A fan warning method, system, device and computer readable storage medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INSPUR (SHANDONG) COMPUTER TECH CO LTD
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-26
Smart Images
Figure CN122083010A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of equipment alarm technology, and more specifically, to a fan alarm method, system, device, and computer-readable storage medium. Background Technology
[0002] In servers, the main function of the fan is to dissipate the heat generated by the CPU (Central Processing Unit) during operation outside the chassis in a timely manner, preventing overheating that could lead to system instability or even damage.
[0003] If a fan fails, it will directly affect the normal operation of the server. Therefore, the fan needs to be troubleshooted, such as replacing it after the fan rotor is detected to be damaged. However, waiting until the fan rotor fails before replacing it can easily affect the system's heat dissipation regulation, causing the server to operate at an unsuitable temperature and leading to the risk of server downtime. Moreover, the time between the fan failure and finding a replacement part is long, and the server will be in a faulty state for a long time, increasing the risk of server instability.
[0004] In conclusion, ensuring the stable operation of servers is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] The purpose of this application is to provide a fan alarm method, which can, to some extent, solve the technical problem of how to ensure the stable operation of a server. This application also provides a fan alarm system, electronic device, and computer-readable storage medium.
[0006] To achieve the above objectives, this application provides the following technical solution: A fan alarm method, applied to the BMC of a server, includes: Acquire server fan performance data over a target duration, including current and speed values at each time step; The performance data within the target time period is processed to generate the target time series; Input the target time series into the target LSTM model; Obtain rotor wear data output by the target LSTM model; Determine whether to issue an alarm based on the rotor wear data; In response to an alarm determined based on the rotor wear data, a fan alarm operation is executed; The target LSTM model determines the degree of rotor wear based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the degree of rotor wear.
[0007] In an exemplary embodiment, the process of processing performance data within a target duration to generate a target time series includes: The performance data within the target time period is sorted according to time order to obtain the initial time series; By using the sliding window 3σ method, abnormal current values and abnormal rotation speed values in the initial time series are removed to obtain candidate time series; The values removed from the candidate time series are interpolated to obtain the target time series.
[0008] In an exemplary embodiment, the step of removing abnormal current values and abnormal rotational speed values from the initial time series using the sliding window 3σ method to obtain a candidate time series includes: The initial time series is divided into a set number of consecutive sliding windows; For any sliding window, generate the local mean and local standard deviation of the current value within the sliding window, and generate the local mean and local standard deviation of the rotational speed value; For any sliding window, an abnormal current threshold is generated based on the local mean and local standard deviation of the current value, and current values outside the abnormal current threshold are removed. For any sliding window, an abnormal speed threshold is generated based on the local mean and local standard deviation of the speed values, and speed values outside the abnormal speed threshold are removed. The values that were not removed from the initial time series are used as candidate time series.
[0009] In an exemplary embodiment, the step of interpolating the values removed from the candidate time series to obtain the target time series includes: For any sliding window, using the removed current value as a position reference, the normal current value within the first position range is determined in the candidate time series; based on the normal current value, the removed current value is interpolated using a polynomial interpolation method. For any sliding window, using the removed rotational speed value as the position reference, the normal rotational speed value within the second position range is determined in the candidate time series; based on the normal rotational speed value, the removed rotational speed value is interpolated using polynomial interpolation to obtain the target time series.
[0010] In an exemplary embodiment, before inputting the target time series into the target LSTM model, the method further includes: Obtain a pre-trained and lightweight target LSTM model; Deploy the target LSTM model.
[0011] In an exemplary embodiment, obtaining the pre-trained and lightweight target LSTM model includes: Obtain the factory setting data of the server fan, which includes the fan rotor speed and current value under different wear levels; The factory setting data is processed to generate a training time series and a rotor wear label for the training time series. The training time series is divided into a training set, a validation set, and a test set according to the chronological order. The initial LSTM model is trained using the training set, its parameters are tuned using the validation set, and it is tested using the test set. If the test set loss no longer decreases in multiple consecutive iterations, the model weights with the optimal validation set loss are used as the weights of the trained initial LSTM model. The trained initial LSTM model is lightweighted to obtain the target LSTM model.
[0012] In an exemplary embodiment, determining whether to issue an alarm based on the rotor wear data includes: The test is to determine whether the rotor wear data indicates severe wear of the fan rotor. If the rotor wear data indicates that the fan rotor is not severely worn, then no alarm will be issued; In response to the rotor wear data indicating severe wear of the fan rotor, the system detects whether the fan rotor is continuously under severe wear within a time threshold, where the time threshold is greater than the target duration. If the fan rotor does not experience continuous heavy wear within the time threshold, no alarm will be triggered. An alarm is triggered in response to continuous heavy wear of the fan rotor within a time threshold.
[0013] A fan alarm system, applied to the BMC of a server, includes: The performance data acquisition module is used to acquire the performance data of the server fan within a target duration, and the performance data includes the current value and speed value at the time step. The performance data processing module is used to process performance data within the target duration and generate the target time series. The target time series input module is used to input the target time series into the target LSTM model; The result acquisition module is used to acquire rotor wear data output by the target LSTM model; The alarm detection module is used to determine whether an alarm should be triggered based on the rotor wear data. The alarm module is used to execute a fan alarm operation in response to an alarm determination based on the rotor wear level data; The target LSTM model determines the degree of rotor wear based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the degree of rotor wear.
[0014] An electronic device, comprising: Memory, used to store computer programs; A processor, configured to implement the steps of any of the above-described fan alarm methods when executing the computer program.
[0015] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described fan alarm methods.
[0016] This application provides a fan alarm method applied to the BMC of a server. The method acquires server fan performance data over a target duration, including current and speed values at each time step. It processes the performance data to generate a target time series, inputs the target time series into a target LSTM model, acquires rotor wear data output by the target LSTM model, determines whether an alarm is triggered based on the rotor wear data, and executes a fan alarm operation in response to the alarm determination based on the rotor wear data. The target LSTM model determines the rotor wear degree based on the increase in current value at the same speed; the greater the increase in current value, the higher the rotor wear degree. This application incorporates an LSTM model into the server's BMC (Browser Control Center). Since, with a constant fan rotor speed, the higher the rotor wear, the greater the increase in fan current over time. Therefore, rotor wear and current exhibit a consistent trend over time. The LSTM model can accurately process this trend, allowing the BMC to accurately identify rotor wear and issue fan alarms. This enables accurate and timely fan fault handling, ensuring stable server operation. The fan alarm system, electronic device, and computer-readable storage medium provided in this application also solve the corresponding technical problems. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0018] Figure 1 A flowchart illustrating a fan alarm method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a fan alarm system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 4 This is another structural schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] Please see Figure 1 , Figure 1 A flowchart of a fan alarm method provided in an embodiment of this application.
[0021] This application provides a fan alarm method applied to the BMC of a server, which may include the following steps: Step S101: Obtain the performance data of the server fan within the target duration. The performance data includes the current value and speed value at each time step.
[0022] In practical applications, the server's BMC (Baseboard Management Controller) acts as an independent intelligent control unit separate from the server system. It is responsible for core functions such as hardware monitoring and management, health status management, power consumption management, and fan control. Therefore, the BMC can be used to handle alarms for server fans. During this process, the BMC needs to acquire performance data of the server fans within a target duration. This performance data includes current and speed values at each time step. The target duration is used to analyze rotor wear and can be flexibly set according to actual needs, such as 5 minutes or 10 minutes.
[0023] Step S102: Process the performance data within the target duration to generate the target time series.
[0024] In practical applications, after analyzing the physical and electronic characteristics of the fan, it is found that, with the fan rotor speed remaining constant, the higher the rotor wear, the greater the increase in fan current value over time. In other words, the essence of rotor wear is increased mechanical resistance. To maintain the same speed, the motor drive current increases. Therefore, the rotor wear and current value show a consistent pattern over time. This pattern can be used to evaluate the rotor wear. Therefore, it is necessary to process the performance data within the target time period to generate a target time series. The target time series is used to reflect the changes in speed and current values over time.
[0025] In an exemplary embodiment, considering that fan rotor wear is a gradual process and the current value will slowly increase over time, static 3σ will identify this trend as an anomaly, leading to false filtering. To avoid this, the performance data within the target time period is processed. In the process of generating the target time series, the performance data within the target time period can be sorted according to time order to obtain an initial time series. Abnormal current values and abnormal speed values in the initial time series are removed by the sliding window 3σ method to obtain a candidate time series. The values removed from the candidate time series are interpolated to obtain the target time series.
[0026] In specific application scenarios, during the process of eliminating abnormal current and speed values from the initial time series using the sliding window 3σ method to obtain candidate time series, the initial time series can be divided into a set number of continuous sliding windows. This set number can be flexibly set according to alarm sensitivity; for example, a larger set number can be used when high alarm sensitivity is required, and vice versa. For any sliding window, local mean and local standard deviation of the current and speed values are generated within the window. For any sliding window, an abnormal current threshold is generated based on the local mean and local standard deviation of the current values, and current values outside this threshold are eliminated. Similarly, for any sliding window, an abnormal speed threshold is generated based on the local mean and local standard deviation of the speed values, and speed values outside this threshold are eliminated. The values not eliminated from the initial time series are then used as candidate time series. In this way, the time series data of current and speed values can be divided into fixed-length sliding windows by using the sliding window 3σ filtering method. The local mean and standard deviation are calculated separately in each sliding window, and the threshold is dynamically updated. Moreover, the threshold of each sliding window can only remove abnormal current and speed values within that sliding window, which can prevent misjudgment and ensure the authenticity and integrity of the time series data.
[0027] In specific application scenarios, to ensure the continuity of timestamps, the values removed from the candidate time series are interpolated to obtain the target time series. For any sliding window, the removed current value can be used as a position reference to determine the normal current value within the first position range in the candidate time series; based on the normal current value, the removed current value is interpolated using polynomial interpolation; for any sliding window, the removed rotational speed value is used as a position reference to determine the normal rotational speed value within the second position range in the candidate time series; based on the normal rotational speed value, the removed rotational speed value is interpolated using polynomial interpolation to obtain the target time series.
[0028] Step S103: Input the target time series into the target LSTM model. The target LSTM model determines the rotor wear degree based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the rotor wear degree.
[0029] Step S104: Obtain rotor wear data output by the target LSTM model.
[0030] In practical applications, considering that LSTM models excel at capturing long-term dependencies and dynamic changes in time-series data, and can effectively identify rotor wear characteristics in this application, BMC can input the target time series into the target LSTM model, and then obtain the rotor wear degree data output by the target LSTM model. The target LSTM model determines the rotor wear degree based on the increase in current value at the same rotational speed; the greater the increase in current value, the higher the rotor wear degree. The type of rotor wear degree data can be flexibly determined according to the application scenario. For example, the rotor wear degree data can be fixed values such as 0, 1, 2, and 3, where 0 represents normal rotor, 1 represents slight rotor wear, 2 represents moderate rotor wear, and 3 represents severe rotor wear, etc.
[0031] In an exemplary embodiment, before inputting the target time series into the target LSTM model, a pre-trained and lightweight target LSTM model can be obtained; the target LSTM model is then deployed so that the BMC can use the target LSTM model normally with low power consumption, saving the server's consumption for fan alarms.
[0032] In a specific application scenario, during the process of acquiring the pre-trained and lightweight target LSTM model, the factory setting data of the server fan can be obtained. This factory setting data includes the fan rotor speed and current values under different wear levels. The factory setting data is processed to generate a training time series and rotor wear labels for the training time series. The training time series is then divided into a training set, a validation set, and a test set according to chronological order. The initial LSTM model is trained using the training set, its parameters are adjusted using the validation set, and it is tested using the test set. If the loss on the test set no longer decreases in multiple iterations, the model weights with the optimal loss on the validation set are used as the weights of the trained initial LSTM model. The trained initial LSTM model is then lightweighted to obtain the target LSTM model. In this way, "splitting the training set, validation set, and test set in chronological order" can avoid the risk of data leakage and ensure the generalization ability of the LSTM model on real time series. Furthermore, if the test set loss does not decrease after multiple iterations, training is immediately terminated and the model weights with the optimal validation set loss during training are restored. This can effectively suppress overfitting of the LSTM model and ensure the generalization ability and wear level discrimination accuracy of the LSTM model on unknown data.
[0033] In specific application scenarios, when inputting the target time series into the target LSTM model, BMC can first generate the variance or gradient values of the current and speed values in the target time series. If both the variance and gradient values of the current and speed values are less than a threshold, the convolutional kernels or attention heads processing high-frequency features in the target LSTM model are automatically pruned. The target time series is then input into the pruned target LSTM model. If either the variance or gradient value of the current or speed value is greater than the threshold, the target time series is input into the complete target LSTM model. In this way, the variance or gradient value can be used to reflect the fluctuation of the signal. Values less than the threshold indicate a stable signal, while values greater than the threshold indicate a drastic fluctuation. Correspondingly, when the current and speed signals are very stable, it is not necessary to activate all neurons in the target LSTM model. Only when the current and speed fluctuate drastically is the entire model required to participate in the calculation. This achieves adaptive model pruning of the target LSTM model, allowing BMC computing power to be allocated as needed, further saving energy consumption for fan alarms.
[0034] Step S105: Determine whether to issue an alarm based on rotor wear data.
[0035] Step S106: In response to an alarm determined based on rotor wear data, a fan alarm operation is performed.
[0036] In practical applications, after obtaining the rotor wear level data output by the target LSTM model, BMC can determine whether to issue an alarm based on the rotor wear level data. In response to the alarm determination based on the rotor wear level data, it will perform fan alarm operations, such as playing fan fault information by voice, displaying fan fault information by text, or sending fan fault information to the user client.
[0037] In an exemplary embodiment, during the process of determining whether to issue an alarm based on rotor wear data, it can detect whether the rotor wear data indicates severe wear of the fan rotor. If the rotor wear data indicates that the fan rotor is not severely worn, then no alarm is issued. If the rotor wear data indicates severe wear of the fan rotor, then it can detect whether the fan rotor experiences continuous severe wear within a time threshold, where the time threshold is greater than a target duration, such as 5 hours, 10 hours, or one day. If the fan rotor does not experience continuous severe wear within the time threshold, then no alarm is issued. If the fan rotor experiences continuous severe wear within the time threshold, then an alarm is issued. In this way, the BMC will only issue a fan alarm when the server fan rotor experiences continuous severe wear within the time threshold, reducing the need for pre-emptive alerts and improving the accuracy of fan alarms.
[0038] This application provides a fan alarm method applied to the BMC of a server. The method acquires server fan performance data over a target duration, including current and speed values at each time step. It processes the performance data to generate a target time series, inputs the target time series into a target LSTM model, acquires rotor wear data output by the target LSTM model, determines whether an alarm is triggered based on the rotor wear data, and executes a fan alarm operation in response to the alarm determination based on the rotor wear data. The target LSTM model determines the rotor wear degree based on the increase in current value at the same speed; the greater the increase in current value, the higher the rotor wear degree. This application incorporates an LSTM model into the server's BMC. Since, with a constant fan rotor speed, the higher the rotor wear, the greater the increase in fan current over time. Therefore, rotor wear and current exhibit a consistent trend over time. The LSTM model can accurately process this trend, enabling the BMC to accurately identify rotor wear and issue fan alarms. This allows for accurate and timely fan fault handling, ensuring stable server operation.
[0039] Please see Figure 2 , Figure 2 This is a schematic diagram of a fan alarm system provided in an embodiment of this application.
[0040] This application provides a fan alarm system applied to the BMC of a server, which may include: The performance data acquisition module 101 is used to acquire the performance data of the server fan within a target duration. The performance data includes the current value and speed value at each time step. The performance data processing module 102 is used to process the performance data within the target duration and generate the target time series. The target time series input module 103 is used to input the target time series into the target LSTM model; The result acquisition module 104 is used to acquire rotor wear data output by the target LSTM model; The alarm detection module 105 is used to determine whether an alarm should be triggered based on rotor wear data; Alarm module 106 is used to execute fan alarm operation in response to an alarm determined based on rotor wear data; The target LSTM model determines the degree of rotor wear based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the degree of rotor wear.
[0041] This application provides a fan alarm system applied to the BMC of a server, wherein the performance data processing module may include: The initial time series generation unit is used to sort the performance data within the target duration according to time order to obtain the initial time series; The outlier filtering unit is used to remove abnormal current and abnormal speed values from the initial time series using the sliding window 3σ method to obtain candidate time series. The interpolation unit is used to interpolate the values that have been removed from the candidate time series to obtain the target time series.
[0042] This application provides a fan alarm system applied to the BMC of a server. The outlier filtering unit is specifically used for: dividing the initial time series to obtain a set number of continuous sliding windows; for any sliding window, generating the local mean and local standard deviation of the current value within the sliding window, and generating the local mean and local standard deviation of the rotational speed value; for any sliding window, generating a current anomaly threshold for the sliding window based on the local mean and local standard deviation of the current value, and removing current values outside the current anomaly threshold; for any sliding window, generating a rotational speed anomaly threshold for the sliding window based on the local mean and local standard deviation of the rotational speed value, and removing rotational speed values outside the rotational speed anomaly threshold; and using the values that were not removed in the initial time series as candidate time series.
[0043] This application provides a fan alarm system applied to the BMC of a server. The interpolation unit is specifically used for: for any sliding window, determining the normal current value within a first position range in the candidate time series based on the removed current value as the position reference; interpolating the removed current value using polynomial interpolation based on the normal current value; for any sliding window, determining the normal rotational speed value within a second position range in the candidate time series based on the removed rotational speed value as the position reference; and interpolating the removed rotational speed value using polynomial interpolation based on the normal rotational speed value to obtain the target time series.
[0044] This application provides a fan alarm system applied to the BMC of a server, which may further include: The model acquisition module is used to acquire a pre-trained and lightweight target LSTM model before the target time series is input into the target LSTM model by the target time series input module. The model deployment module is used to deploy the target LSTM model.
[0045] This application provides a fan alarm system applied to the BMC of a server. The model acquisition module may include: The factory setting data acquisition unit is used to acquire the factory setting data of the server fan, which includes the fan rotor speed and current value under different wear levels. The training time series generation unit is used to process the factory setting data, generate the training time series, and generate rotor wear labels for the training time series. The partitioning unit is used to divide the training time series into training set, validation set and test set according to the time sequence; The processing unit is used to train the initial LSTM model using the training set, adjust the parameters of the initial LSTM model using the validation set, and test the initial LSTM model using the test set. The filtering unit is used to select the model weights with the optimal loss on the validation set as the weights of the trained initial LSTM model if the loss on the test set no longer decreases in multiple consecutive iterations. The lightweight unit is used to perform lightweight processing on the trained initial LSTM model to obtain the target LSTM model.
[0046] This application provides a fan alarm system applied to the BMC of a server. The alarm detection module may include: The severe wear detection unit is used to detect whether the rotor wear level data indicates severe wear of the fan rotor. If the rotor wear level data indicates that the fan rotor is not severely worn, no alarm is triggered. If the rotor wear level data indicates severe wear of the fan rotor, the unit detects whether the fan rotor is continuously severely worn within a time threshold, where the time threshold is greater than the target duration. If the fan rotor is not continuously severely worn within the time threshold, no alarm is triggered. If the fan rotor is continuously severely worn within the time threshold, an alarm is triggered.
[0047] This application also provides an electronic device and a computer-readable storage medium, both of which have the corresponding effects of the fan alarm method provided in the embodiments of this application. Please refer to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0048] An electronic device provided in this application includes a memory 201 and a processor 202. The memory 201 stores a computer program, and when the processor 202 executes the computer program, it implements the steps of the fan alarm method described in any of the above embodiments.
[0049] Please see Figure 4 Another electronic device provided in this application embodiment may further include: an input port 203 connected to the processor 202 for transmitting commands input from the outside to the processor 202; a display unit 204 connected to the processor 202 for displaying the processing results of the processor 202 to the outside; and a communication module 205 connected to the processor 202 for enabling communication between the electronic device and the outside. The display unit 204 may be a display panel, a laser scanner, or the like; the communication method used by the communication module 205 includes, but is not limited to, Mobile High-Definition Link (MHL), Universal Serial Bus (USB), High-Definition Multimedia Interface (HDMI), wireless connectivity: Wireless Fidelity (WiFi), Bluetooth communication technology, Bluetooth Low Energy communication technology, and communication technology based on IEEE 802.11s.
[0050] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the fan alarm method described in any of the above embodiments.
[0051] The computer-readable storage media involved in this application include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs (compact disc read-only memory), or any other form of storage media known in the art.
[0052] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the fan alarm method described in any of the above embodiments.
[0053] For descriptions of relevant parts of the fan alarm system, electronic device, and computer-readable storage medium provided in this application's embodiments, please refer to the detailed description of the corresponding parts in the fan alarm method provided in this application's embodiments, which will not be repeated here. Furthermore, parts of the technical solutions provided in this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.
[0054] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0055] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A fan alarm method, characterized in that, BMCs used in servers include: Acquire server fan performance data over a target duration, including current and speed values at each time step; The performance data within the target time period is processed to generate the target time series; Input the target time series into the target LSTM model; Obtain rotor wear data output by the target LSTM model; Determine whether to issue an alarm based on the rotor wear data; In response to an alarm determined based on the rotor wear data, a fan alarm operation is executed; The target LSTM model determines the degree of rotor wear based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the degree of rotor wear.
2. The method according to claim 1, characterized in that, The process of processing performance data within the target duration to generate a target time series includes: The performance data within the target time period is sorted according to time order to obtain the initial time series; By using the sliding window 3σ method, abnormal current values and abnormal rotation speed values in the initial time series are removed to obtain candidate time series; The values removed from the candidate time series are interpolated to obtain the target time series.
3. The method according to claim 2, characterized in that, The process of using a sliding window 3σ method to remove abnormal current and abnormal speed values from the initial time series yields a candidate time series, including: The initial time series is divided into a set number of consecutive sliding windows; For any sliding window, generate the local mean and local standard deviation of the current value within the sliding window, and generate the local mean and local standard deviation of the rotational speed value; For any sliding window, an abnormal current threshold is generated based on the local mean and local standard deviation of the current value, and current values outside the abnormal current threshold are removed. For any sliding window, an abnormal speed threshold is generated based on the local mean and local standard deviation of the speed values, and speed values outside the abnormal speed threshold are removed. The values that were not removed from the initial time series are used as candidate time series.
4. The method according to claim 3, characterized in that, The step of interpolating the removed values from the candidate time series to obtain the target time series includes: For any sliding window, using the removed current value as a position reference, the normal current value within the first position range is determined in the candidate time series; based on the normal current value, the removed current value is interpolated using a polynomial interpolation method. For any sliding window, using the removed rotational speed value as the position reference, the normal rotational speed value within the second position range is determined in the candidate time series; based on the normal rotational speed value, the removed rotational speed value is interpolated using polynomial interpolation to obtain the target time series.
5. The method according to claim 1, characterized in that, Before inputting the target time series into the target LSTM model, the method further includes: Obtain a pre-trained and lightweight target LSTM model; Deploy the target LSTM model.
6. The method according to claim 5, characterized in that, The process of obtaining the pre-trained and lightweight target LSTM model includes: Obtain the factory setting data of the server fan, which includes the fan rotor speed and current value under different wear levels; The factory setting data is processed to generate a training time series and a rotor wear label for the training time series. The training time series is divided into a training set, a validation set, and a test set according to the chronological order. The initial LSTM model is trained using the training set, its parameters are tuned using the validation set, and it is tested using the test set. If the test set loss no longer decreases in multiple consecutive iterations, the model weights with the optimal validation set loss are used as the weights of the trained initial LSTM model. The trained initial LSTM model is lightweighted to obtain the target LSTM model.
7. The method according to claim 1, characterized in that, The process of determining whether to issue an alarm based on the rotor wear data includes: The test is to determine whether the rotor wear data indicates severe wear of the fan rotor. If the rotor wear data indicates that the fan rotor is not severely worn, then no alarm will be issued; In response to the rotor wear data indicating severe wear of the fan rotor, the system detects whether the fan rotor is continuously under severe wear within a time threshold, where the time threshold is greater than the target duration. If the fan rotor does not experience continuous heavy wear within the time threshold, no alarm will be triggered. An alarm is triggered in response to continuous heavy wear of the fan rotor within a time threshold.
8. A fan alarm system, characterized in that, BMCs used in servers include: The performance data acquisition module is used to acquire the performance data of the server fan within a target duration, and the performance data includes the current value and speed value at the time step. The performance data processing module is used to process performance data within the target duration and generate the target time series. The target time series input module is used to input the target time series into the target LSTM model; The result acquisition module is used to acquire rotor wear data output by the target LSTM model; The alarm detection module is used to determine whether an alarm should be triggered based on the rotor wear data. The alarm module is used to execute a fan alarm operation in response to an alarm determination based on the rotor wear level data; The target LSTM model determines the degree of rotor wear based on the increase in current value at the same rotational speed; and the greater the increase in current value, the higher the degree of rotor wear.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the fan alarm method as described in any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the fan alarm method as described in any one of claims 1 to 7.