Intelligent analysis and service optimization system for campus informatization maintenance work order

Abstract

The invention relates to the technical field of intelligent processing of campus operation and maintenance work orders, and discloses an intelligent analysis and service optimization system for campus informatization maintenance work orders. According to the system, standardization processing of multi-source work order data is achieved through a work order information aggregation module; the intelligent priority evaluation module adopts a deep learning model to analyze fault text semantics, generates a dynamic priority coefficient and improves the accuracy of work order grading; the resource scheduling optimization module generates an optimal allocation scheme based on the priority and the personnel state; and the redundant communication management module constructs an active-active architecture by using a software defined network, so that load balancing and rapid fault switching of a data stream and a control signaling are realized, and the communication reliability of the system is effectively guaranteed. According to the invention, the problems of inaccurate priority judgment and insufficient communication guarantee of a traditional work order system are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of campus operation order intelligent processing, in particular to a campus information maintenance order intelligent analysis and service optimization system. BACKGROUND

[0002] In the campus information operation and maintenance management, the processing efficiency of the maintenance order directly determines the quality and stability of the campus information service. The existing order processing system generally adopts a priority judgment mechanism based on predefined rules. This static rule library method is difficult to cope with complex and variable actual fault scenarios and cannot understand the context semantics of the fault description. The rule engine may not effectively distinguish between single user terminal problems, floor switch failures or campus network backbone link congestion, resulting in priority misjudgment, causing critical order response delay and secondary order occupying limited maintenance resources.

[0003] The reliability design of the underlying communication architecture of the existing system has limitations. Most systems use master-slave server switching or simple network link redundancy, and the focus of their protection is the reliability of data storage, and there is a lack of effective protection for the transmission quality of control signaling generated in real time during the order processing process. The delay or loss of control signaling will cause the resource scheduling instructions to be unable to be issued in time and accurately, and the maintenance personnel state update will be out of sync, thereby causing scheduling conflicts or resource idling. When the network fluctuates locally, the system is difficult to realize the rapid and lossless path switching of the order data stream and the control signaling, and the real-time and determinacy of the entire scheduling process cannot be guaranteed.

[0004] The current campus information maintenance order system faces two core challenges: how to break through the limitations of the rule engine and realize intelligent understanding and accurate priority evaluation of the order content semantics; how to build a highly reliable communication bottom layer to ensure that intelligent scheduling instructions can be executed in real time and reliably, and avoid the decline of the overall efficiency of the system due to communication problems. SUMMARY

[0005] The purpose of the present application is to provide a campus information maintenance order intelligent analysis and service optimization system to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a campus information maintenance order intelligent analysis and service optimization system, which comprises: an order information aggregation module, an intelligent priority evaluation module, a resource scheduling optimization module and a redundant communication management module; The order information aggregation module is used to asynchronously collect maintenance order data from multiple campus information subsystems and perform multi-source fusion processing on the order data to generate standardized order records. The intelligent priority evaluation module is used for analyzing fault text description and equipment type in standardized work order records through a deep learning model, automatically generating a work order feature vector, and calculating a dynamic priority coefficient based on the feature vector; The resource scheduling optimization module is used for generating a multi-objective resource allocation scheme according to the dynamic priority coefficient and a real-time maintenance personnel state, and executing a resource scheduling instruction; The redundant communication management module is used for constructing a dual-active communication architecture based on a software-defined network, and realizing load balancing and fault switching of work order data flow and control signaling.

[0007] Preferably, the work order information aggregation module asynchronously collects maintenance work order data from multiple campus informatization subsystems, specifically: Distributed data collection agents are deployed, each agent monitors work order generation events of a specific informatization subsystem, and captures work order raw data through an event-driven mechanism; Semantic analysis and field mapping are performed on the work order raw data to eliminate data mode differences between different subsystems, and a consistent hash algorithm is applied to distributed deduplication of the work order; The deduplicated work order data is encapsulated into messages in a unified data format, and is asynchronously transmitted to a central data warehouse for persistent storage through a message queue.

[0008] Preferably, the intelligent priority evaluation module analyzes fault text description and equipment type in standardized work order records through a deep learning model, specifically: A sequence-to-sequence model containing an encoder-decoder structure is constructed, the encoder converts the fault text description into a context-aware semantic embedding vector; The device type data is processed using a graph convolution network to extract device-related topological features and generate a device topology embedding vector; The semantic embedding vector and the device topology embedding vector are spliced, and the fault influence factor and the repair urgency factor are calculated through a fully connected neural network to generate a work order feature vector.

[0009] Preferably, the method for calculating a dynamic priority coefficient based on a feature vector includes: The work order feature vector is input into a time recurrent neural network with an attention mechanism to dynamically learn the weight distribution of the work order features over time; External environmental parameters including campus activity schedules and device usage frequency are introduced, and an environmental influence correction value is calculated through a feature cross layer; The weight distribution and the environmental influence correction value are point multiplied to output a dynamic priority coefficient, the dynamic priority coefficient is a continuous numerical value used to represent the priority of work order processing.

[0010] Preferably, the resource scheduling optimization module generates a multi-objective resource allocation scheme based on dynamic priority coefficients and real-time maintenance personnel states, specifically: Establish a maintenance personnel state profile, including a skill matrix, current location coordinates, and a workload index, and update the state profile in real time; Use the dynamic priority coefficient as the weight to construct a resource allocation optimization objective function, and the optimization objectives include minimizing the average response time and maximizing the resource utilization rate; Use a genetic algorithm to solve the objective function to generate a resource allocation scheme that meets multiple constraint conditions, including a skill matching threshold and a geographic distance upper limit.

[0011] Preferably, the method for executing resource scheduling instructions in the resource scheduling optimization module includes: Encode the resource allocation scheme into a machine-readable scheduling instruction set, which includes work order assignment objects, estimated processing time, and optimal path planning; Distribute the scheduling instruction set to the maintenance personnel mobile terminal through the application programming interface, and activate the instruction execution state machine; Real-time monitoring of instruction execution progress, when detecting work order state changes or maintenance personnel location updates, triggering resource allocation scheme recalculation process.

[0012] Preferably, the redundant communication management module constructs a dual-active communication architecture based on software-defined networks, specifically: Deploy a software-defined network controller in the campus network infrastructure, which dynamically manages the data plane of two independent physical links; Assign a main communication path to the work order data stream, which uses multi-protocol label switching technology to ensure transmission quality; Assign a backup communication path to the control signaling, which is based on the user datagram protocol to achieve low-latency communication and maintain path availability through a heartbeat detection mechanism.

[0013] Preferably, the method for implementing load balancing and fault switching of work order data streams and control signaling in the redundant communication management module includes: Integrate a load balancing algorithm in the software-defined network controller, which dynamically adjusts packet distribution strategies based on real-time network traffic statistics; Set a path health index, when the index of the main communication path is below the threshold, the controller automatically switches the work order data stream to the backup communication path; The control signaling uses a multi-copy synchronization mechanism to ensure the continuity and consistency of signaling transmission during path switching.

[0014] Preferably, the intelligent priority evaluation module is also used to train a fault prediction model based on historical work order data; The failure prediction model adopts a long short-term memory network structure, and input sequence data includes device operation logs, maintenance records and environmental sensor readings. The model output is a device failure probability distribution, which is used to correct the calculation process of the dynamic priority coefficient.

[0015] Preferably, the resource scheduling optimization module is also used to dynamically adjust the resource allocation strategy according to the work order processing feedback. Real-time acquisition of work order processing full-link data, including maintenance operation records, resource consumption indicators and user evaluation information; Using online learning algorithm to analyze full-link data, dynamically updating parameter weight in resource allocation optimization objective function, generating adaptive resource scheduling strategy.

[0016] Compared with the prior art, the beneficial effects of the present application are: The deep learning model is used to analyze the semantic analysis of the unstructured text description in the work order, and the high-dimensional work order feature vector is generated combined with the device type information. The core is to understand the deep semantic association of natural language by using neural network model, instead of relying on surface keyword matching. Through training, the model can identify the complexity of the severity, impact range and urgency implied in the fault description. The calculation of the dynamic priority coefficient is a continuous learning and adjustment process, and the model weight is optimized according to the processing results and effectiveness feedback of historical work orders. This makes the priority evaluation no longer a set of fixed rules, but an intelligent system that can adapt to new fault types and changes in campus operation and maintenance environment. The accuracy of priority judgment is improved qualitatively, and the system can effectively distinguish between superficially similar but essentially different faults, so as to guide the maintenance resources to the truly critical tasks.

[0017] A dedicated dual-active communication architecture based on software-defined network is constructed, and the data plane and control plane of the work order system are separated and uniformly managed. The centralized controller obtains the global network state view in real time, and dynamically allocates the optimal transmission path for the work order data flow and control signaling according to the preset strategy. It not only realizes the on-demand allocation and load balancing of network bandwidth resources, but more importantly, provides high-priority and exclusive channel protection for scheduling instructions, state synchronization and other control signaling. When detecting network node or link failure, the controller can recalculate and issue flow table within milliseconds to realize seamless switching of business flow and signaling flow. Eliminate the problems of scheduling instruction loss and state inconsistency caused by communication interruption or delay, provide stable and reliable transmission protection for upper-layer intelligent scheduling decision, and enhance the response speed and robustness of the whole system. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The working principle diagram of the campus informationization maintenance work order intelligent analysis and service optimization system of the present application; Figure 2 Flow chart for data collection and processing of work order information aggregation module; Figure 3 Flow chart for feature extraction of intelligent priority evaluation module; Figure 4 Skill matching degree and actual processing time regression analysis chart; Figure 5 Dual-active communication path bandwidth utilization rate and load balancing weight trend chart. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0020] Please refer to Figure 1 The present application provides a campus information maintenance work order intelligent analysis and service optimization system, which comprises a work order information aggregation module, an intelligent priority evaluation module, a resource scheduling optimization module and a redundant communication management module.

[0021] The work order information aggregation module asynchronously collects maintenance work order data from multiple campus information subsystems, and performs multi-source fusion processing on the work order data to generate standardized work order records. The intelligent priority evaluation module analyzes the fault text description and device type in the standardized work order records through a deep learning model, automatically generates a work order feature vector, and calculates a dynamic priority coefficient based on the feature vector. The resource scheduling optimization module generates a multi-objective resource allocation scheme according to the dynamic priority coefficient and the real-time maintenance personnel state, and executes resource scheduling instructions. The redundant communication management module constructs a dual-active communication architecture based on software-defined networks to achieve load balancing and fault switching of work order data flow and control signaling.

[0022] Embodiment 1: refer to Figure 2In a specific implementation, deploying distributed data collection agents is the basis for realizing data collection, each of which is configured to monitor ticket generation events of a specific informationized subsystem, which can be triggered by a workflow engine or user operation within the subsystem, and the distributed data collection agent captures ticket raw data in real time through an event-driven mechanism, which relies on the application programming interface callback or database transaction log monitoring technology of the subsystem. In some embodiments, the distributed data collection agent runs independently in a microservice architecture, interacts with the subsystem through a lightweight communication protocol, and ensures low invasiveness and high scalability of the collection process. Optionally, the event-driven mechanism can implement event publishing and subscribing based on a message middleware, thereby further improving the decoupling capability and reliability of the system.

[0023] Semantic parsing utilizes natural language processing techniques to analyze the work order description text, identifying and extracting key information entities such as faulty equipment and problem symptoms. Field mapping, through predefined mapping rules, transforms the data patterns of different subsystems into a unified data model, eliminating pattern differences. A consistent hashing algorithm is applied for distributed deduplication of work orders. This deduplication operation occurs after the original work order data has undergone semantic parsing and field mapping, at which point the work order data has been transformed into a unified data model. A deduplication key is generated for each standardized work order record, calculated based on a combination of the work order's core fields, such as the work order number, faulty equipment identifier, and work order creation timestamp, ensuring that the same work order has the same deduplication key across different subsystems. The consistent hashing algorithm organizes the entire hash value space into a virtual ring structure, dividing the hash ring into multiple consecutive intervals, each corresponding to a distributed node responsible for deduplication. The hash value of the deduplication key is calculated and mapped to a specific position on the hash ring. Based on the mapping position, the target node responsible for deduplication checking for that work order is determined. In some embodiments, the hash calculation of the deduplication key can use hash functions such as MD5 or SHA-1 to generate a fixed-length hash value. Optionally, the generation of the deduplication key can ignore some variable fields, such as non-keywords in the work order description text, to improve deduplication accuracy. After a work order is mapped to a target node on the hash ring, the node queries its locally maintained processed work order index, which stores the deduplication key hash values ​​of recently processed work orders. If the same deduplication key hash value already exists, the current work order is determined to be a duplicate record, and a deduplication operation is performed, discarding the newly arrived work order data or merging it with an existing work order record. It can be understood that the introduction of the consistent hashing algorithm allows the deduplication task to be distributed across multiple nodes for parallel processing, improving system throughput. In specific implementations, nodes on the hash ring typically correspond to different data collection agents or dedicated deduplication service instances, with each node responsible for the work order deduplication task for a certain interval on the ring. The storage of work order deduplication key hash values ​​in the node index can be set with an expiration time, automatically cleaning up historical data to control the storage size. In some embodiments, the consistent hashing algorithm achieves better load balancing through virtual node technology, mapping a physical node to multiple virtual nodes on the ring, thereby avoiding large-scale data remapping caused by changes in the number of nodes. Optionally, the deduplication judgment logic can be combined with a time window, such as comparing only work orders processed in the past 24 hours, to balance the deduplication effect and system overhead.

[0024] The de-duplicated work order data is encapsulated into a message in a unified data format, which adopts a structured data mode and contains fixed fields such as work order number, fault type and timestamp. The encapsulation process includes data serialization and metadata addition. The message is transmitted asynchronously to the central data warehouse through a message queue, which adopts a publish-subscribe mode and supports parallel processing of multiple consumers to ensure high throughput and fault tolerance of data transmission. Optionally, the message queue can implement persistent storage and retry mechanism to prevent data loss. The central data warehouse uses a distributed database system for persistent storage, providing efficient query and data management capabilities. It can be understood that the asynchronous transmission mechanism reduces the coupling between system components and improves overall maintainability. In specific implementation, the message format can choose JavaScript Object Notation or Protocol Buffer to adapt to the performance requirements in different scenarios.

[0025] Embodiment 2: see Figure 3 In specific implementation, building a sequence-to-sequence model containing an encoder-decoder structure is the core of parsing fault text description. The encoder part usually adopts a bidirectional long short-term memory network to process the input fault text description word sequence step by step and finally convert it into a semantic embedding vector containing context information. The decoder part can be configured according to actual task requirements, such as work order classification or summary generation. In some embodiments, the output state of the encoder will be weighted and summarized through an attention layer to highlight the influence of key words in the fault description on the final semantic representation. Optionally, the sequence-to-sequence model can be pre-trained on a large-scale campus work order corpus to obtain higher-quality semantic embedding vectors.

[0026] A graph convolutional network is used to process device type data. The graph convolutional network abstracts the information devices in the campus and their connection relationships into a graph structure. The nodes in the graph represent specific devices and are attached with device type attributes, and the edges represent the physical or logical connection relationships between devices. The graph convolutional network aggregates the feature information of adjacent nodes by performing convolution operations on the adjacency matrix of the graph, thereby extracting device association topology features and generating device topology embedding vectors. It can be understood that the graph convolutional network can effectively capture the potential propagation risk of device faults, for example, the failure of a core switch may affect multiple access layer switches connected to it. The semantic embedding vector and the device topology embedding vector are spliced to form a composite vector representation that integrates fault semantics and device topology. The composite vector is then input into a fully connected neural network. The fully connected neural network calculates the fault influence factor and the repair urgency factor through multiple layers of nonlinear transformation, and finally fuses to generate a work order feature vector. The fault influence factor quantifies the range and severity of the possible impact of the fault, and the repair urgency factor reflects the urgency of the problem to be solved.

[0027] The intelligent priority evaluation module also trains a failure prediction model based on historical work order data. The failure prediction model adopts a long short-term memory network structure. The long short-term memory network effectively learns and captures long-term dependency patterns in time series data such as device operation logs, maintenance records, and environmental sensor readings through its internal gating mechanism. The model output is a device failure probability distribution, which is a multi-dimensional vector representing the likelihood of various types of failure events occurring within a specific time period in the future. The device failure probability distribution is used to modify the calculation process of the dynamic priority coefficient. For example, when the failure prediction model outputs that a certain device has a high failure probability, even if the current work order feature vector indicates a low priority, the dynamic priority coefficient may be adjusted accordingly to achieve preventive maintenance. In specific implementations, the calculation of the dynamic priority coefficient can introduce a correction term based on the failure probability. The calculation formula of the correction term is as follows:

[0028] wherein: represents the priority correction value, represents the total number of failure types, represents the preset weight coefficient of the type of failure, represents the probability of the type of failure occurring output by the failure prediction model. It can be understood that by introducing the failure prediction model, the system can shift from passive response to active early warning, optimizing the configuration efficiency of maintenance resources. In some embodiments, the training data of the failure prediction model needs to be strictly cleaned and labeled to ensure that the time series data accurately corresponds to the subsequent actual failure work orders. Optionally, the long short-term memory network can be stacked with multiple layers to enhance its modeling capability, and techniques such as dropout can be used to prevent overfitting.

[0029] In some embodiments, the work order feature vector is input into a time recurrent neural network with attention mechanism, which is a key step to calculate the dynamic priority coefficient. The time recurrent neural network with attention mechanism can analyze the time dependence of the work order feature vector sequence, and the attention mechanism assigns dynamic weights to the features of different time steps, thereby capturing the law of the evolution of the work order priority over time, and the hidden state of the time recurrent neural network records the influence of historical work order features. External environmental parameters are introduced, including campus activity schedules and device usage frequency. The campus activity schedule data is derived from the school's academic and activity management systems, and the device usage frequency data is obtained from device monitoring logs. The environmental impact correction value is calculated through a feature cross layer, which interacts the work order features with external environmental parameters, such as calculating the combined effect of a specific device malfunction during a large event. In some embodiments, the feature cross layer can use a tensor product or a factorization machine model to construct the cross relationship between features. Optionally, the external environmental parameters can also include weather data, academic calendar, and other more dimensional information.

[0030] The weight distribution is point multiplied with the environmental impact correction value. The point multiplication operation multiplies the time weights output by the time recurrent neural network with attention mechanism and the environmental impact correction value calculated by the feature cross layer element by element, obtaining an adjusted weight that integrates the time sequence dynamics and external environmental impact. The dynamic priority coefficient is output, which is a continuous numerical value representing the priority of work order processing. It can be understood that this calculation method enables the work order priority to respond to changes in campus operating status in real time. The resource scheduling optimization module also dynamically adjusts the resource allocation strategy based on work order processing feedback, and collects real-time work order processing full-link data. The full-link data covers the complete life cycle from work order creation, assignment, maintenance to closure. The maintenance operation record details the specific steps performed by the maintenance personnel and the spare parts used. The resource consumption index quantifies the time, manpower, and material costs. User evaluation information is obtained through a satisfaction survey questionnaire after the work order is closed. Online learning algorithms are used to analyze the full-link data. The online learning algorithm processes newly arrived work order feedback data in an incremental update manner, dynamically updates the parameter weights in the resource allocation optimization objective function, and generates an adaptive resource scheduling strategy. In some embodiments, the online learning algorithm can use a stochastic gradient descent method or an adaptive moment estimation algorithm to minimize a loss function defined on the feedback data. The loss function L can be represented as:

[0031] wherein: represents the loss function, represents the number of work order feedback samples in the current batch, represents the actual processing efficiency score of the th sample, a prediction score of the resource scheduling model for the a work order feature and a resource allocation scheme, a parameter weight to be updated by the model, a regularization coefficient to control the complexity of the model. It can be understood that through continuous learning of work order processing feedback, the resource allocation strategy can be continuously self-optimized to better adapt to actual operation and maintenance scenarios. Optionally, the collection of full-link data can be realized automatically by integrating data collection points at each node of the work order flow.

[0032] In a specific implementation, the establishment of the maintenance personnel state portrait is the basis for generating the resource allocation scheme, and the maintenance personnel state portrait is dynamically constructed by integrating multiple data sources, including obtaining a skill matrix from a human resource system, collecting current location coordinates through a global positioning system module, and calculating a work load index from a work order processing system. The skill matrix stores the skill types and proficiency levels mastered by each maintenance personnel in the form of a two-dimensional table, the current location coordinates are updated in real time in the form of longitude and latitude data, and the work load index is calculated based on the number of unfinished work orders of the maintenance personnel and the weighted calculation of the estimated processing time. The maintenance personnel state portrait is continuously updated by receiving real-time event streams through a message queue. In some embodiments, the maintenance personnel state portrait can be stored in an in-memory database to realize high-speed read-write access, supporting the low-latency requirements of the resource scheduling process. Optionally, the update of the skill matrix can be automatically synchronized through regular training records or certification results.

[0033] With the dynamic priority coefficient as the weight, a resource allocation optimization objective function is constructed, and the optimization objectives include minimizing the average response time and maximizing the resource utilization rate. The average response time refers to the time interval from the generation of a work order to the arrival of a maintenance personnel at the scene, and the resource utilization rate is defined as the ratio of the effective working time of the maintenance personnel to the total working hours. The role of the resource allocation optimization objective function is to provide mathematical optimization criteria for the resource scheduling optimization module. The resource allocation optimization objective function converts the dynamic priority coefficient of the work order, the maintenance personnel state, and the system operation target into a quantifiable mathematical expression, so that the genetic algorithm can perform directional search in the vast possible solution space based on this expression. The value of the resource allocation optimization objective function serves as a unified scale for evaluating the pros and cons of different resource allocation schemes, and the allocation scheme with a smaller function value is considered to be better. The resource allocation optimization objective function fuses multiple optimization objectives, i.e., minimizing the average response time and maximizing the resource utilization rate, and introduces the dynamic priority coefficient as a weight factor, ensuring that high-priority work orders obtain greater weight in resource allocation, thereby guiding the system to prioritize processing of faults that have a greater impact on campus informatization services.

[0034] ​The specific form of the resource allocation optimization objective function defines the mathematical essence of the optimization problem. The weight coefficients a and b in the function allow the system administrator to adjust the relative importance of different objectives according to the actual operation strategy, for example, a higher weight can be given to the average response time to ensure service during the peak period of the school opening, and the weight of resource utilization can be increased to optimize labor costs during the daily maintenance stage. The resource allocation optimization objective function works together with various constraints, such as skill matching threshold and geographical distance upper limit, to ensure that the generated resource allocation scheme has practical feasibility in addition to theoretical optimality. In some embodiments, the parameters of the resource allocation optimization objective function can be dynamically adjusted through online learning based on historical work order processing results. Optionally, the resource allocation optimization objective function can be extended to include more optimization objectives, such as minimizing the overall maintenance cost or maximizing user satisfaction. The scalar result output by the resource allocation optimization objective function enables the genetic algorithm to sort the chromosomes in the population by fitness, thereby driving the selection, crossover, and mutation operations to continuously approach a better allocation scheme.

[0035] The genetic algorithm encodes the resource allocation scheme as a chromosome, each chromosome representing a possible combination of work order assignments, and iteratively evolves the population through selection, crossover, and mutation operations to generate a resource allocation scheme that satisfies various constraints, including a skill matching threshold and a geographical distance upper limit. The skill matching threshold requires the skill proficiency of the assigned maintenance personnel to be higher than the minimum standard required by the work order, and the geographical distance upper limit ensures that the Euclidean distance between the current location of the maintenance personnel and the fault point does not exceed the pre-set maximum value. It can be understood that the genetic algorithm can effectively handle the combinatorial optimization characteristics of the resource allocation problem. The form of the resource allocation optimization objective function is as follows:

[0036] wherein: represents the objective function value, and are dimensionless weight coefficients for balancing the importance of different optimization objectives, represents the response time of the i-th work order, represents the maximum response time threshold allowed by the system, represents the resource utilization of the j-th maintenance personnel, represents the total number of work orders to be allocated, represents the total number of available maintenance personnel.

[0037] ​​The method of executing the resource scheduling instruction encodes the resource allocation scheme into a machine-readable scheduling instruction set in a structured data format, which contains fields such as work order assignment object, estimated processing time, and optimal path planning. The work order assignment object specifies the specific maintenance personnel identifier responsible, the estimated processing time is predicted based on the time consumption statistics of similar historical work orders, and the optimal path planning calculates the shortest feasible route from the current location of the maintenance personnel to the fault point using a path planning algorithm. The scheduling instruction set is delivered to the maintenance personnel mobile terminal through an application programming interface, which uses the representational state transfer architecture and activates the instruction execution state machine, which defines the life cycle state transition of the work order processing. Real-time monitoring of instruction execution progress, monitoring data comes from the operation feedback and location update service of the maintenance personnel mobile terminal, when the work order state changes or the maintenance personnel location updates are detected, the resource allocation scheme re-computation process is triggered. In some embodiments, the optimal path planning can integrate real-time traffic data. Optionally, the scheduling instruction set can be encoded in a lightweight data interchange format. It can be understood that the real-time monitoring and re-computation mechanism ensures the adaptability of the resource scheduling system, see Table 1.

[0038] Table 1: Maintenance personnel state portrait data table

[0039] Referring to Figure 4 , the correlation between skill matching degree and actual processing time in the campus information maintenance work order scenario is presented. In the figure, the skill matching degree is taken as the horizontal axis, the actual processing time (hours) is taken as the vertical axis, the blue data points represent the actual observed values of different work orders, the red regression line reflects the linear trend of the two, and the pink area is the confidence interval. From a professional point of view, the skill matching degree reflects the degree of fit between the maintenance personnel's skills and the work order requirements, and the actual processing time is the time consumption from the assignment to the completion of the work order. The regression line shows an upward trend, indicating that as the skill matching degree improves, the actual processing time as a whole shows an increasing trend, which may be due to high matching degree work orders often involve more complex fault scenarios. The width of the confidence interval reflects the dispersion degree of the data in different skill matching degree intervals. The narrower the interval, the stronger the data consistency, and the wider the interval, the higher the dispersion. Such analysis provides data support for the resource scheduling optimization module, which can assist the genetic algorithm in generating resource allocation schemes to more accurately weigh the relationship between skill matching degree and processing time, for example, when setting the skill matching degree threshold constraint, the parameter weight in the objective function can be optimized in combination with the figure to achieve the multi-objective balance of minimizing the average response time and maximizing resource utilization.

[0040] In some embodiments, the dual-active communication architecture based on software-defined network is implemented by deploying a software-defined network controller in the campus network infrastructure, which dynamically manages the data plane of two independent physical links as a central control unit. The two independent physical links usually use different transmission media or operator lines to enhance redundancy. The software-defined network controller communicates with network devices such as switches through a southbound interface and issues flow table rules to guide data forwarding. The primary communication path is allocated for the work order data stream, which uses multi-protocol label switching technology to ensure transmission quality. The multi-protocol label switching technology establishes a label switching path for the work order data stream at the network layer, providing a low-delay and high-reliability transmission channel. The standby communication path is allocated for the control signaling, which is implemented based on the user datagram protocol for low-delay communication. The connectionless nature of the user datagram protocol is beneficial for the rapid transmission of control signaling, and the path availability is maintained through a heartbeat detection mechanism that periodically sends probe packets and monitors responses to assess the status of the standby communication path. In some embodiments, the software-defined network controller can be deployed in a high-availability cluster mode to prevent single-point failures. Optionally, the primary communication path and the standby communication path can be physically isolated based on different network topologies, further improving fault tolerance.

[0041] The method for load balancing and fault switching of work order data streams and control signaling integrates a load balancing algorithm in the software-defined network controller, which dynamically adjusts the data packet distribution strategy according to real-time network traffic statistics, including link bandwidth utilization, packet loss rate, and transmission delay. The load balancing algorithm calculates the current load status of each path and allocates work order data streams. A path health index is set, which is a calculated value of comprehensive link quality parameters, such as a weighted score based on bandwidth utilization and error rate. When the path health index of the primary communication path is below a preset threshold, the software-defined network controller automatically switches the work order data stream to the standby communication path. The control signaling uses a multi-copy synchronization mechanism that sends copies through both the primary and standby paths simultaneously when transmitting control signaling, ensuring the continuity and consistency of signaling transmission during path switching. It can be understood that the dual-active communication architecture can effectively handle network link failures and ensure the continuous operation of the work order management system. The calculation formula of the path selection weight in the load balancing algorithm is as follows:

[0042] wherein: represents the load balancing weight of path represents the current available bandwidth of path represents the current transmission delay of path The software-defined network controller calculates the load balancing weight of path ​​The value proportion distribution work order data stream, the path with high weight undertakes more traffic.

[0043] In a specific implementation, the calculation of the path health index can integrate multiple network performance parameters, and the path health index The path health index is obtained by periodic sampling, and the calculation formula is:

[0044] Wherein: represents the bandwidth utilization rate, represents the data packet loss rate, and is a weight coefficient. When the value is lower than the threshold value, the fault switching process is triggered.

[0045] In some embodiments, the threshold value of the path health index is dynamically adjusted according to the network condition, which is realized by the threshold management strategy built in the software defined network controller. The software defined network controller continuously monitors the network performance parameters and calculates the real-time path health index. The threshold management strategy divides the current network state into different operation modes, such as normal mode, busy mode and congestion mode, and each mode corresponds to a path health index threshold value. When the software defined network controller detects that the average bandwidth utilization rate of the network continuously exceeds the preset high water line, the system determines that the network enters the busy mode, at this time, the switching threshold value of the path health index is automatically increased, so that the path fault tolerance standard is more strict, so as to trigger the fault switching earlier when the network load is high, and avoid further performance deterioration. It can be understood that this dynamic adjustment mechanism enables the system to adapt to the periodic fluctuation of network traffic.

[0046] In specific implementations, the dynamic adjustment of the path health indicator threshold value follows a rule set based on historical data and prediction models, which takes into account time factors, network events, and traffic trends. The software-defined network controller analyzes recent path health indicator time series data, calculates its moving average and standard deviation, and moderately increases the threshold value when the indicator volatility increases to avoid frequent false switching when the network condition is unstable. In some embodiments, the threshold value adjustment can be associated with campus schedules, such as during important online exams or video conferences, to adopt more conservative threshold settings to maximize communication stability. Optionally, the software-defined network controller can integrate a simple machine learning model to predict the optimal threshold level for the network in the future period based on historical failure records. The dynamic adjustment process itself also has a protection mechanism to prevent the threshold value from changing too drastically in a short period of time, ensuring the predictability of system behavior. The ultimate goal of dynamic adjustment of the path health indicator threshold value is to balance the sensitivity of failure switching and system stability. For example, a more stringent threshold value is adopted during the peak period of campus activities. Optionally, the failure switching process can set a delay trigger mechanism to avoid false switching due to transient network fluctuations. The multiple copy synchronization mechanism of control signaling is repeated at the receiving end to ensure that only valid signaling is processed. It can be understood that the multiple copy synchronization mechanism increases the transmission reliability of control signaling. The software-defined network controller maintains a path state machine to manage the active state and switching logic of the primary and backup paths.

[0047] Referring to Figure 5 , which shows the bandwidth utilization of the primary communication path, the backup communication path, and the change trend of the primary path weight and the backup path weight over time in a software-defined network-based dual-active communication architecture. From the bandwidth utilization dimension, the primary communication path (blue solid line) has a low bandwidth utilization at 18:00, followed by significant fluctuations in the night period, such as around 20:00, reaching a high value, and then gradually decreasing, and in the early morning period, it is at a low level, and in the daytime period, it shows fluctuations; the backup communication path (green dotted line) has relatively low overall bandwidth utilization, with small fluctuations in some periods, reflecting the difference in bandwidth resource occupation between the primary and backup paths. From the load balancing weight dimension, the primary path weight (red dotted line) and the backup path weight (purple dotted line) show obvious time correlation and complementarity. For example, in the early morning period, the primary path weight is high, and the backup path weight is relatively stable; while in some daytime periods, the backup path weight has a significant increase, and the primary path weight adjusts accordingly, which is consistent with the mechanism of the load balancing algorithm in the software-defined network controller dynamically adjusting the data packet distribution strategy based on real-time network traffic statistics (such as link bandwidth utilization, transmission delay, etc.). Through this dynamic weight distribution, load balancing of ticket data flow and control signaling between the primary and backup paths is achieved, and when the primary communication path health indicator is below the threshold value, failure switching can be triggered to ensure the continuity and reliability of communication.

[0048] It has to be noted that, in the present document, the terms "first", "second", etc. merely serve to identify a subject or action, without necessarily requiring or implying any such actual relationship or order between such subjects or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0049] While embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, combinations, and variations of the embodiments can be undertaken without departing from the spirit and scope of the present application, which is defined by the appended claims and their equivalents.

Claims

1. A campus information system maintenance work order intelligent analysis and service optimization system, characterized in that, It includes a work order information aggregation module, an intelligent priority evaluation module, a resource scheduling optimization module, and a redundant communication management module; The work order information aggregation module is used to asynchronously collect maintenance work order data from multiple campus information subsystems, and perform multi-source fusion processing on the work order data to generate standardized work order records. The intelligent priority evaluation module is used to parse the fault text description and equipment type in the standardized work order record through a deep learning model, automatically generate a work order feature vector, and calculate a dynamic priority coefficient based on the feature vector. The resource scheduling optimization module is used to generate a multi-objective resource allocation scheme based on dynamic priority coefficients and real-time maintenance personnel status, and to execute resource scheduling instructions. The redundant communication management module is used to build a dual-active communication architecture based on software-defined networking to achieve load balancing and fault switching of work order data streams and control signaling.

2. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The work order information aggregation module asynchronously collects maintenance work order data from multiple campus information subsystems, specifically as follows: Deploy distributed data acquisition agents, with each agent monitoring work order generation events of a specific information subsystem and capturing raw work order data through an event-driven mechanism; Semantic parsing and field mapping are performed on the original work order data to eliminate data pattern differences between different subsystems, and a consistent hashing algorithm is applied to perform distributed deduplication of work orders; The deduplicated work order data is encapsulated into messages in a unified data format and asynchronously transmitted to the central data warehouse for persistent storage via a message queue.

3. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The intelligent priority evaluation module uses a deep learning model to parse the fault text description and equipment type in the standardized work order record, specifically: Construct a sequence-to-sequence model with an encoder-decoder structure, where the encoder converts fault text descriptions into context-aware semantic embedding vectors; Graph convolutional networks are used to process device type data, extract device association topology features, and generate device topology embedding vectors. The semantic embedding vector and the device topology embedding vector are concatenated, and the fault impact factor and repair urgency factor are calculated through a fully connected neural network, and then fused to generate the work order feature vector.

4. The campus information maintenance work order intelligent analysis and service optimization system according to claim 3, characterized in that, The method for calculating dynamic priority coefficients based on feature vectors includes: The work order feature vector is input into a time recurrent neural network with an attention mechanism to dynamically learn the weight distribution of work order features over time. External environmental parameters, including campus activity schedules and equipment usage frequency, are introduced, and environmental impact correction values ​​are calculated through a feature cross-layer. The weight distribution is multiplied by the environmental impact correction value to output a dynamic priority coefficient, which is a continuous value used to represent the priority of work order processing.

5. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The resource scheduling optimization module generates a multi-objective resource allocation scheme based on dynamic priority coefficients and real-time maintenance personnel status, specifically as follows: Establish a status profile for maintenance personnel, including a skill matrix, current location coordinates, and workload index, and update the status profile in real time; Using dynamic priority coefficients as weights, a resource allocation optimization objective function is constructed. The optimization objectives include minimizing the average response time and maximizing resource utilization. The objective function is solved using a genetic algorithm to generate a resource allocation scheme that satisfies various constraints, including a skill matching threshold and a geographical distance limit.

6. The campus information maintenance work order intelligent analysis and service optimization system according to claim 5, characterized in that, The method for executing resource scheduling instructions in the resource scheduling optimization module includes: The resource allocation scheme is encoded into a machine-readable set of scheduling instructions, which includes the work order assignment object, the estimated processing time, and the optimal path planning. The scheduling instruction set is sent to the maintenance personnel's mobile terminal via the application programming interface, and the instruction execution state machine is activated. The system monitors the execution progress of instructions in real time. When a change in work order status or an update in the location of maintenance personnel is detected, the system triggers a recalculation process for the resource allocation scheme.

7. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The redundant communication management module constructs a dual-active communication architecture based on software-defined networking, specifically as follows: Deploy a software-defined network controller in the campus network infrastructure; the controller dynamically manages the data plane of two independent physical links. A main communication path is allocated to the work order data stream, and the main communication path uses multi-protocol label switching technology to ensure transmission quality. To control signaling, a backup communication path is allocated. The backup communication path implements low-latency communication based on the User Datagram Protocol and maintains path availability through a heartbeat detection mechanism.

8. The campus information maintenance work order intelligent analysis and service optimization system according to claim 7, characterized in that, The methods for load balancing and fault switching of work order data streams and control signaling in the redundant communication management module include: A load balancing algorithm is integrated into the software-defined network controller. The algorithm dynamically adjusts the packet distribution strategy based on real-time network traffic statistics. Set a path health metric. When the metric of the main communication path is lower than the threshold, the controller will automatically switch the work order data stream to the backup communication path. The control signaling adopts a multi-copy synchronization mechanism to ensure the continuity and consistency of signaling transmission during path switching.

9. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The intelligent priority evaluation module is also used to train a fault prediction model based on historical work order data. The fault prediction model adopts a long short-term memory network structure, and the input sequence data includes equipment operation logs, maintenance records and environmental sensor readings. The model output is a device failure probability distribution, which is used to correct the calculation process of the dynamic priority coefficient.

10. The campus information maintenance work order intelligent analysis and service optimization system according to claim 1, characterized in that, The resource scheduling optimization module is also used to dynamically adjust the resource allocation strategy based on work order processing feedback. Real-time collection of data across the entire work order processing chain, including maintenance operation records, resource consumption indicators, and user evaluation information; Online learning algorithms are used to analyze end-to-end data, dynamically update the parameter weights in the resource allocation optimization objective function, and generate an adaptive resource scheduling strategy.

Images