A computer room fault alarm system based on supporting maintenance information records

Abstract

The present invention relates to the field of fault alarm technology, and in particular to a computer room fault alarm system based on supporting maintenance information recording, which collects and organizes the maintenance activity data of each host and its components in the computer room through a supporting maintenance information recording module, including daily maintenance, repair and component replacement. The hosts in the computer room are classified and arranged in a matrix according to temperature zones. The life cycle module predicts the service life of each component through a neural network model based on component type, temperature zone, maintenance type, maintenance times and maintenance frequency. The component anomaly assessment module monitors the power consumption and error log of the component in real time, and generates anomaly assessment scores based on the frequency of anomalies. When a host shutdown failure is detected, the intelligent fault judgment module will weight the failure probability score and anomaly assessment score of each component to identify possible faulty components, thereby guiding operation and maintenance personnel to quickly and accurately handle the fault.

Description

Technical Field

[0001] The present invention relates to the technical field of fault alarm, and in particular to a computer room fault alarm system based on matching maintenance information records. Background Art

[0002] In the modern information technology architecture, the computer room is an important part of the enterprise and data center. Its stable operation is crucial to ensure business continuity and data security. Cloud computing, big data and digital services, including digital factories, all need to be based on data rooms. However, with the expansion of the scale of computer rooms and the increase in the number of equipment, the complexity of computer room management and fault handling has also increased. In particular, when a host in the computer room goes down or fails, it is important to quickly and accurately locate the cause of the failure and perform corresponding repairs or replacements to maintain the normal operation of the computer room.

[0003] In the traditional computer room management model, after a fault occurs, operation and maintenance personnel usually need to manually check the various components of the affected host to determine the source of the fault. This process is not only inefficient, but also extremely time-consuming and labor-intensive when faced with a large number of hosts in the computer room. In addition, due to the lack of an effective early warning mechanism and fault pre-judgment alarm technology, operation and maintenance personnel often find it difficult to determine the specific location of the fault in the first place, resulting in a longer fault recovery time, which in turn affects the operating labor cost and service quality of the entire computer room. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a computer room fault alarm system based on supporting maintenance information records.

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

[0006] A computer room fault alarm system based on supporting maintenance information recording, comprising a supporting maintenance information recording module, a life cycle module, a component abnormality assessment module, an intelligent fault judgment module and an alarm module, wherein the supporting maintenance information recording module is respectively connected to the life cycle module and the intelligent fault judgment module, and the life cycle module, the component abnormality assessment module and the alarm module are respectively connected to the intelligent fault judgment module;

[0007] The supporting maintenance information recording module is used to divide the host in the computer room into temperature zones according to different heights, arrange several components in several hosts in the same zone in a matrix, and generate a supporting maintenance information recording table. The supporting maintenance information recording table is used to digitally mark the maintenance activities of the components, and the maintenance activities include daily maintenance, repairing components and replacing components;

[0008] The life cycle module is used to obtain the maintenance times and repair times of each component according to the matching maintenance information record table, perform neural network training based on the component category, temperature zone, maintenance times, repair times and maintenance activity frequency changes, and output the component's service life cycle through the neural network;

[0009] The component anomaly assessment module is used to capture component power consumption anomalies and operation error logs, and generate component anomaly assessment scores according to the occurrence frequencies of power consumption anomalies and operation error logs;

[0010] The intelligent fault judgment module is used to obtain the fault probability score through functionalization based on the output of the life cycle module and the working time of each component. When a host shutdown failure occurs, the intelligent fault judgment module weights the fault probability scores and abnormal assessment scores of several components to obtain several component fault judgment scores, selects suspected faulty components based on the fault judgment scores, and controls the broadcast module to issue an alarm.

[0011] Furthermore, the matrix arrangement of the components in the plurality of hosts in the same area comprises the following steps:

[0012] Identify and record the location information of each host, including the rack, row, and column where the host is located;

[0013] Determine the component information of each host, including the type and quantity of components inside the host;

[0014] Based on the location information and component information of each host, the hosts and their internal components in the same temperature zone are arranged in a matrix, wherein the rows of the matrix map different hosts, and the columns of the matrix map the components inside the host.

[0015] Furthermore, the supporting maintenance information record table is used to digitally mark the maintenance type of the component, including the following steps:

[0016] Assign a unique identifier to each maintenance type and record the specific date and time of the maintenance;

[0017] Each maintenance type is assigned a corresponding preset numeric code.

[0018] Furthermore, obtaining the maintenance times and repair times of each component according to the matching maintenance information record table includes the following steps:

[0019] Extracting the maintenance activities of each component from the supporting maintenance information record table, including identifying each row of data in the table and accumulating the number of maintenance activities belonging to the same component;

[0020] The time of component replacement is taken as the end node of the life cycle of the previous component, and as the beginning node of the life cycle of this component to count the number of different types of maintenance activities in the component life cycle, including routine maintenance times and component repair times.

[0021] Furthermore, the step of acquiring the change in the maintenance activity frequency includes:

[0022] Based on the specific date and time of maintenance, the number of maintenance activities within the time window is calculated by presetting the time window and moving step;

[0023] Calculate the difference in the number of maintenance activities within the time window and generate a difference change function;

[0024] The maintenance activity frequency change score was obtained by normalizing the rate of change of the tangent slope of the difference change function.

[0025] Further, the neural network training based on component category, temperature zone, maintenance times, repair times, maintenance activity frequency changes and corresponding life cycle includes:

[0026] The component category, temperature zone, maintenance times within the component life cycle, repair times and maintenance activity frequency change scores are used as the input elements of the neural network, and the component life cycle is used as the output of the neural network for training;

[0027] The network parameters, including weights and biases, are adjusted through back-propagation.

[0028] Further, generating an abnormality assessment score of a component according to the frequency of occurrence of power consumption abnormalities and operation error logs includes:

[0029] For each component, continuously monitor and record its power consumption data and generated operation error logs and mark them as abnormal events;

[0030] For the identified abnormal events, calculate their occurrence frequency within a preset time period;

[0031] The occurrence frequency of abnormal events is standardized to obtain the abnormal evaluation score of the component, and the occurrence frequency of the abnormal events is positively correlated with the abnormal evaluation score of the component.

[0032] Furthermore, the calculation formula of the failure probability score is as follows:

[0033] When T a ≤T p hour,

[0034] When T a >T p hour,

[0035] Among them, T a is the actual working time, T p is the remaining life cycle, k and λ are adjustment coefficients, P failure is the failure probability score.

[0036] Further, selecting a suspected faulty component according to the fault judgment score includes:

[0037] According to the fault judgment score, the suspected faulty components are selected using the maximum value selection principle.

[0038] The beneficial effect of the present invention is that the present invention collects maintenance activity data of each host and its components in the computer room through a supporting maintenance information recording module, including but not limited to information such as daily maintenance, repair components and replacement components. Since the temperature of the host in the computer room affects the life of the host components, and the temperature of the host in the middle part of the large rack in the computer room is often higher, while the temperature of the host close to the ground and the ceiling is relatively low due to the influence of air conditioning or other heat dissipation equipment, the supporting maintenance information recording module classifies the hosts in the computer room according to different temperature zones, and arranges several components in several hosts in the same area in a matrix, and generates a detailed supporting maintenance information recording table. The staff records each maintenance through the supporting maintenance information recording table, which can not only record the maintenance information but also provide basic data for subsequent fault analysis and prediction. The life cycle module uses the collected maintenance information to train the model through a neural network to predict the service life cycle of each component. This module considers changes in component category, temperature zone, maintenance times, repair times and maintenance activity frequency, and the output life cycle data provides an important basis for prediction of the intelligent fault judgment module. The component anomaly assessment module then monitors the power consumption anomalies and operation error logs of the components, and generates component anomaly assessment scores based on the frequency of these anomalies and errors, further refining the accuracy of fault prediction. When a host computer shuts down, the intelligent fault judgment module weights the fault probability scores and anomaly assessment scores of several components to obtain the fault judgment scores of each component. Based on the fault judgment scores, the system can select suspected faulty components and issue warnings through the alarm module to guide operation and maintenance personnel to conduct targeted maintenance, thereby improving the efficiency and accuracy of fault handling and shortening fault recovery time. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a structural schematic diagram of a computer room fault alarm system based on supporting maintenance information records in the present invention.

[0040] Figure 2 It is a flowchart of the steps of obtaining the maintenance times and repair times of each component according to the matching maintenance information record table in the present invention.

[0041] Figure 3 It is a flow chart of the steps of obtaining the change of the frequency of maintenance activities in the present invention. DETAILED DESCRIPTION

[0042] See also Figure 1-3 As shown, the present invention relates to a computer room fault alarm system based on supporting maintenance information recording, including a supporting maintenance information recording module, a life cycle module, a component abnormality assessment module, an intelligent fault judgment module and an alarm module, wherein the supporting maintenance information recording module is respectively connected to the life cycle module and the intelligent fault judgment module, and the life cycle module, the component abnormality assessment module and the alarm module are respectively connected to the intelligent fault judgment module;

[0043] The supporting maintenance information recording module is used to divide the host in the computer room into temperature zones according to different heights, arrange several components in several hosts in the same zone in a matrix, and generate a supporting maintenance information recording table. The supporting maintenance information recording table is used to digitally mark the maintenance activities of the components, and the maintenance activities include daily maintenance, repairing components and replacing components;

[0044] The life cycle module is used to obtain the maintenance times and repair times of each component according to the matching maintenance information record table, perform neural network training based on the component category, temperature zone, maintenance times, repair times and maintenance activity frequency changes, and output the component's service life cycle through the neural network;

[0045] The component anomaly assessment module is used to capture component power consumption anomalies and operation error logs, and generate component anomaly assessment scores according to the occurrence frequencies of power consumption anomalies and operation error logs;

[0046] The intelligent fault judgment module is used to obtain the fault probability score through functionalization based on the output of the life cycle module and the working time of each component. When a host shutdown failure occurs, the intelligent fault judgment module weights the fault probability scores and abnormal assessment scores of several components to obtain several component fault judgment scores, selects suspected faulty components based on the fault judgment scores, and controls the broadcast module to issue an alarm.

[0047] In some embodiments, the supporting maintenance information recording module will serve as the basis for data collection. It can use the temperature sensors deployed in the computer room to collect real-time temperature data, and automatically divide the temperature zones by analyzing these data. For example, the system can identify areas in the computer room where the temperature is continuously high and mark them as high temperature zones. Correspondingly, areas with lower temperatures are divided into low temperature zones. In some cases, the automated tool may not accurately reflect the actual temperature distribution, or it may need to be fine-tuned according to specific maintenance strategies and historical experience. At this time, the system provides the function of manually dividing the temperature zones. The operation and maintenance personnel can manually adjust or define the temperature zones based on their understanding of the computer room environment and the sensitivity of specific equipment. For example, if a high-performance computing cluster generates more heat due to its high power density, even if the automated tool does not mark the surrounding area as a high temperature zone, the operation and maintenance personnel can divide it into a specially monitored high temperature zone according to the actual situation. The supporting maintenance information recording module records the maintenance activities of each host and its components in the computer room, including routine maintenance, repairs, and replacement operations. After being digitized, the information is divided and arranged in a matrix according to the temperature zone where the host is located, forming a comprehensive supporting maintenance information record sheet. The staff records each maintenance through the supporting maintenance information record sheet to obtain supporting information, which can not only record the maintenance information but also provide basic data for subsequent fault analysis and prediction. For example, for a server, the replacement records of key components such as the CPU, memory bar, and hard disk will be recorded and classified according to the rack location of the server.

[0048] Based on the information in the above record table, the life cycle module uses deep learning technology to predict the life cycle of each component. The neural network model here has been trained with a large amount of historical data and can identify and learn the relationship between maintenance activities and component life, thereby outputting a predicted service life for each component. At the same time, the component anomaly assessment module monitors the power consumption data and operation error logs of each component in real time, analyzes these data using advanced anomaly detection algorithms, and calculates an anomaly assessment score. For example, if a server's memory bar suddenly has a high-frequency error log, the module will capture this anomaly and increase the anomaly assessment score of the memory bar accordingly.

[0049] Furthermore, the matrix arrangement of the components in the plurality of hosts in the same area comprises the following steps:

[0050] Identify and record the location information of each host, including the rack, row, and column where the host is located;

[0051] Determine the component information of each host, including the type and quantity of components inside the host;

[0052] Based on the location information and component information of each host, the hosts and their internal components in the same temperature zone are arranged in a matrix, wherein the rows of the matrix map different hosts, and the columns of the matrix map the components inside the host.

[0053] In some embodiments, the system needs to accurately identify the location of each host in the computer room. This step is usually completed by reading the existing data in the computer room management system or by on-site confirmation by the operation and maintenance personnel. The location information includes the rack number where the host is located and the row and column position in the rack. For example, a host may be located in the second row and fourth column of the third rack. These location information not only reflects the distribution of the host in the physical space, but is also closely related to the temperature zone in which it is located, because hosts in different locations may be affected by different degrees of temperature. Subsequently, the system will determine the component information inside each host in detail, including various types of components and their quantities. This step requires the operation and maintenance personnel to conduct a detailed inspection of each host, or automatically obtain it through the host's management interface. The component information covers various hardware components such as CPU, memory bar, hard disk, etc., and the specifications and quantity of each component will be recorded. For example, a server may be equipped with two CPUs, eight memory bars and four hard disks. Finally, the system will arrange all hosts and their components in the same temperature zone in a matrix based on the location information and component information of each host. In this step, the rows of the matrix represent different hosts, and the columns of the matrix map to the components inside the host. In this way, the system can clearly represent the distribution and status information of each host and its components. For example, in a matrix consisting of three hosts, the first row represents the status of the first host and its internal components, and so on.

[0054] Furthermore, the supporting maintenance information record table is used to digitally mark the maintenance type of the component, including the following steps:

[0055] Assign a unique identifier to each maintenance type and record the specific date and time of the maintenance;

[0056] Each maintenance type is assigned a corresponding preset numeric code.

[0057] In some embodiments, the system initially categorizes all possible maintenance activities in the room, including but not limited to routine maintenance, repairing components, and replacing components. Each maintenance activity is assigned a unique identifier based on its nature. For example, routine maintenance may be labeled "MT001" while repairing components may be labeled "MT002". This classification and coding scheme makes it simple and clear to track and reference specific maintenance activities. Whenever a maintenance activity is performed, the system records the specific date and time that the activity occurred. This step is critical because it provides a timeline of maintenance activities, which helps analyze the historical maintenance records of components and predict their future maintenance needs. The records of dates and times are usually stored in the form of timestamps to ensure the accuracy and consistency of the information. In addition to the unique identifier, each maintenance activity is also assigned a preset numeric code for subsequent data processing and analysis. The data in the supporting maintenance information record table is not only highly organized and searchable, but also can interact efficiently with other modules in the system (such as the fault prediction module).

[0058] Furthermore, obtaining the maintenance times and repair times of each component according to the matching maintenance information record table includes the following steps:

[0059] Extracting the maintenance activities of each component from the supporting maintenance information record table, including identifying each row of data in the table and accumulating the number of maintenance activities belonging to the same component;

[0060] The time of component replacement is taken as the end node of the life cycle of the previous component, and as the beginning node of the life cycle of this component to count the number of different types of maintenance activities in the component life cycle, including routine maintenance times and component repair times.

[0061] In some embodiments, the system needs to extract all maintenance activity records for each component from the matching maintenance information record table. This record table is stored in the form of a database or spreadsheet, and each row of data contains detailed information about a single maintenance activity, such as maintenance date, maintenance type (routine maintenance, fault repair, component replacement, etc.), maintenance performer, and related component identifier. The system traverses these records and compares them with the unique identifier of the component to identify and accumulate all maintenance activities belonging to the same component. For example, if a server's hard disk has been maintained three times in the past year (including two routine inspections and one fault repair), the system accumulates these activities to establish a comprehensive maintenance record overview for the hard disk. After extracting the maintenance activity data of each component, the time point of component replacement will be further analyzed. For components that have been replaced, the replacement time point is regarded as the end node of the life cycle of the previous component. Newly installed components are counted into the new life cycle from this time point and begin to accumulate new maintenance times. In this way, the system can clearly distinguish the different life cycle stages of each component and their corresponding maintenance activities. For example, a network switch may undergo multiple software upgrades and hardware overhauls during its five-year lifecycle, and each component replacement will reset its maintenance count, ensuring the data is accurate and relevant.

[0062] Furthermore, the step of acquiring the change in the maintenance activity frequency includes:

[0063] Based on the specific date and time of maintenance, the number of maintenance activities within the time window is calculated by presetting the time window and moving step;

[0064] Calculate the difference in the number of maintenance activities within the time window and generate a difference change function;

[0065] The maintenance activity frequency change score was obtained by normalizing the rate of change of the tangent slope of the difference change function.

[0066] In some embodiments, it is necessary to set a preset time window (such as quarterly, monthly or weekly) based on the specific date and time of the maintenance activity, and define a moving step (for example, one day, one week or one month), depending on the intensity of equipment use and the sensitivity of the maintenance strategy. In each time window, the system will accumulate and record the number of all maintenance activities that occurred in the window. For example, if the selected time window is one month and the moving step is one week, the system will calculate the number of maintenance activities starting from the beginning of the month and accumulated every week to the end of the month. This allows the system to capture the changes in maintenance requirements of each device in different time periods, thereby evaluating the trends and patterns of maintenance activities. After obtaining the number of maintenance activities in each time window, the system further calculates the difference in the number of maintenance times between adjacent time windows. This step is to identify the changing trend of maintenance activities, that is, whether the maintenance demand increases or decreases. The system analyzes these differences by establishing a difference change function, which can be linear or nonlinear, and selects the most appropriate model based on the distribution and change characteristics of the actual data. Next, the system calculates the tangent slope of the change function, that is, the rate of change, which represents the rate of increase or decrease in maintenance activities. Finally, these change rates are normalized to ensure that they are comparable across different equipment and time spans. The normalized score, the maintenance activity frequency change score, provides a quantitative measure that reflects the dynamic changes in equipment maintenance needs.

[0067] Further, the neural network training based on component category, temperature zone, maintenance times, repair times, maintenance activity frequency changes and corresponding life cycle includes:

[0068] The component category, temperature zone, maintenance times within the component life cycle, repair times and maintenance activity frequency change scores are used as the input elements of the neural network, and the component life cycle is used as the output of the neural network for training;

[0069] The network parameters, including weights and biases, are adjusted through back-propagation.

[0070] In some embodiments, the input elements of the neural network are first defined, including: Component category: different types of components may have different failure rates and maintenance requirements, for example, hard disks and memory have different maintenance cycles and failure modes. Temperature zone: the ambient temperature of the device has a significant impact on its performance and life. Devices in higher temperature areas may require maintenance more frequently than devices in relatively lower temperature areas. Maintenance times and repair times: reflect the maintenance history of the component, and the maintenance times usually indicate potential reliability problems. Maintenance activity frequency change score: this value provides a dynamic view of changes in component maintenance requirements and is a key indicator for assessing the possibility of future failures. The output of the neural network is the predicted life cycle of the component, that is, the length of time that the component may continue to operate without replacement failures in a certain period of time in the future. By training the neural network model to approximate this output, a scientific basis can be provided for maintenance decisions. The training process of the neural network includes two main steps: forward propagation and back propagation: in forward propagation, the input data is passed through the network layer, and each layer of the network will weight and activate the data, and finally output the predicted life cycle. The purpose of back propagation is to minimize the error between the predicted output and the actual life cycle by adjusting the weights and biases in the network. By calculating the gradient of the error and updating the network parameters, the model gradually learns how to accurately predict the life cycle of the component. Using optimization methods such as the gradient descent algorithm, the system is able to find the optimal network parameter configuration to minimize the prediction error. This process is repeated until the model's prediction performance reaches an acceptable level or a preset number of iterations is reached.

[0071] Further, generating an abnormality assessment score of a component according to the frequency of occurrence of power consumption abnormalities and operation error logs includes:

[0072] For each component, continuously monitor and record its power consumption data and generated operation error logs and mark them as abnormal events;

[0073] For the identified abnormal events, calculate their occurrence frequency within a preset time period;

[0074] The occurrence frequency of abnormal events is standardized to obtain the abnormal evaluation score of the component, and the occurrence frequency of the abnormal events is positively correlated with the abnormal evaluation score of the component.

[0075] In some embodiments, the power consumption data of each component needs to be continuously monitored. Power consumption monitoring is usually achieved through sensors integrated on the device, which can record the energy consumption of the device in real time. At the same time, the system also monitors and records all generated operation error logs. These logs contain key information about the operating status of the component, such as detailed records of unexpected restarts, performance degradation, or functional failures. For example, for a server, the system may monitor the power consumption of its CPU and record all logs about memory errors or hard disk read and write anomalies. Whenever a sudden increase in power consumption or an error in the operation log is detected, the system automatically identifies it as an abnormal event and records the specific time and details of the event. Once the abnormal event is identified, the system will then calculate the frequency of occurrence of these events within a preset time period. This step involves matching the time range and counting the number of each identified abnormal event. By setting a time window (for example, daily, weekly, or monthly), the system will count the number of occurrences of various abnormal events during this period. The calculated frequency of abnormal events is standardized to generate an abnormal evaluation score for each component. Standardization is necessary because it can eliminate the difference in data magnitude between different devices or under different operating conditions, making the evaluation scores comparable. Normalization is usually achieved by dividing the actual frequency value by a benchmark value (such as the average failure frequency of similar devices). The abnormality assessment score is a quantitative indicator that is positively correlated with the frequency of abnormal events, that is, the higher the frequency, the higher the assessment score. This score can intuitively reflect the health status and failure risk of the component. For example, if the abnormality assessment score of a server continues to increase, it may indicate that its hardware part is gradually approaching the edge of failure.

[0076] Furthermore, the calculation formula of the failure probability score is as follows:

[0077] When T a ≤T p hour,

[0078] When T a >T p hour,

[0079] Among them, T a is the actual working time, T p is the remaining life cycle, k and λ are adjustment coefficients, P failure is the failure probability score.

[0080] It should be noted that when T a ≤T p The formula controls the growth curve of the failure probability by adjusting the coefficient k, where k is a positive coefficient. The calculation in this part reflects the failure probability of the component before reaching its life cycle. As the actual working time Ta Near the remaining life cycle T p , the probability of failure gradually increases.

[0081] When T a >T p When the actual working time exceeds the remaining life cycle, an exponential function is used to calculate the failure probability after the actual working time exceeds the remaining life cycle, where λ is a positive coefficient that controls the growth rate. The exponential function ensures that the failure probability is calculated after the actual working time exceeds the remaining life cycle T p The two formulas combine two common situations in actual operation: one is that the equipment has not reached the predetermined life threshold, and the other is that the equipment has exceeded the predetermined life. By adjusting the two parameters k and λ, the evaluation model of the failure probability can be customized according to the characteristics of different equipment and historical maintenance data.

[0082] Then, the weights of the fault probability score and the abnormality assessment score are determined according to the impact of the component life cycle and the impact of the component abnormality assessment to calculate the fault judgment score. The weight value can be further adjusted through multiple training and back propagation to achieve a more accurate calculation of the fault judgment score.

[0083] Further, selecting a suspected faulty component according to the fault judgment score includes:

[0084] According to the fault judgment score, the suspected faulty components are selected using the maximum value selection principle.

[0085] In some embodiments, the fault judgment score of each component reflects its current fault probability, and the higher the value, the greater the possibility of its failure. In actual operation, when a host device in the computer room is down, the system administrator needs to quickly identify the root cause of the problem in order to restore service as soon as possible. Adopting the maximum value selection principle is an efficient method, that is, selecting the highest one (or several) from the fault judgment scores of all components as the suspected fault component. This method is based on the assumption that the highest fault judgment score usually corresponds to the component that is most likely to fail. First, ensure that the fault judgment scores of all components are updated and available, which may involve refreshing or calculating the data for the last time. Sort the fault judgment scores of all components and identify the component with the highest score. After identifying the component with the highest score, the system will automatically mark it as a suspected fault component and control the broadcast module to alert the personnel.

[0086] 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. A computer room fault alarm system based on supporting maintenance information records, characterized in that: It includes a supporting maintenance information recording module, a life cycle module, a component abnormality assessment module, an intelligent fault judgment module and an alarm module, wherein the supporting maintenance information recording module is respectively connected to the life cycle module and the intelligent fault judgment module, and the life cycle module, the component abnormality assessment module and the alarm module are respectively connected to the intelligent fault judgment module; The supporting maintenance information recording module is used to divide the hosts in the computer room into temperature zones, arrange several components in several hosts in the same zone in a matrix, and generate a supporting maintenance information recording table. The supporting maintenance information recording table is used to digitally mark the maintenance activities of the components, and the maintenance activities include daily maintenance, repairing components, and replacing components; The life cycle module is used to obtain the maintenance times and repair times of each component according to the matching maintenance information record table, perform neural network training based on the component category, temperature zone, maintenance times, repair times and maintenance activity frequency changes, and output the component's service life cycle through the neural network; The component anomaly assessment module is used to capture component power consumption anomalies and operation error logs, and generate component anomaly assessment scores according to the occurrence frequencies of power consumption anomalies and operation error logs; The intelligent fault judgment module is used to obtain the fault probability score through functionization according to the output of the life cycle module and the working time of each component. When a shutdown fault occurs in the host, the intelligent fault judgment module weights the fault probability scores and abnormal evaluation scores of several components to obtain several component fault judgment scores, selects suspected fault components according to the fault judgment scores, and controls the broadcast module to issue an alarm; The supporting maintenance information record table is used to digitally mark the maintenance type of the component and includes the following steps: Assign a unique identifier to each maintenance type and record the specific date and time of the maintenance; Assign a corresponding preset numeric code to each maintenance type; The step of obtaining the change in the maintenance activity frequency includes: Based on the specific date and time of maintenance, the number of maintenance activities within the time window is calculated by presetting the time window and moving step; Calculate the difference in the number of maintenance activities within the time window and generate a difference change function; The maintenance activity frequency change score is obtained by normalizing the change rate of the tangent slope of the difference change function; The calculation formula of the failure probability score is as follows: when hour, ; when hour, ; in, is the actual working time, is the remaining life cycle, and is the adjustment coefficient, is the failure probability score.

2. A computer room fault alarm system based on supporting maintenance information recording according to claim 1, characterized in that: The step of arranging several components in several hosts in the same area into a matrix comprises the following steps: Identify and record the location information of each host, including the rack, row, and column where the host is located; Determine the component information of each host, including the type and quantity of components inside the host; Based on the location information and component information of each host, the hosts and their internal components in the same temperature zone are arranged in a matrix, wherein the rows of the matrix map different hosts, and the columns of the matrix map the components inside the host.

3. A computer room fault alarm system based on supporting maintenance information recording according to claim 1, characterized in that: The method of obtaining the maintenance times and repair times of each component according to the matching maintenance information record table includes the following steps: Extracting the maintenance activities of each component from the supporting maintenance information record table, including identifying each row of data in the table and accumulating the number of maintenance activities belonging to the same component; The time of component replacement is taken as the end node of the life cycle of the previous component, and as the beginning node of the life cycle of this component to count the number of different types of maintenance activities in the component life cycle, including routine maintenance times and component repair times.

4. A computer room fault alarm system based on supporting maintenance information recording according to claim 1, characterized in that: The neural network training based on component category, temperature zone, maintenance times, repair times, maintenance activity frequency changes and corresponding life cycle includes: The component category, temperature zone, maintenance times within the component life cycle, repair times and maintenance activity frequency change scores are used as the input elements of the neural network, and the component life cycle is used as the output of the neural network for training; The network parameters, including weights and biases, are adjusted through back-propagation.

5. A computer room fault alarm system based on supporting maintenance information recording according to claim 1, characterized in that: Generating the abnormality evaluation score of the component according to the occurrence frequency of the power consumption abnormality and the operation error log includes: For each component, continuously monitor and record its power consumption data and generated operation error logs and mark them as abnormal events; For the identified abnormal events, calculate their occurrence frequency within a preset time period; The occurrence frequency of abnormal events is standardized to obtain the abnormal evaluation score of the component, and the occurrence frequency of the abnormal events is positively correlated with the abnormal evaluation score of the component.

6. A computer room fault alarm system based on supporting maintenance information recording according to claim 1, characterized in that: The selecting of suspected faulty components according to the fault judgment score includes: According to the fault judgment score, the suspected faulty components are selected using the maximum value selection principle.

Images