An IT operation and maintenance intelligent management system

Abstract

The present invention relates to the field of IT operation and maintenance technology, and in particular to an IT operation and maintenance intelligent management system, which realizes the precise positioning of equipment in an office building through an IT equipment positioning module. Each device is equipped with a Bluetooth device to synchronize the location information and operating status to a cloud management module in real time. When a device fails, the system quickly obtains the device location information through automated monitoring, which reduces the reliance on manual troubleshooting and significantly shortens the time for maintenance personnel to locate the fault. By tracking the location of maintenance personnel in real time and combining the data of faulty equipment, the optimal scheduling plan is automatically generated using a reinforcement learning model to ensure that the nearest maintenance personnel arrive at the faulty equipment in the shortest time, thereby improving operation and maintenance efficiency. Based on operation and maintenance data analysis, the optimal maintenance plan is generated using an equipment similarity algorithm, the faulty equipment is matched with historical maintenance records, and efficient maintenance methods are automatically recommended to reduce human judgment errors and further shorten fault repair time.

Description

Technical Field

[0001] The present invention relates to the field of IT operation and maintenance technology, and in particular to an IT operation and maintenance intelligent management system. Background Art

[0002] As the degree of enterprise informatization continues to improve, IT infrastructure has become a core component of enterprise operations. However, IT equipment such as printers and computers in large office buildings are of various types and brands, widely distributed and complex to manage. IT operation and maintenance personnel need to face the needs of efficient management and processing of multiple equipment types and potential fault types.

[0003] In the traditional IT operation and maintenance method, the staff in the building usually notify the operation and maintenance department personnel, who need to go on site to determine the type of IT equipment failure and then solve it in a targeted manner. However, due to the complex network environment, the distribution and diversity of equipment, it is often difficult for maintenance personnel to respond quickly when faced with IT equipment maintenance in large office buildings, making it difficult for operation and maintenance personnel to locate faulty IT equipment. At the same time, it is difficult to handle nearby equipment with simultaneous failures nearby or give priority to them, which prolongs the fault repair time, affects the stability of operation and maintenance, and increases labor costs. Summary of the invention

[0004] To solve the above problems, the present invention provides an IT operation and maintenance intelligent management system, which solves the problem of difficult positioning through Bluetooth networking technology, generates decision suggestions for faults in advance, and performs personnel scheduling through reinforcement learning, thereby reducing fault resolution time and reducing operation and maintenance manpower costs.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] An IT operation and maintenance intelligent management system, comprising: an IT equipment positioning module, a personnel positioning module and a cloud management module, wherein the IT equipment positioning module and the personnel positioning module are respectively connected to the cloud management module, and the cloud management module comprises an IT operation and maintenance information input unit, a decision suggestion unit, a priority determination unit and a personnel scheduling unit;

[0007] The IT equipment positioning module is used to locate IT equipment in the building and synchronize it to the cloud management module;

[0008] The personnel positioning module is used to locate personnel and synchronize them to the cloud management module;

[0009] The IT operation and maintenance information input unit is used to obtain the operation and maintenance data of IT equipment, wherein the operation and maintenance data includes equipment information, equipment update data, equipment failure data and maintenance record data;

[0010] The decision suggestion unit is used to calculate the equipment similarity based on the operation and maintenance data of the faulty IT equipment and the IT equipment set and generate a similar equipment set of the faulty IT equipment, and based on the similar equipment set of the faulty IT equipment, weightedly calculate the maintenance plan by the equipment similarity to match the maintenance plan of the faulty IT equipment;

[0011] The priority determination unit is used to determine the maintenance priority score according to the equipment type and the fault type;

[0012] The personnel scheduling unit is used to calculate the maintenance distance according to the maintenance personnel location and IT equipment location, determine the reward function and construct a reinforcement learning model through the maintenance priority score, maintenance time, number of scheduling personnel and maintenance distance of the fault task, and generate a personnel scheduling plan by training the reinforcement learning model.

[0013] Furthermore, the IT equipment positioning module includes several Bluetooth devices installed on the IT equipment, and also includes several Bluetooth transfer terminals installed on each floor of the building. The Bluetooth devices are used to store operation and maintenance data of the corresponding IT equipment and to establish a Bluetooth network by establishing a Bluetooth connection with other Bluetooth devices and Bluetooth transfer terminals within the Bluetooth signal range. The Bluetooth transfer terminals are used to exchange Bluetooth data with the Bluetooth devices installed on the IT equipment and to communicate with the cloud management module.

[0014] Furthermore, the personnel positioning module includes a plurality of Bluetooth mobile terminals, and the Bluetooth mobile terminals are used to obtain their own positioning and the position of the equipment to be maintained by connecting to the Bluetooth network for calculation.

[0015] Furthermore, the step of calculating the device similarity based on the operation and maintenance data of the faulty IT device and the IT device set and generating a similar device set of the faulty IT device includes the following steps:

[0016] Clean and normalize the operation and maintenance data and standardize it into operation and maintenance data vectors;

[0017] Calculate the cosine similarity of the operation and maintenance data vectors of the faulty IT equipment and the equipment in the IT equipment cluster;

[0018] The first ten devices whose cosine similarity is higher than a preset threshold are divided into a similar device set of the faulty IT device.

[0019] Furthermore, performing weighted calculation on the maintenance plans by device similarity to match the maintenance plan of the faulty IT device includes:

[0020] According to the cosine similarity of each device in the similar device set of the faulty IT device, the cosine similarity value is used as a weight to weight the maintenance plan corresponding to the fault type in the maintenance record data of the device in the similar device set, and the scores of several maintenance plans are calculated;

[0021] A maintenance plan with the largest score is selected from the scores of the several maintenance plans as the maintenance plan for the faulty IT equipment.

[0022] Further, determining the maintenance priority score according to the equipment type and the fault type includes: presetting a priority score value interval according to the equipment type, and presetting the maintenance priority score value in the priority score value interval according to the impact of the fault type.

[0023] Furthermore, the maintenance distance is generated by determining the path in the building according to the location of IT equipment and personnel in the building and performing distance ratio calculation on the path.

[0024] Furthermore, determining the reward function and constructing the reinforcement learning model by the maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task includes the following steps:

[0025] Define the scheduling status, including the current location and availability of all maintenance personnel, the location of the faulty IT equipment, and the maintenance priority score of each fault task;

[0026] Define scheduling actions, including assigning faulty IT equipment to maintenance personnel;

[0027] The maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task are weighted to generate a reward function;

[0028] A reinforcement learning model is built based on the scheduling state, scheduling action and reward function, and the scheduling strategy is optimized through the reward function.

[0029] Furthermore, the formula of the reinforcement learning model is as follows:

[0030] ;

[0031] Among them, S is the scheduling state; A is the scheduling action; is the optimal policy function, which indicates the scheduling action A that should be selected under a certain scheduling state S; is the maximum expected reward for scheduling action A; represents expected value calculation; is the scheduling state at time t And select the scheduling action at time t The instant reward you get when you is the discount factor; is the expression of cumulative reward, which means the sum of total rewards from time t=0 to time t=T. All are discounted according to time t.

[0032] The beneficial effect of the present invention is that the present invention realizes the precise positioning of IT equipment distributed in an office building through an IT equipment positioning module. Each device is equipped with a Bluetooth device, which can synchronize its location information and operating status to the cloud management module in real time. When a device fails, the system reduces reliance on manual troubleshooting and obtains the location information of the faulty device in a timely manner through automated monitoring. This greatly shortens the time for maintenance personnel to determine the fault location and avoids the response delay problem caused by the wide distribution of equipment. Secondly, the system tracks the location information of maintenance personnel in real time through the personnel positioning module, and combined with the positioning data of the faulty IT equipment, the system can automatically generate the optimal scheduling plan through a reinforcement learning model. The scheduling plan takes into account the maintenance distance, maintenance time, number of scheduling personnel and task priority, ensuring that the nearest maintenance personnel can reach the faulty equipment in the shortest time for processing. This optimization process avoids the inefficiency of maintenance personnel repeatedly searching for faulty equipment in a complex environment, and can simultaneously handle multiple faults nearby or preferentially, effectively improving the overall operation and maintenance efficiency. In addition, this solution analyzes the operation and maintenance data of the faulty IT equipment and the IT equipment set through the decision-making recommendation unit in the cloud management module to generate the optimal maintenance plan. The decision-making recommendation unit uses the equipment similarity algorithm to match the faulty equipment with the historical maintenance records, and automatically recommends targeted and efficient maintenance plans, reducing the possibility of human misjudgment and further shortening the fault repair time. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a structural diagram of an IT operation and maintenance intelligent management system in the present invention.

[0034] Figure 2 It is a flowchart of the steps of determining the reward function and constructing the reinforcement learning model through the maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task in the present invention. DETAILED DESCRIPTION

[0035] See also Figure 1-2 As shown, the present invention relates to an IT operation and maintenance intelligent management system, comprising: an IT equipment positioning module, a personnel positioning module and a cloud management module, wherein the IT equipment positioning module and the personnel positioning module are respectively connected to the cloud management module, and the cloud management module comprises an IT operation and maintenance information input unit, a decision suggestion unit, a priority determination unit and a personnel scheduling unit;

[0036] The IT equipment positioning module is used to locate IT equipment in the building and synchronize it to the cloud management module;

[0037] The personnel positioning module is used to locate personnel and synchronize them to the cloud management module;

[0038] The IT operation and maintenance information input unit is used to obtain the operation and maintenance data of IT equipment, wherein the operation and maintenance data includes equipment information, equipment update data, equipment failure data and maintenance record data;

[0039] The decision suggestion unit is used to calculate the equipment similarity based on the operation and maintenance data of the faulty IT equipment and the IT equipment set and generate a similar equipment set of the faulty IT equipment, and based on the similar equipment set of the faulty IT equipment, weightedly calculate the maintenance plan by the equipment similarity to match the maintenance plan of the faulty IT equipment;

[0040] The priority determination unit is used to determine the maintenance priority score according to the equipment type and the fault type;

[0041] The personnel scheduling unit is used to calculate the maintenance distance according to the maintenance personnel location and IT equipment location, determine the reward function and construct a reinforcement learning model through the maintenance priority score, maintenance time, number of scheduling personnel and maintenance distance of the fault task, and generate a personnel scheduling plan by training the reinforcement learning model.

[0042] It should be noted that the IT equipment positioning module realizes accurate positioning of the equipment through the Bluetooth device deployed on each IT equipment in the office building. Each Bluetooth device includes a low-power Bluetooth (BLE) chip, a signal processing unit and a built-in storage unit, which is used to store the basic information and operation and maintenance data of the equipment, and can synchronize the real-time location information, operation status and other data of the equipment to the cloud management module through the Bluetooth relay device. The Bluetooth device builds a Bluetooth network by regularly broadcasting its location signal to other Bluetooth devices to ensure a wide range of Bluetooth signal distance. The Bluetooth relay device is located at a key position on each floor of the office building, responsible for receiving Bluetooth signals from multiple devices, and transmitting this information to the cloud management module through a wired or wireless network. The IT operation and maintenance information entry unit in the cloud management module collects this data in real time and stores it in the database for subsequent decision-making. The personnel positioning module realizes real-time tracking of the location of maintenance personnel through Bluetooth mobile devices. The Bluetooth mobile device worn by the maintenance personnel obtains the location data of the maintenance personnel in real time by communicating with the Bluetooth relay device in the building and the Bluetooth device of the IT equipment, and synchronizes this data to the cloud management module. In this way, when a device failure occurs, the cloud management module can immediately determine the distance between the faulty device and the maintenance personnel, and generate the optimal personnel scheduling plan based on the task priority, device type and maintenance personnel status. The cloud management module is the core control unit of the system, which contains multiple functional units for processing and analyzing the collected data. First, the IT operation and maintenance information entry unit collects various types of operation and maintenance data reported by personnel, including the real-time status of the equipment, historical maintenance records and the location of maintenance personnel. The decision-making and recommendation unit analyzes the historical operation and maintenance data of the faulty device and the IT equipment set through a similarity algorithm to generate a similar device set. Based on the historical maintenance records of these similar devices, the decision-making and recommendation unit can calculate the feasibility of each maintenance plan and recommend the optimal maintenance plan for the faulty device. The priority determination unit is used to evaluate the urgency of different fault tasks. According to the type of equipment (such as server, printer, computer, etc.) and the type of fault (such as hardware fault, software fault, network connection problem, etc.), the system assigns a priority score to each fault. The priority score will be used to determine the processing order of tasks and the priority of resource allocation during the scheduling process. The personnel dispatching unit calculates the distance between the maintenance personnel and the faulty equipment based on the output of the priority determination unit and the data from the personnel positioning module. Through the reinforcement learning model, the system can automatically generate the optimal personnel dispatching plan based on factors such as priority, maintenance time, number of dispatching personnel, and maintenance distance. The reinforcement learning model gradually optimizes the dispatching strategy through repeated training, allowing the system to continuously improve response speed and maintenance efficiency during actual operation and maintenance.

[0043] Furthermore, the IT equipment positioning module includes several Bluetooth devices installed on the IT equipment, and also includes several Bluetooth transfer terminals installed on each floor of the building. The Bluetooth devices are used to store operation and maintenance data of the corresponding IT equipment and to establish a Bluetooth network by establishing a Bluetooth connection with other Bluetooth devices and Bluetooth transfer terminals within the Bluetooth signal range. The Bluetooth transfer terminals are used to exchange Bluetooth data with the Bluetooth devices installed on the IT equipment and to communicate with the cloud management module.

[0044] Specifically, the Bluetooth chip uses a mature low-power Bluetooth chip on the market, such as Nordic Semiconductor's nRF52 series chip. The chip has the characteristics of low power consumption and high data transmission rate, and supports multi-device connection and self-organizing network functions. The Bluetooth chip is responsible for periodically broadcasting the location information of the device and receiving signals from other devices or transfer terminals to form a local area network based on the Bluetooth protocol. The signal processing module integrates a digital signal processor (DSP) to process the signal data received by the Bluetooth chip. By analyzing the Bluetooth signal strength (RSSI), the Bluetooth device can estimate its relative position in the floor, thereby providing more accurate device positioning. The signal processing module is also responsible for filtering and denoising to improve positioning accuracy. The memory includes EEPROM or flash memory, which is used to store basic information of the device, historical operation and maintenance data, and real-time status information. The key to the memory design is to support power-off protection and fast data recovery functions to ensure that the device can quickly restore data synchronization capabilities when the Bluetooth signal is interrupted or the device is restarted. The power management module uses an efficient power management chip, such as TI's BQ series, in conjunction with a battery management unit (BMS) to ensure that the Bluetooth device can operate stably for a long time. The power management module integrates energy-saving algorithms to dynamically adjust the working mode of the Bluetooth chip (such as deep sleep mode) to extend the battery life of the device. High-performance Bluetooth receiving module: The transfer end uses a multi-channel Bluetooth receiving module that can simultaneously receive and process signals from multiple Bluetooth devices. This design ensures that the transfer end can maintain efficient data collection capabilities in complex environments, especially in scenarios where high-density equipment is deployed. The entire IT equipment positioning module is designed to achieve real-time, efficient positioning and management of equipment in office buildings. The system's workflow is as follows: The Bluetooth device periodically sends device location information, and the Bluetooth transfer end receives and processes these signals to upload the device location information and operation and maintenance data to the cloud management module. After receiving this data, the cloud management module can update the device's location information and operating status in real time, and quickly locate the device when a fault occurs, so that the operation and maintenance personnel can use it for scheduling.

[0045] Furthermore, the personnel positioning module includes a plurality of Bluetooth mobile terminals, and the Bluetooth mobile terminals are used to obtain their own positioning and the position of the equipment to be maintained by connecting to the Bluetooth network for calculation.

[0046] It should be noted that the Bluetooth mobile terminal can be a mobile phone or any mobile device that supports Bluetooth communication and data interaction. The Bluetooth mobile terminal dynamically obtains the current location of the maintenance personnel by connecting to the Bluetooth transfer terminal and other Bluetooth devices in the office building, and calculates the precise distance and direction between the maintenance personnel and the equipment to be maintained by combining the device location information obtained by the IT equipment positioning module in the Bluetooth network. Specifically, the Bluetooth mobile terminal calculates the current location of the maintenance personnel using a triangulation positioning algorithm based on the received Bluetooth signal strength indication (RSSI) and the location information of the Bluetooth transfer terminal. Combined with the location information of the target device, the Bluetooth mobile terminal can calculate the straight-line distance between the maintenance personnel and the target device and the best path in real time. With the support of the cloud management module, the Bluetooth mobile terminal device keeps in sync with the cloud database to ensure the real-time and accuracy of the positioning data. The cloud management module can push the optimal task scheduling plan for each Bluetooth mobile terminal device according to the priority, distance and current load of the maintenance personnel of the maintenance task, and update the task status in real time. Through this implementation plan, the system can effectively solve the problem of difficult personnel positioning in the traditional IT operation and maintenance process, especially in multi-story office buildings or equipment-intensive areas. The Bluetooth mobile terminal device can help the operation and maintenance personnel quickly find the faulty equipment and complete the troubleshooting task in the shortest time. This system greatly improves operation and maintenance efficiency and reduces time loss caused by complex equipment distribution.

[0047] Furthermore, the step of calculating the device similarity based on the operation and maintenance data of the faulty IT device and the IT device set and generating a similar device set of the faulty IT device includes the following steps:

[0048] Clean and normalize the operation and maintenance data and standardize it into operation and maintenance data vectors;

[0049] Calculate the cosine similarity of the operation and maintenance data vectors of the faulty IT equipment and the equipment in the IT equipment cluster;

[0050] The first ten devices whose cosine similarity is higher than a preset threshold are divided into a similar device set of the faulty IT device.

[0051] In some embodiments, first, the processing of operation and maintenance data starts with data cleaning. Since the data from different devices may contain noise, outliers or missing values, the system uses data cleaning technology to ensure the accuracy and completeness of the input data. The data cleaning process includes outlier detection and elimination, missing data completion, and invalid data filtering. For example, a detection method based on local outlier factor (LOF) is used to identify and eliminate abnormal data; for missing data, the system can complete the data by interpolation or the average method based on historical data. After data cleaning, the system normalizes the processed operation and maintenance data. Operation and maintenance data may involve multiple features, such as the operating time, failure frequency, maintenance time, etc. of the equipment, and the dimensions of each feature are different. To ensure that different features have consistent influence in similarity calculation, the system uses the minimum-maximum normalization method to convert each feature value to an interval between 0 and 1. The normalized data eliminates the dimension difference, making different features comparable in subsequent similarity calculations. After data normalization, the system converts the data of each device into an operation and maintenance data vector. Assuming that the data of each device contains n features, the data vector of each device is , for the faulty device, the system will generate a corresponding feature vector and calculate the similarity with the feature vectors of other devices. By calculating the cosine similarity between the faulty device and other devices, the system can quantify the similarity between the faulty device and other devices. After the similarity calculation, the system sorts the similarities of all devices and selects the top ten devices with similarities higher than the preset threshold as the similar device set of the faulty device. The threshold is set based on empirical data or dynamically adjusted according to operation and maintenance requirements to ensure that the similar device set can represent the typical characteristics of the faulty device.

[0052] Furthermore, performing weighted calculation on the maintenance plans by device similarity to match the maintenance plan of the faulty IT device includes:

[0053] According to the cosine similarity of each device in the similar device set of the faulty IT device, the cosine similarity value is used as a weight to weight the maintenance plan corresponding to the fault type in the maintenance record data of the device in the similar device set, and the scores of several maintenance plans are calculated;

[0054] A maintenance plan with the largest score is selected from the scores of the several maintenance plans as the maintenance plan for the faulty IT equipment.

[0055] It should be noted that, first, a similar device set of the faulty IT device is generated in the previous step, and the device set contains devices with high similarity to the faulty device. The similarity value of each similar device is calculated by cosine similarity, which indicates the similarity between these devices and the faulty device in terms of operating status, historical fault records, maintenance data, etc. The higher the similarity value, the higher the matching degree of the device with the faulty device in the feature dimension. On this basis, the system performs weighted processing on the maintenance record data of each device in the similar device set by using the cosine similarity values ​​of these similar devices as weights. Specifically, the system extracts the maintenance plan related to the current fault type from the maintenance record of the similar device, and weights the maintenance plan with the corresponding similarity value. Score the corresponding maintenance plan in the maintenance record of the device. This weighted calculation method ensures that the maintenance plan of the device with high similarity to the faulty device has a greater impact on the solution selection of the faulty device. After the weighting is completed, the system sorts the weighted scores of all candidate maintenance plans, and finally selects the maintenance plan with the highest score as the recommended solution for the faulty IT device. This maintenance plan combines the historical maintenance records and execution results of multiple similar devices, so it has high applicability and reliability, and can effectively guide maintenance personnel to quickly resolve fault problems.

[0056] Further, determining the maintenance priority score according to the equipment type and the fault type includes: presetting a priority score value interval according to the equipment type, and presetting the maintenance priority score value in the priority score value interval according to the impact of the fault type.

[0057] In some embodiments, first, the system presets a priority score interval for different device types. Device types include but are not limited to servers, printers, computers, network devices, etc. Different types of devices have different importance in the system, and the priority score intervals are also different. For example, the core server may have a higher priority score interval, set to 80 to 100, because it has a greater impact on the entire system, while the priority score interval of ordinary office printers may be lower, set to 20 to 40. These value intervals are preset based on the role and importance of the device in the entire system, which can reflect the degree of impact of the device on the operation of the enterprise. After the device type priority score interval is determined, the system further combines the specific fault type to finely preset the priority score. Different fault types have different degrees of impact on the device. The system dynamically adjusts the score by analyzing the impact of the fault on the normal operation of the device. For example, a hardware failure on the server may cause the entire system to stop operating. The priority of this type of fault should be a higher value in the device type score interval, such as 100; while a relatively easy-to-solve and less impactful fault such as a software update failure may only need to be assigned a medium or low priority, such as 80. Assign according to the severity of the impact of the fault on the device. Fault types with higher severity will have their priority scores increased to ensure that they are handled first.

[0058] Furthermore, the maintenance distance is generated by determining the path in the building according to the location of IT equipment and personnel in the building and performing distance ratio calculation on the path.

[0059] In some embodiments, the system obtains the real-time location of all IT equipment and maintenance personnel in the building through the IT equipment positioning module and the personnel positioning module. These location data are captured through Bluetooth positioning technology and recorded in the cloud management module in the form of coordinates. To improve positioning accuracy, the system will calculate the specific location of each device and personnel in the building based on the Bluetooth signal strength indication (RSSI) combined with the triangulation positioning algorithm. After obtaining the preliminary location data, the system accesses the floor plan or building information model (BIM) in the building, which contains the actual layout information of the office building, such as the location and connectivity of corridors, rooms, stairs and elevators. The system combines this location information with the building layout to generate an actual feasible path rather than a simple straight-line distance. Specifically, the system determines the shortest feasible path from the current location of the maintenance personnel to the faulty IT equipment in two-dimensional or three-dimensional space through a path planning algorithm such as the Dijkstra algorithm or the A* algorithm. After the path planning is completed, the system calculates the distance ratio of the path. Since there may be differences in the difficulty of access to different areas in the office building, such as elevator connections between floors, stair walking time, etc., the simple geometric distance cannot fully reflect the time and resource consumption required for maintenance personnel to reach the equipment. To this end, the system introduces a distance ratio calculation model, which adjusts the time cost of different path nodes through preset weights. For example, the system can set higher weights for the use of elevators between different floors, because waiting for elevators and moving between floors usually takes more time than walking in a straight line on the same floor. Correspondingly, the walking time of the stairs will also be given a higher weight, and the system will adjust the distance of each section in the path according to these weights to generate a proportional maintenance distance. Once the maintenance distance calculation is completed, the system uses the result as an important input parameter of the scheduling algorithm, and combines factors such as the priority of the maintenance task and the current status of the maintenance personnel to generate the optimal personnel scheduling plan to ensure that operation and maintenance resources are reasonably allocated and faulty equipment can be handled in the shortest time.

[0060] Furthermore, determining the reward function and constructing the reinforcement learning model by the maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task includes the following steps:

[0061] Define the scheduling status, including the current location and availability of all maintenance personnel, the location of the faulty IT equipment, and the maintenance priority score of each fault task;

[0062] Define scheduling actions, including assigning faulty IT equipment to maintenance personnel;

[0063] The maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task are weighted to generate a reward function;

[0064] A reinforcement learning model is built based on the scheduling state, scheduling action and reward function, and the scheduling strategy is optimized through the reward function.

[0065] In some embodiments, first, the system defines a scheduling state S, which contains all the key information that affects the scheduling decision. Specifically, the scheduling state includes the following: Location of maintenance personnel: The geographical location information of maintenance personnel is obtained in real time through the personnel positioning module, accurate to the floor and specific location. The system uses Bluetooth positioning technology or Wi-Fi positioning technology to record the current location of each maintenance personnel as a state variable. Available status of maintenance personnel: The current task status of each maintenance personnel (such as whether the task is being performed) is also included in the state variable. The system uses this status information to determine which personnel can immediately respond to new fault tasks. Location and status of faulty IT equipment: The location information of the faulty equipment is obtained through the equipment positioning module. Maintenance priority score of fault tasks: The system assigns a priority score to each fault task, which is preset based on the equipment type and fault type. Secondly, the system defines a scheduling action A. In each state, the scheduling action includes the following: Matching of maintenance personnel and tasks: Assign the most suitable fault task to each available maintenance personnel. The system generates multiple optional scheduling schemes based on the distance between the maintenance personnel and the faulty equipment, the skill matching degree of the maintenance personnel, and the priority score of the current task. Each scheme corresponds to a specific scheduling action A. Task adjustment and resource reallocation: When multiple simultaneous faults occur, the system needs to dynamically adjust the task allocation of maintenance personnel. For example, when a high-priority fault occurs, the system can interrupt some low-priority tasks and reallocate related resources to high-priority tasks. Based on the above state and action definitions, the system designs a reward function R to evaluate the effectiveness of each scheduling action. The reward function is the key to the optimization of the reinforcement learning model, and its design needs to take into account multiple factors: Maintenance priority score: The system gives higher rewards to scheduling actions that successfully solve high-priority tasks. This part of the reward ensures that the system can prioritize the fault tasks that have the greatest impact on the operation of the enterprise. Maintenance time: The shorter the time to complete the task, the higher the reward. The system improves the overall operation and maintenance efficiency by shortening the task response time and execution time. Number of dispatchers: The system encourages the use of fewer maintenance personnel to complete tasks in order to improve resource utilization efficiency. If the same task is completed by fewer personnel, the system will give higher rewards. Maintenance distance: The shorter the distance between the maintenance personnel and the faulty equipment, the faster the response time, and the system gives higher rewards for such scheduling. The system calculates the actual walking distance through the path planning algorithm and rewards short-distance scheduling behavior. Finally, the system optimizes the scheduling strategy through a model based on deep reinforcement learning. The model uses the deep Q network (DQN) or the proximal policy optimization (PPO) algorithm. During the training process, the model continuously interacts with the operation and maintenance environment, accumulates experience data, and continuously updates the strategy based on the feedback of the reward function.The specific steps of model training are as follows: Experience accumulation: After each execution of the scheduling action A, the system stores the current state S, action A, reward R, and next state S′ into the experience replay pool. These experience data will be used for model training. Policy update: The model estimates the Q value or optimizes the strategy by extracting samples from the experience replay pool using a deep neural network. In the DQN algorithm, the model updates the strategy by minimizing the Bellman error of the Q value; in the PPO algorithm, the model improves the long-term cumulative reward through policy gradient optimization. Strategy evaluation and improvement: The system continuously adjusts the weight parameters in the reward function by evaluating the actual task completion effect, and tests the new scheduling strategy in the new task environment, gradually approaching the global optimal solution.

[0066] Furthermore, the formula of the reinforcement learning model is as follows:

[0067] ;

[0068] Among them, S is the scheduling state; A is the scheduling action; is the optimal policy function, which indicates the scheduling action A that should be selected under a certain scheduling state S; is the maximum expected reward for scheduling action A; represents expected value calculation; is the scheduling state at time t And select the scheduling action at time t The instant reward you get when you is the discount factor; is the expression of cumulative reward, which means the sum of total rewards from time t=0 to time t=T. All are discounted according to time t.

[0069] It should be noted that the optimal strategy function is the core goal of the model, which indicates which scheduling action A the system should choose to maximize the future cumulative reward under a certain scheduling state S. This policy function is continuously optimized through the reinforcement learning model, aiming to help the system find the optimal scheduling decision, thereby improving the overall operation and maintenance efficiency. The immediate reward is used to evaluate the immediate effect of the system after executing a certain scheduling action at each point in time. The design of the reward takes into account multiple factors, including the priority of the task, the completion time, the efficiency of resource utilization (such as the number of maintenance personnel used), and the maintenance distance. The immediate reward is the basis of model training. Through the reward function, the system can understand the quality of each scheduling decision and make strategy adjustments. Used to balance the impact of current rewards and future rewards. A larger discount factor means that the system pays more attention to long-term returns, while a smaller discount factor means that the system pays more attention to immediate benefits. In the calculation of cumulative rewards, future rewards are discounted according to time, which can guide the system to focus on both short-term goals and long-term optimization.

[0070] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An IT operation and maintenance intelligent management system, characterized in that: include: An IT equipment positioning module, a personnel positioning module and a cloud management module, wherein the IT equipment positioning module and the personnel positioning module are respectively connected to the cloud management module, and the cloud management module includes an IT operation and maintenance information input unit, a decision suggestion unit, a priority determination unit and a personnel scheduling unit; The IT equipment positioning module is used to locate IT equipment in the building and synchronize it to the cloud management module; The personnel positioning module is used to locate personnel and synchronize them to the cloud management module; The IT operation and maintenance information input unit is used to obtain the operation and maintenance data of IT equipment, wherein the operation and maintenance data includes equipment information, equipment update data, equipment failure data and maintenance record data; The decision suggestion unit is used to calculate the equipment similarity based on the operation and maintenance data of the faulty IT equipment and the IT equipment set and generate a similar equipment set of the faulty IT equipment, and based on the similar equipment set of the faulty IT equipment, weightedly calculate the maintenance plan by the equipment similarity to match the maintenance plan of the faulty IT equipment; The priority determination unit is used to determine the maintenance priority score according to the equipment type and the fault type; The personnel scheduling unit is used to calculate the maintenance distance according to the maintenance personnel location and IT equipment location, determine the reward function and construct a reinforcement learning model through the maintenance priority score, maintenance time, number of scheduling personnel and maintenance distance of the fault task, and generate a personnel scheduling plan by training the reinforcement learning model.

2. An IT operation and maintenance intelligent management system according to claim 1, characterized in that: The IT equipment positioning module includes several Bluetooth devices installed on the IT equipment, and also includes several Bluetooth transfer terminals installed on each floor of the building. The Bluetooth devices are used to store the operation and maintenance data of the corresponding IT equipment and to establish a Bluetooth network by establishing a Bluetooth connection with other Bluetooth devices and Bluetooth transfer terminals within the Bluetooth signal range. The Bluetooth transfer terminal is used to exchange Bluetooth data with the Bluetooth devices installed on the IT equipment and to communicate with the cloud management module.

3. An IT operation and maintenance intelligent management system according to claim 2, characterized in that: The personnel positioning module includes a plurality of Bluetooth mobile terminals, and the Bluetooth mobile terminals are used to obtain their own positioning and the position of the equipment to be maintained by connecting to the Bluetooth network for calculation.

4. The IT operation and maintenance intelligent management system according to claim 1, characterized in that: The step of calculating the device similarity based on the operation and maintenance data of the faulty IT device and the IT device set and generating a similar device set of the faulty IT device comprises the following steps: Clean and normalize the operation and maintenance data and standardize it into operation and maintenance data vectors; Calculate the cosine similarity of the operation and maintenance data vectors of the faulty IT equipment and the equipment in the IT equipment cluster; The first ten devices whose cosine similarity is higher than a preset threshold are divided into a similar device set of the faulty IT device.

5. An IT operation and maintenance intelligent management system according to claim 4, characterized in that: The weighted calculation of the maintenance plan by the equipment similarity to match the maintenance plan of the faulty IT equipment includes: According to the cosine similarity of each device in the similar device set of the faulty IT device, the cosine similarity value is used as a weight to weight the maintenance plan corresponding to the fault type in the maintenance record data of the device in the similar device set, and the scores of several maintenance plans are calculated; A maintenance plan with the largest score is selected from the scores of the several maintenance plans as the maintenance plan for the faulty IT equipment.

6. An IT operation and maintenance intelligent management system according to claim 1, characterized in that: Determining the maintenance priority score according to the equipment type and the fault type includes: presetting a priority score value interval according to the equipment type, and presetting the maintenance priority score value in the priority score value interval according to the impact degree of the fault type.

7. An IT operation and maintenance intelligent management system according to claim 1, characterized in that: The maintenance distance is generated by determining the path in the building according to the location of IT equipment and personnel in the building and performing distance ratio calculation on the path.

8. An IT operation and maintenance intelligent management system according to claim 7, characterized in that: Determining the reward function and building a reinforcement learning model based on the maintenance priority score, maintenance time, number of dispatchers, and maintenance distance of the fault task includes the following steps: Define the scheduling status, including the current location and availability of all maintenance personnel, the location of the faulty IT equipment, and the maintenance priority score of each fault task; Define scheduling actions, including assigning faulty IT equipment to maintenance personnel; The maintenance priority score, maintenance time, number of dispatchers and maintenance distance of the fault task are weighted to generate a reward function; A reinforcement learning model is built based on the scheduling state, scheduling action and reward function, and the scheduling strategy is optimized through the reward function.

9. An IT operation and maintenance intelligent management system according to claim 8, characterized in that: The formula of the reinforcement learning model is as follows: ； Among them, S is the scheduling state; A is the scheduling action; is the optimal policy function, which indicates the scheduling action A that should be selected under a certain scheduling state S; is the maximum expected reward for scheduling action A; represents expected value calculation; is the scheduling state at time t And select the action to be scheduled at time t The instant reward you get when you is the discount factor; is the expression of cumulative reward, which means the sum of total rewards from time t=0 to time t=T. All are discounted according to time t.

Images