device digital twin sub-model is constructed based on the target device's real-time IoT time-series data, historical maintenance data, and static device archives. It is used to characterize the device's real-time health status, performance degradation trend, and historical fault map.

[0060] The engineer digital twin sub-model is built based on the engineer's real-time location, skill tags, historical work order execution data, task queue, and implicit experience features extracted from maintenance process data. It is used to dynamically evaluate the engineer's availability, skill proficiency, and task load.

[0061] The supply chain digital twin sub-model is built based on dynamic inventory data of spare parts in warehouses at all levels, in-transit logistics information and supplier delivery cycle data, and is used to characterize the real-time availability and replenishment expectations of spare parts.

[0062] For ease of understanding, the following terms in this embodiment will be explained first:

[0063] Static equipment profile: Fixed attribute information of the equipment, including equipment model, manufacturing date, installation location, technical parameters, maintenance history, etc., used for equipment identification and performance benchmark setting.

[0064] Real-time health status: A health score of the device calculated based on real-time operating data, reflecting the current operating status of the device, usually expressed as 0-100 points.

[0065] Historical Fault Map: Structured data that records the time, type, cause, and maintenance plan of historical equipment faults, used to analyze fault patterns and optimize maintenance strategies.

[0066] Skill tags: The professional skills and ability levels of engineers, such as "Advanced Elevator Maintenance" or "Intermediate Air Conditioning System," used to match work orders with engineers.

[0067] Historical work order execution data: Records of work orders completed by engineers in the past, including completion rate, average response time, customer rating, repair quality, etc., used to evaluate the engineer's work ability and efficiency.

[0068] Task Queue: A list of tasks currently pending by the engineer, including the number of work orders, urgency level, and estimated completion time, used to assess the engineer's workload.

[0069] Implicit experience characteristics: Experience and knowledge accumulated by engineers in long-term maintenance practice that are difficult to describe with explicit skill labels, such as familiarity with specific equipment models and the ability to quickly diagnose specific fault modes.

[0070] Available status: Whether the engineer is currently available to accept new tasks, including idle, busy, offline, etc.

[0071] Skill proficiency: The degree to which an engineer has mastered a particular skill, usually expressed on a scale of 0-100, based on historical work order execution data and maintenance quality assessment.

[0072] Task load: The amount of work currently undertaken by an engineer, usually expressed as the current number of tasks divided by the maximum capacity, used for task allocation and load balancing.

[0073] Dynamic inventory data: The real-time inventory quantity of spare parts in warehouses at all levels, including inventory information of central warehouses, regional warehouses, and field warehouses, which will change in real time with inbound and outbound operations.

[0074] In-transit logistics information: Status information of spare parts during transportation, including current location, mode of transportation, and estimated arrival time.

[0075] Supplier delivery cycle: The time required for a supplier to complete delivery from receiving an order, including production cycle and transportation time, used for spare parts demand forecasting.

[0076] Real-time availability: Whether spare parts are currently available and for how long, taking into account factors such as inventory quantity, logistics in transit, and supplier delivery time.

[0077] Supply forecasting: Based on demand forecasts and supply chain capabilities, predict the timing and quantity of future spare parts availability for inventory management and allocation decisions.

[0078] Furthermore, the dynamic digital twin model consists of three core sub-models: an equipment digital twin sub-model, an engineer digital twin sub-model, and a supply chain digital twin sub-model. The equipment digital twin sub-model addresses the problem of "unclear equipment status," enabling real-time perception of equipment health status and fault prediction, providing a technological foundation for predictive maintenance. The engineer digital twin sub-model addresses the problem of "unclear engineer capabilities," enabling dynamic assessment of engineer skills and task load balancing, improving the accuracy of person-job matching. The supply chain digital twin sub-model addresses the problem of "insufficient spare parts availability," enabling real-time monitoring of spare parts availability and demand forecasting, ensuring timely supply of maintenance resources. These sub-models respectively map to physical equipment entities, human resources, and material resources, and interact through a unified data interface to form a comprehensive simulation system.

[0079] Specifically, the process of constructing each sub-model is as follows:

[0080] The equipment digital twin sub-model collects real-time equipment operation data through IoT sensors (vibration sensors, temperature sensors, pressure sensors, etc.) to form a time-series data stream; it retrieves historical maintenance records (repair time, replaced parts, fault descriptions) from the equipment management system; and it extracts static equipment information (model, manufacturing date, installation location, technical parameters) from the equipment archive. Employing deep learning models such as LSTM networks or Transformers, the input is multi-dimensional time-series data, and the output is an equipment health score (0-100 points) and a predicted remaining service life. A fault map is constructed based on historical fault data, recording the frequency of different fault modes, their correlations, and maintenance plans. This sub-model is used to monitor the equipment's operating status in real time, triggering an alert when the health score falls below a threshold; it predicts future fault time windows based on performance degradation trends, providing a basis for predictive maintenance; and it analyzes equipment weaknesses through historical fault maps to guide maintenance strategy optimization.

[0081] The engineer's digital twin model uses a GPS positioning module to obtain the engineer's real-time location; skill tags (e.g., "Advanced Elevator Repair," "Intermediate Air Conditioning System Repair") are obtained from the HR system; historical work order execution data (completion rate, average response time, customer ratings) is obtained from the work order system; the current task queue (number of pending work orders, estimated completion time) is obtained from the task management module; and maintenance process data (on-site photos, videos, fault descriptions, and replacement part records) is collected from the maintenance terminal. Natural language processing technology is used to analyze the differences between solutions in maintenance records and standard solutions in the knowledge base, identifying the engineer's personalized handling methods; image recognition technology is used to analyze fault characteristic multimedia data, establishing a correspondence between fault phenomena and root causes; and maintenance quality is evaluated based on equipment performance recovery curves to generate an engineer's experience score. An engineer profile matrix is ​​constructed, with dimensions including skill proficiency (0-100 points), task load (current number of tasks / maximum capacity), availability status (idle / busy / offline), and expertise scores for different equipment models and fault modes.

[0082] The supply chain digital twin sub-model connects to warehouse management systems at all levels via API interfaces to obtain real-time spare parts inventory data (inventory quantity and age in central, regional, and field warehouses); it accesses the logistics system to obtain information on spare parts in transit (transportation status and estimated arrival time); and it extracts historical supplier delivery cycle data (average delivery time and reliability score) from the supplier management system. Based on inventory optimization algorithms, it calculates spare parts safety stock and reorder points; considering in-transit logistics and supplier delivery cycles, it predicts spare parts availability time; and it establishes a spare parts demand forecasting model to generate a spare parts demand list based on equipment failure prediction results. This model is used to monitor spare parts inventory status in real time, automatically triggering replenishment when inventory falls below the safety threshold; it provides spare parts availability information for dispatch decisions, preventing engineers from arriving on-site unable to perform repairs due to insufficient spare parts; and it optimizes spare parts allocation routes to reduce logistics costs.

[0083] Therefore, by using real-time health status assessment and fault prediction through equipment digital twin sub-models, potential faults can be identified in advance and early warning work orders can be generated, realizing a shift from passive response to proactive early warning and reducing equipment downtime. Engineer digital twin sub-models provide accurate profiles of engineer capabilities, and by combining real-time location and task load information, appropriate work orders can be assigned to the most suitable engineers, improving the accuracy of person-job matching and reducing response time. Supply chain digital twin sub-models monitor spare parts inventory status and logistics information in real time, providing spare parts availability information for dispatching decisions, preventing engineers from arriving on-site unable to perform repairs due to insufficient spare parts, and improving the first-time repair rate.

[0084] In one embodiment, the extraction of implicit experiential features from the engineer's digital twin sub-model includes:

[0085] Analyze the differences between the final solutions adopted by engineers in their historical maintenance records and the standard solutions in the knowledge base;

[0086] Analyze the correspondence between the fault characteristic multimedia data uploaded by engineers through the terminal during the maintenance process and the associated root causes of the fault;

[0087] Analyze the completeness and stability curves of IoT performance indicators after the equipment has been repaired by engineers;

[0088] Based on all the analysis results, in addition to the basic skill tags for engineers, implicit expertise tags for specific equipment models and specific failure modes are generated for engineers through dynamic weighting.

[0089] Specifically, for each "equipment model-failure mode" combination handled by the engineer, it is denoted as... The study analyzed the differences between the actual maintenance solutions adopted and the standard solutions in the knowledge base.

[0090] 1) If an engineer frequently uses a solution that differs from the standard procedure but ultimately succeeds in fixing the same problem in multiple instances, then they are considered to possess the necessary skills for addressing that specific issue. Special experience.

[0091] Define the engineer in Effective deviation for:

[0092]

[0093] The higher the score, the more skilled the engineer is at solving problems in non-standard but effective ways.

[0094] 2) During the repair process, engineers upload photos or videos of the fault scene via mobile devices. Image recognition models are then used to determine whether this multimedia content contains typical features relevant to the ultimately confirmed root cause of the fault.

[0095] Define its diagnostic accuracy for:

[0096]

[0097] This indicator reflects whether engineers can accurately capture key fault clues and is an important manifestation of implicit diagnostic capabilities.

[0098] 3) After the work order is completed, analyze the completeness and stability curve of the IoT performance indicators restored by the engineer. Continuously monitor the performance of the device's IoT performance indicators (such as operating efficiency, vibration level, energy consumption, etc.) within the next preset time period, usually 24 hours. If the device performance quickly recovers to normal levels and remains stable, the repair quality is considered high. This is the stability indicator. The calculation method is as follows:

[0099]

[0100] 4) Considering the above three indicators, the calculation engineer's... Hidden expertise score :

[0101]

[0102] in, , and Let be the weights of the three indicators, and let the sum of the three indicators equal 1.

[0103] when When the score exceeds a set threshold, such as 0.7, the system automatically adds an implicit expertise tag, for example: "This engineer is proficient in handling [fault mode f] of [equipment model m]". This tag will be updated in real time and used as part of the engineer's digital twin sub-model for skill matching calculations in subsequent dispatch decisions.

[0104] This embodiment achieves an objective assessment of an engineer's implicit maintenance capabilities by quantifying their performance across three dimensions: solution innovation, on-site diagnostic accuracy, and repair result reliability. Compared to traditional static evaluation methods that rely solely on qualification certificates or the number of work orders, this approach establishes a more accurate profile of the engineer's capabilities, thereby improving the first-time repair rate. Specifically, it prioritizes matching engineers with relevant implicit expertise when dispatching work orders, reducing rework. Furthermore, as new work order data accumulates, implicit tags are dynamically updated, correspondingly enhancing the intelligence of the dispatching system. Therefore, this method, without increasing the engineer's workload, fully utilizes existing maintenance data to automate the mining and application of implicit experience, significantly improving the intelligence level of the after-sales service system and customer satisfaction.

[0105] In one embodiment, prior to responding to a received user repair order or an alert order, the method further includes an automated order classification step:

[0106] For user repair work orders, based on the repair text description and the real-time abnormal data stream provided by the digital twin sub-model of the equipment, a multimodal fusion classification model is used to output classification labels containing the root cause of the fault, the required skills, and key spare parts.

[0107] For early warning work orders, classification labels are generated based on the results of collaborative anomaly analysis.

[0108] In the traditional model, customer service representatives or dispatchers need to manually determine the fault type, required skills, and spare parts based on user descriptions and equipment information, which is time-consuming and prone to errors. Inconsistent judgment standards among different personnel for the same fault lead to chaotic work order classification, affecting subsequent dispatching and statistical analysis. Therefore, this embodiment aims to use automated classification technology to accurately "identify" each work order before it enters the dispatching decision-making process, solving the problems of low efficiency, poor accuracy, and inconsistent standards in traditional manual classification.

[0109] Repair request text description: The text describing the fault submitted by the user through channels such as telephone, APP, WeChat, etc., such as "The equipment makes abnormal noises and cannot operate normally".

[0110] Real-time abnormal data stream: Abnormal segments identified in the time-series data collected by the device's sensors, including information such as the type of abnormality, severity, and duration.

[0111] Root cause of failure: The fundamental reason for equipment failure, such as "motor bearing wear" or "control board failure", which is different from the surface phenomenon.

[0112] Key spare parts: The core spare parts most likely to be needed to repair this fault, such as "inverter module" and "sensor probe".

[0113] Specifically, for user repair work orders, the first step is to process the user repair text: this includes using natural language processing techniques to segment the repair description, remove stop words, and perform entity recognition; extracting key information such as equipment model, fault symptoms, occurrence time, and affected area; and constructing text feature vectors, such as TF-IDF and word embedding. Then, real-time time-series data (vibration, temperature, current, etc.) of the target equipment is obtained from the equipment's digital twin model. Abnormal features are extracted: abnormal occurrence time, duration, abnormal amplitude, and associated sensors; and a time-series feature vector is constructed. Based on timestamps, the repair text is time-aligned with the equipment abnormal data to establish a correspondence between the text description and the equipment abnormality pattern.

[0114] A multimodal fusion classification model typically employs a dual-tower structure: a text encoder (such as TextCNN) processes the repair request text, while a temporal encoder (such as LSTM or Transformer) processes the equipment anomaly data. Multimodal fusion is achieved through attention mechanisms or feature concatenation. The output layer uses a softmax classifier to output classification labels across three dimensions: root cause of the fault, required skills, and key spare parts. Historical work order data is used as the training set, including repair request text, equipment anomaly data, actual root causes of the fault, and final repair solutions. Multi-task learning is employed, simultaneously optimizing the loss functions for the three classification tasks. Overfitting is prevented through cross-validation and early stop strategies. After model training, the model is applied. For user repair request classification, upon receiving a work order, the classification model is automatically triggered. The model input typically consists of the repair request text and real-time equipment anomaly data. The model output includes a root cause label (e.g., "motor bearing wear"), a required skill label (e.g., "advanced mechanical repair"), and a key spare parts label (e.g., "bearing model XXX"). A confidence threshold is then set; when the confidence level falls below the threshold, the work order is marked as "pending manual review." As for the classification of early warning work orders, these orders are automatically generated by collaborative anomaly analysis and already contain fault type information. They are directly mapped to the standard classification label system based on the collaborative anomaly analysis results. For example: abnormal vibration of group equipment → root cause of fault (transmission system fault) → required skills (mechanical maintenance) → critical spare parts (drive shaft).

[0115] After classifying the two types of work orders, the classification labels serve as input features for the work order optimization model, the root cause labels are used to match engineer skill specialties, and the key spare parts labels are used for spare parts demand forecasting and inventory preparation. To continuously optimize the model, actual repair results (final root cause of failure, actual spare parts used, and repair effectiveness) are collected; the actual results are compared with the classification predictions as new samples for model training. Model parameters are updated regularly to improve classification accuracy.

[0116] Through the above embodiments, work order classification is automated, reducing manual intervention and significantly improving classification speed. The multimodal fusion model comprehensively utilizes text descriptions and equipment operation data, achieving a classification accuracy significantly higher than that of a single information source. Based on a unified model and tagging system, it ensures consistent classification results for the same fault across different times and personnel. Fault root causes and required skill tags provide accurate matching criteria for dispatching, avoiding secondary dispatches due to skill mismatches and improving service quality. Simultaneously, the classification model can quickly identify fault types, shorten fault diagnosis time, and greatly improve service response efficiency.

[0117] In one embodiment, the multi-resource collaborative optimization of dispatch decision-making specifically includes:

[0118] To minimize the overall system response cost and risk, a multi-objective optimization model is constructed and solved.

[0119] The decision variables of the optimization model include the matching relationship between work orders and engineers, and the supply path of required spare parts; the constraints of the optimization model are derived from the dynamic parameters of the dynamic digital twin model.

[0120] Solve the multi-objective optimization model and output a dispatch plan that includes a specified engineer, a recommended spare parts retrieval plan, and a suggested execution sequence.

[0121] Preferably, the overall system response cost includes:

[0122] Explicit costs: Engineer travel costs, time costs, spare parts logistics costs;

[0123] Hidden risk costs include: downtime losses predicted by the equipment's digital twin model, backlog of work orders due to task delays, and maintenance failure risk costs due to skill mismatch.

[0124] The dynamic parameters include:

[0125] The system includes the engineer's real-time location and future availability window, the predicted rate of equipment failure deterioration, the real-time inventory and lock-up status of spare parts, and the dynamic skill matching degree of the engineer for the root cause of the current work order failure.

[0126] Specifically, the decision variables in this embodiment include:

[0127] : Indicates whether to submit the work order Assigned to engineers ;

[0128] : indicates a work order Are the required spare parts available from a warehouse or supply chain? supply;

[0129] The optimization objective is to minimize the weighted comprehensive cost function. Its expression is:

[0130] ;

[0131] :engineer Go to work order The estimated travel cost for the location is calculated based on real-time map navigation distance, traffic conditions, and cost per unit mileage.

[0132] The estimated service time is predicted by the device's digital twin sub-model based on the current device status, fault type, and historical similar work order data.

[0133] :engineer The unit time labor cost;

[0134] From the supply path Transfer work order The total cost of the required spare parts, including warehousing, transportation and emergency procurement premiums;

[0135] The value range is [0,1]. (Engineer) For work orders The dynamic skill matching degree is generated in real time by the engineer's digital twin sub-model based on the matching degree between implicit expertise tags and root causes of failures;

[0136] :satisfy Work order The risk weighting coefficient is determined by the economic losses from equipment downtime, the customer's SLA level, and the scope of abnormal propagation in the equipment group.

[0137] The model satisfies the following constraints:

[0138] 1) Only one engineer will be assigned to each work order: ;

[0139] 2) Spare parts requirements for each work order are met via a single path: ;

[0140] 3) The total time spent on an engineer's tasks does not exceed its available time window:

[0141] ;

[0142] In the formula, For engineers The earliest available end time, For the current moment, For engineers Go to the work order from your current location or the location where the previous task ended. The estimated commute time required to reach the location.

[0143] 4) Spare parts allocation shall not exceed the dynamic available inventory: ;

[0144] In the formula, For work orders Required number of spare parts Its expected execution time, For warehouse The available inventory at that moment is predicted in real time by the supply chain digital twin model;

[0145] 5) Dispatch low-skilled players to orders:

[0146] hour, ;

[0147] In the formula, The minimum skill matching threshold preset by the system;

[0148] After establishing the constraints, real-time parameters are extracted from three digital twin sub-models: equipment, engineers, and supply chain. These parameters are initialized, and a mixed-integer linear programming (MILP) model is constructed. An optimization solver is then invoked to solve the problem within a preset time limit, yielding the optimal decision variables. and Finally, based on the optimal solution, a dispatch plan is output, including the designated engineer, spare parts source warehouse, suggested departure time, and execution order.

[0149] In this embodiment, a multi-objective optimization model is constructed with the goal of minimizing the overall system response cost and risk. The overall system response cost includes not only explicit costs such as engineer travel, time, and spare parts logistics, but also implicit costs such as downtime losses predicted by the equipment digital twin sub-model, work order backlog caused by task delays, and maintenance failure risks due to skill mismatch. This achieves a comprehensive trade-off between cost and risk, improving the scientific nature of decision-making. Dynamic parameters include the engineer's real-time location and available time window, the predicted speed of equipment failure deterioration, the real-time inventory and lockout status of spare parts, and the dynamic skill matching degree of the engineer for the root cause of the current work order failure. These provide real-time and accurate constraints for the multi-objective optimization model, ensuring the executability and optimality of the dispatching plan.

[0150] See Figure 2 In one embodiment, the present invention also provides an automatic classification and dispatch optimization system for after-sales work orders based on Internet of Things (IoT) analysis, the system comprising:

[0151] The model building unit 100 is used to build a dynamic digital twin model of after-sales work order service. The dynamic digital twin model is used to represent the equipment status, service resource status and material supply chain status.

[0152] The anomaly analysis unit 200 is used to perform collaborative anomaly analysis on the equipment group based on the equipment status data in the dynamic digital twin model. When a collaborative anomaly is detected in the equipment group that exceeds a preset threshold, an early warning work order is automatically generated and a predictive spare parts demand list is marked on the early warning work order.

[0153] The work order response unit 300 is used to respond to the received user repair work order or the early warning work order, and to make multi-resource collaborative optimization dispatch decision based on the dynamic constraint parameters provided by the dynamic digital twin model and the predictive spare parts demand list as input.

[0154] The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list.

[0155] The model update unit 400 is used to collect feedback data after performing services according to the dispatch scheme, and to feed the feedback data back to the dynamic digital twin model to drive model updates.

[0156] It is understood that the system provided in this embodiment has functions or includes modules that can be used to execute the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0157] The present invention also provides an electronic device including a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device performs a method as described in any of the above possible implementations.

[0158] The present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor of an electronic device, cause the processor to perform a method as described in any of the above possible implementations.

[0159] Those skilled in the art will 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, or a combination of computer software and electronic hardware. 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 implementations should not be considered beyond the scope of this invention.

Claims

1. A method for automatically classifying and optimizing after-sales work orders based on Internet of Things (IoT) analytics, characterized in that: The method includes: Construct a dynamic digital twin model for after-sales work order services. The dynamic digital twin model is used to represent the status of equipment, service resources, and material supply chain. Based on the equipment status data in the dynamic digital twin model, a collaborative anomaly analysis is performed on the equipment group. When a collaborative anomaly is detected in the equipment group that exceeds a preset threshold, an early warning work order is automatically generated, and a predictive spare parts demand list is marked on the early warning work order. In response to the received user repair work order or the warning work order, based on the dynamic constraint parameters provided by the dynamic digital twin model, the predictive spare parts demand list is used as input to perform multi-resource collaborative optimization dispatch decision-making; The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list. After the service is executed according to the dispatch plan, feedback data is collected and fed back to the dynamic digital twin model to drive model updates.

2. The method for automatic classification and dispatch optimization of after-sales work orders based on IoT analysis according to claim 1, characterized in that, The dynamic digital twin model of the after-sales work order service includes: The device digital twin sub-model is constructed based on the target device's real-time IoT time-series data, historical maintenance data, and static device archives. It is used to characterize the device's real-time health status, performance degradation trend, and historical fault map. The engineer digital twin sub-model is built based on the engineer's real-time location, skill tags, historical work order execution data, task queue, and implicit experience features extracted from maintenance process data. It is used to dynamically evaluate the engineer's availability, skill proficiency, and task load. The supply chain digital twin sub-model is built based on dynamic inventory data of spare parts in warehouses at all levels, in-transit logistics information and supplier delivery cycle data, and is used to characterize the real-time availability and replenishment expectations of spare parts.

3. The method for automatic classification and dispatch optimization of after-sales work orders based on IoT analysis according to claim 2, characterized in that, The methods for extracting implicit empirical features from the engineer's digital twin model include: Analyze the differences between the final solutions adopted by engineers in their historical maintenance records and the standard solutions in the knowledge base; Analyze the correspondence between the fault characteristic multimedia data uploaded by engineers through the terminal during the maintenance process and the associated root causes of the fault; Analyze the completeness and stability curves of IoT performance indicators after the equipment has been repaired by engineers; Based on all the analysis results, in addition to the basic skill tags for engineers, implicit expertise tags for specific equipment models and specific failure modes are generated for engineers through dynamic weighting.

4. The method for automatic classification and dispatch optimization of after-sales work orders based on IoT analysis according to claim 2, characterized in that, Before responding to the received user repair work order or the warning work order, the method further includes an automated work order classification step: For user repair work orders, based on the repair text description and the real-time abnormal data stream provided by the digital twin sub-model of the equipment, a multimodal fusion classification model is used to output classification labels containing the root cause of the fault, the required skills, and key spare parts. For early warning work orders, classification labels are generated based on the results of collaborative anomaly analysis.

5. The method for automatic classification and dispatch optimization of after-sales work orders based on IoT analysis according to claim 1, characterized in that, The multi-resource collaborative optimization of order dispatch decision-making is specifically as follows: To minimize the overall system response cost and risk, a multi-objective optimization model is constructed and solved. The decision variables of the optimization model include the matching relationship between work orders and engineers, and the supply path of required spare parts; the constraints of the optimization model are derived from the dynamic parameters of the dynamic digital twin model. Solve the multi-objective optimization model and output a dispatch plan that includes a specified engineer, a recommended spare parts retrieval plan, and a suggested execution sequence.

6. The method for automatic classification and dispatch optimization of after-sales work orders based on Internet of Things analysis according to claim 5, characterized in that, The overall system response cost includes: Explicit costs: Engineer travel costs, time costs, spare parts logistics costs; Hidden risk costs include: downtime losses predicted by the equipment's digital twin model, backlog of work orders due to task delays, and maintenance failure risk costs due to skill mismatch.

7. The method for automatic classification and dispatch optimization of after-sales work orders based on Internet of Things analysis according to claim 5, characterized in that, The dynamic parameters include: The system includes the engineer's real-time location and future availability window, the predicted rate of equipment failure deterioration, the real-time inventory and lock-up status of spare parts, and the dynamic skill matching degree of the engineer for the root cause of the current work order failure.

8. A system for automatically classifying and optimizing after-sales work orders based on Internet of Things (IoT) analytics, characterized in that: The system includes: The model building unit is used to build a dynamic digital twin model of after-sales work order service. The dynamic digital twin model is used to represent the equipment status, service resource status and material supply chain status. An anomaly analysis unit is used to perform collaborative anomaly analysis on equipment groups based on equipment status data in the dynamic digital twin model. When a collaborative anomaly is detected in equipment within the group that exceeds a preset threshold, an early warning work order is automatically generated, and a predictive spare parts demand list is marked on the early warning work order. The work order response unit is used to respond to the received user repair work order or the warning work order, and to make multi-resource collaborative optimization dispatch decisions based on the dynamic constraint parameters provided by the dynamic digital twin model and the predictive spare parts demand list as input. The multiple resources include at least engineer resources and spare parts resources. The dispatch decision is used to ensure that when an engineer performs a task, the spare parts resources specified in the dispatch plan match or can be allocated synchronously with the predictive spare parts demand list. The model update unit is used to collect feedback data after performing services according to the dispatch scheme, and to feed the feedback data back to the dynamic digital twin model to drive model updates.

9. An electronic device, characterized in that, include: A processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device executes the after-sales work order automatic classification and dispatch optimization method based on Internet of Things analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which includes program instructions that, when executed by a processor of an electronic device, cause the processor to perform the after-sales work order automatic classification and dispatch optimization method based on Internet of Things analysis as described in any one of claims 1 to 7.

Images