Server fault detection method and device, electronic equipment and storage medium

Abstract

The invention provides a server fault detection method and device, electronic equipment and a storage medium, and relates to the field of server detection.The method comprises the steps that in the running process of a server, sound signals generated by the server are collected; inputting the sound signal into a trained server detection model to obtain an actual operation state of the server; wherein the server detection model is a neural network model, and is obtained by training standard sound signals generated by the server in various preset operation states; determining a fault state of the server according to the actual operation state; the trained neural network model can be used for performing feature extraction on the sound signal actually generated by the server, and the actual operation state corresponding to the server is flexibly determined, so that the fault state of the server can be determined according to the actual operation state. Therefore, the adaptation degree of server fault detection to a complex server operation environment can be improved.

Description

Technical Field

[0001] The present invention relates to the field of server detection, and in particular to a server fault detection method, device, electronic equipment and storage medium. Background Art

[0002] With the continuous development of computer technology, servers have become an important infrastructure in enterprises, and the stable operation of servers is of great significance to the normal operation of enterprises. In related technologies, server fault detection is generally performed by comparing the threshold of the server's operating data. However, this method has low reliability and flexibility for server fault detection. Summary of the invention

[0003] The purpose of the present invention is to provide a server fault detection method, device, electronic device and storage medium, which can use the sound signal actually generated by the server to perform server fault detection, thereby improving the adaptability of server fault detection to complex server operating environments.

[0004] In order to solve the above technical problems, the present invention provides a server fault detection method, comprising:

[0005] During the operation of the server, collecting the sound signal generated by the server;

[0006] Inputting the sound signal into a trained server detection model to obtain the actual operating state of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operating states;

[0007] The fault status of the server is determined according to the actual operating status.

[0008] Optionally, inputting the sound signal into a trained server detection model to obtain an actual operating state of the server includes:

[0009] Encoding the sound signal to obtain input data;

[0010] Inputting the input data into the server detection model, so that the input layer, the hidden layer and the output layer in the server detection model process the input data in sequence to obtain the detection probability of the server corresponding to various preset operating states;

[0011] An actual operating state of the server is determined according to the detection probability.

[0012] Optionally, determining the fault state of the server according to the actual operating state includes:

[0013] Recording changes in the actual operating status of the server;

[0014] A failure state of the server is predicted according to the change.

[0015] Optionally, the training process of the server detection model includes:

[0016] Acquire standard sound signals generated by the server under multiple preset operating states, and divide the standard sound signals into a training set and a validation set according to a preset data division ratio;

[0017] Obtain a neural network model to be trained, randomly initialize the model parameters of the neural network model, and perform multiple rounds of model training processing on the neural network model using the training set and a preset optimization algorithm, until the loss value generated by the neural network model is less than a preset threshold value, and then stop the model training processing; wherein each round of the model training processing includes a forward propagation processing and a back propagation processing, and the loss value is calculated by using the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal;

[0018] Determine the performance index value corresponding to the trained neural network model using the validation set;

[0019] When it is determined that the performance indicator value meets the preset condition, the neural network model is set as the server detection model.

[0020] Optionally, after obtaining the actual running status of the server, the method further includes:

[0021] receiving a manual review result of the actual operating status;

[0022] If the manual review result is that the detection is correct, combining the corresponding relationship between the sound signal and the actual operating state into model update data;

[0023] If the manual review result is a detection error, combining the corresponding relationship between the sound signal and the correct operating state recorded in the manual review result into model update data;

[0024] The server detection model is updated using the model update data.

[0025] Optionally, collecting the sound signal generated by the server includes:

[0026] Collecting original sound signals generated by various hardware components in the server;

[0027] The original sound signal is sequentially subjected to audio enhancement processing, audio noise reduction processing and analog-to-digital conversion processing to obtain the sound signal.

[0028] Optionally, during the operation of the server, collecting a sound signal generated by the server includes:

[0029] When the server executes a self-check process, collecting a self-check sound signal generated by the server;

[0030] The step of inputting the sound signal into a trained server detection model to obtain an actual operating state of the server includes:

[0031] Inputting the self-checking sound signal into the server detection model to obtain the actual self-checking status of the server;

[0032] Determining the fault state of the server according to the actual operating state includes:

[0033] Determining whether the actual self-test state is a self-test abnormal state;

[0034] If the actual self-check state is the self-check abnormal state, it is determined that the server has a self-check fault;

[0035] If the actual self-check state is not the self-check abnormal state, determining whether the server's state retention time in the actual self-check state reaches a preset maximum time;

[0036] If the state is maintained for a period of time that reaches the preset maximum period of time, it is determined that the server has a self-test failure;

[0037] If the state retention time does not reach the preset maximum time, the step of determining whether the actual self-check state is a fault state is entered.

[0038] Optionally, after determining that the server has a self-check failure, the method further includes:

[0039] The server is controlled to restart so that the server re-executes the self-check process.

[0040] The present invention also provides a server fault detection device, comprising:

[0041] A collection module, used to collect sound signals generated by the server during the operation of the server;

[0042] An operation status detection module, used for inputting the sound signal into a trained server detection model to obtain the actual operation status of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operation statuses;

[0043] A fault detection module is used to determine the fault status of the server according to the actual operating status.

[0044] The present invention also provides an electronic device, comprising:

[0045] Memory for storing computer programs;

[0046] A processor is used to implement the server failure detection method as described above when executing the computer program.

[0047] The present invention also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the server fault detection method as described above is implemented.

[0048] The present invention provides a server fault detection method, comprising: collecting sound signals generated by the server during the operation of the server; inputting the sound signals into a trained server detection model to obtain the actual operation status of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under a variety of preset operation statuses; and determining the fault status of the server according to the actual operation status.

[0049] It can be seen that the present invention can firstly actively collect the sound signals generated by the server during the operation of the server, and these sound signals can truly reflect the operation and fault status of the server. Subsequently, the present invention can input these sound signals into the trained server detection model to obtain the actual operation status of these servers, and can determine the fault status of the server according to the actual operation status. Among them, the server detection model is a neural network model trained using standard sound signals generated by the server under a variety of preset operation states. In other words, the present invention can use the trained neural network model to extract features from the sound signals actually generated by the server and flexibly determine the actual operation status corresponding to the server, and can determine the fault status of the server according to the actual operation status, thereby improving the adaptability of server fault detection to complex server operation environments, and can improve the reliability and flexibility of detection. The present invention also provides a server fault detection device, an electronic device, and a computer-readable storage medium, which have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0051] Figure 1 A flow chart of a server fault detection method provided by an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of the structure of a server detection model provided by an embodiment of the present invention;

[0053] Figure 3 A training flow chart of a server detection model provided by an embodiment of the present invention;

[0054] Figure 4 A structural block diagram of a server fault detection device provided by an embodiment of the present invention;

[0055] Figure 5 The present invention provides a structural block diagram of an electronic device. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] With the continuous development of computer technology, servers have become an important infrastructure in enterprises, and the stable operation of servers is of great significance to the normal operation of enterprises. In related technologies, server fault detection is generally performed by comparing the threshold of the server's operating data. However, this method has low reliability and flexibility for server fault detection.

[0058] In view of this, in order to solve the technical problem of how to improve the reliability and flexibility of server fault detection, the present invention can provide a server fault detection method, which can use a preset neural network model to perform server fault detection on the sound signal actually generated by the server, thereby improving the adaptability of server fault detection to complex server operating environments.

[0059] For easier understanding, please refer to Figure 1 , Figure 1A flow chart of a server fault detection method provided by an embodiment of the present invention, the method may include:

[0060] S101. During the operation of a server, a sound signal generated by the server is collected.

[0061] Unlike the related art, this embodiment may not use the operating data generated by the server (such as CPU frequency, memory usage ratio, fan speed, etc.) to detect server faults, but may use the sound signals generated by the server during operation to detect server faults. These sound signals can be generated by various hardware components in the server, such as fans, mechanical hard disks, buzzers, etc. in the server. At the same time, these sound signals are closely related to the operating status of the server. For example, when the server is fully loaded, the server fan will run at full speed and generate a lot of wind noise; conversely, when the server is running at a low load, the server fan will run at a low speed and only generate a small amount of noise. For another example, if a mechanical hard disk fails, it will generate an abnormal sound. For another example, when the server encounters a fault, the buzzer can sound an alarm. Therefore, this embodiment can use the sound signals generated by various hardware components in the server to detect its operating status from the periphery of the server to enhance the flexibility of server fault detection.

[0062] It should be noted that in order to improve the quality of sound signals, a highly sensitive microphone or audio sensor can be set inside or near the server chassis, and used to collect the sound signals generated by the hardware components in the server to ensure that the sound signals generated by the hardware components in the server can be clearly captured. In addition, considering that the environment of the server room is relatively noisy, in order to improve the availability of sound signals, after collecting the original sound signals generated by each hardware component, they can also be subjected to audio enhancement processing, audio noise reduction processing and analog-to-digital conversion processing in turn to obtain a usable sound signal. The intensity of both audio enhancement processing and audio noise reduction processing can be dynamically adjusted according to the noise level of different environments. In other words, a microphone for collecting environmental noise can also be set, and the algorithm parameters used for audio enhancement processing and audio noise reduction processing can be adjusted according to the intensity of the environmental noise collected by the microphone, thereby ensuring the quality of sound signals under different environmental noise conditions.

[0063] Based on this, collecting the sound signal generated by the server may include:

[0064] Step 11: Collecting original sound signals generated by various hardware components in the server;

[0065] Step 12: sequentially perform audio enhancement processing, audio noise reduction processing, and analog-to-digital conversion processing on the original sound signal to obtain the sound signal.

[0066] It should be noted that this embodiment does not limit the specific audio enhancement algorithm, audio noise reduction algorithm, and analog-to-digital conversion method, which can be set according to actual application requirements. For example, the audio noise reduction algorithm can be frequency domain noise reduction, time domain noise reduction, etc.; analog-to-digital conversion can be achieved through an analog-to-digital converter (ADC).

[0067] S102. Input the sound signal into a trained server detection model to obtain the actual operating status of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operating states.

[0068] In this embodiment, in order to detect the actual operating status of the server using the sound signal generated by the server, a server detection model is specially set. The model is a neural network model, and is trained using standard sound signals generated by the server under a variety of preset operating states. For example, six states of power-on, self-check, normal, load, overload, and crash can be pre-set, and the standard sound signals generated by the server in these six states can be collected, and then the neural network model can be trained using the standard sound signals in these six states, so that the model can automatically determine the corresponding operating state according to the input sound signal, thereby achieving the effect of detecting the operating state of the server using the sound signal generated by the server. Since the server detection model can be trained using a large number of labeled sound signals under different preset operating states, the reliability of server fault detection can be improved.

[0069] It should be noted that the present embodiment does not limit the specific structure of the server detection model, which can be set according to actual application requirements. For example, the server detection model can be a BP neural network model (back propagation, multi-layer feedforward neural network), which can include an input layer, a hidden layer and an output layer. For easy understanding of its model structure, please refer to Figure 2 , Figure 2A structural diagram of a server detection model provided for an embodiment of the present invention may include an input layer, a hidden layer, and an output layer. The number of nodes in the input layer may be determined according to the dimension of the sound signal feature after preprocessing. For example, if MFCC (Mel Frequency Cepstrum Coefficient) is used as the sound signal feature, each sample may have 13 or more features, so the number of input layer nodes may be 13. The number of nodes in the hidden layer needs to be determined through experiments. The initial number of nodes can be set to 2 to 3 times the number of nodes in the input layer, or a cross-validation method (such as a k-fold cross-validation method) can be used to find the optimal number of nodes. The number of nodes in the output layer is equal to the number of server operating states. For example, if there are six states (power-on, self-test, normal, load, overload, and crash), the number of nodes in the output layer is 6. It should be noted that after the output of the output layer is processed by Softmax, the detection probability of the server corresponding to various preset operating states can be obtained. The actual operating state of the server can be determined based on the detection probability. For example, the preset operating state corresponding to the maximum detection probability can be determined as the actual operating state, or the preset operating state corresponding to the detection probability with a value greater than the preset threshold can be determined as the actual operating state. It can be set according to actual application requirements.

[0070] It should also be noted that in order to ensure that the server detection model can process sound signals, the sound signals need to be encoded into a data format that the model can process, and the input data is obtained, and then the input data is input into the model for inference processing.

[0071] Based on this, inputting the sound signal into a trained server detection model to obtain the actual operating status of the server may include:

[0072] Step 21: Encode the sound signal to obtain input data;

[0073] Step 22: inputting the input data into the server detection model, so that the input layer, the hidden layer and the output layer in the server detection model process the input data in sequence to obtain the detection probability of the server corresponding to various preset operating states;

[0074] Step 23: Determine the actual operating status of the server according to the detection probability.

[0075] It should also be noted that the present embodiment does not limit specific preset operating states, and in addition to the above six states, other states may be added according to actual application requirements.

[0076] S103: Determine the fault status of the server according to the actual operating status.

[0077] In this embodiment, after determining the actual operating state of the server, the fault state of the server can be determined according to the actual operating state. For example, when the preset operating state includes an overload state and a dead state, if the actual operating state belongs to an overload state or a dead state, it can be determined that the server has encountered a fault. Of course, this embodiment can also record the changes in the actual operating state of the server and predict the fault state of the server based on the changes. The changing state can be the change of the state itself, or it can further include the change of the detection probability of each state. For example, if the server is in a self-check state for a long time, it can be determined that the server has encountered a self-check fault; for another example, although the server has not been determined to be in an overload state, the detection probability of the overload state has always maintained an increasing state in the past multiple time periods, so it can be determined that the server is about to encounter an overload fault. Furthermore, the fault condition of the server can be perceived in advance to improve the reliability of the detection.

[0078] Based on this, determining the fault state of the server according to the actual operating state may include:

[0079] Step 31: Recording changes in the actual operating status of the server;

[0080] Step 32: predicting the failure state of the server according to the change.

[0081] Based on the above embodiments, the present invention can firstly actively collect the sound signals generated by the server during the operation of the server, and these sound signals can truly reflect the operation and fault status of the server. Subsequently, the present invention can input these sound signals into the trained server detection model to obtain the actual operation status of these servers, and can determine the fault status of the server according to the actual operation status. Among them, the server detection model is a neural network model trained using standard sound signals generated by the server under a variety of preset operating conditions. In other words, the present invention can use the trained neural network model to extract features from the sound signals actually generated by the server and flexibly determine the actual operation status corresponding to the server, and can determine the fault status of the server according to the actual operation status, thereby improving the adaptability of server fault detection to complex server operating environments, and can improve the reliability and flexibility of detection.

[0082] Based on the above embodiment, the training process of the server detection model is described in detail below. In one possible scenario, the training process of the server detection model may include:

[0083] S201. Obtain standard sound signals generated by the server under multiple preset operating states, and divide the standard sound signals into a training set and a validation set according to a preset data division ratio.

[0084] In this embodiment, first, the standard sound signals generated by the server under various preset operating states can be collected, and the standard sound signals can be divided into a training set and a verification set according to a preset data division ratio. Among them, the standard sound signals of different preset operating states in the training set are used to train the model, and the standard sound signals of different preset operating states in the verification set are used to verify the model performance. This embodiment does not limit the specific data division ratio. For example, the standard sound signals can be divided into a training set and a verification set according to a ratio of 8:2.

[0085] S202. Obtain a neural network model to be trained, randomly initialize model parameters of the neural network model, and perform multiple rounds of model training on the neural network model using the training set and a preset optimization algorithm, until the loss value generated by the neural network model is less than a preset threshold, and then stop the model training process; wherein each round of the model training process includes forward propagation processing and back propagation processing, and the loss value is calculated by using the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal.

[0086] In this embodiment, the model parameters of the neural network model may be initialized first, and specifically, the weights in the model may be initialized using a Gaussian random distribution or a uniform random distribution to avoid symmetry problems. In addition, the threshold used by the model is also initialized and may be set to a random value with a smaller value.

[0087] Subsequently, the present embodiment can use the training set and the preset optimization algorithm to perform multiple rounds of model training processing on the neural network model until the loss value generated by the neural network model is less than the preset threshold value, and then stop the model training processing. Among them, each round of model training processing includes forward propagation processing and back propagation processing. Forward propagation processing can pass the input data through the input layer to the hidden layer, and then to the output layer to obtain the output result; back propagation processing can calculate the gradient according to the error, and adjust the weights and thresholds of each layer of the model according to the gradient in reverse. When optimizing the model, the preset optimization algorithm can be used to improve the training efficiency, for example, batch gradient descent, stochastic gradient descent, and small batch gradient descent can be used to optimize the model training. Among them, batch gradient descent can use all training samples to calculate the gradient and update the weight; stochastic gradient descent can use one training sample at a time to calculate the gradient and update the weight; small batch gradient descent can use a small batch of training samples at a time to calculate the gradient and update the weight, which combines the advantages of batch and stochastic gradient descent. Finally, the loss value is calculated by using the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal. That is, the loss value is used to measure the difference between the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal. If the difference is large, it means that the model needs to continue training; otherwise, if the difference is small, it means that the model has completed training.

[0088] S203: Determine the performance index value corresponding to the trained neural network model using the verification set.

[0089] In this embodiment, in order to determine whether the trained neural network model meets the application requirements, the performance index value of the model can also be determined using the validation set. This embodiment does not limit the specific performance index value type, which can be set according to actual application requirements. For example, accuracy, precision, recall and F1 score can be used to evaluate the accuracy and performance of the model.

[0090] S204: When it is determined that the performance indicator value meets a preset condition, the neural network model is set as the server detection model.

[0091] In this embodiment, when it is determined that the performance indicator value meets the preset condition, the neural network model is set as the server detection model. It should be noted that this embodiment does not limit the specific preset condition. For example, a preset threshold value can be set for the performance indicator value. When the indicator value of the model meets the requirements of the corresponding preset threshold value, it can be determined that the preset condition is met. For another example, if there are multiple performance indicator values, the preset condition can also include the number of performance indicator values ​​that meet the standard. The preset condition is only determined to be met when multiple performance indicator values ​​of the model meet the standard.

[0092] For easier understanding, please refer to Figure 3 , Figure 3 A training flow chart of a server detection model provided in an embodiment of the present invention.

[0093] Furthermore, considering that the working conditions of the server are more variable and complex, and the standard sound signal is difficult to fully cover the various working conditions of the server, which in turn affects the detection effect of the server detection model. In view of this, this embodiment can also use the sound signal generated by the server during the actual operation process to further update and train the server to improve the detection reliability of the model. Based on this, after obtaining the actual operating status of the server, the method can also include:

[0094] S301: Receive a manual review result for the actual operating status.

[0095] In this embodiment, an operator may manually review the detection result of the model to obtain a manual review result.

[0096] S302: If the manual review result is that the detection is correct, the corresponding relationship between the sound signal and the actual operating state is combined into model update data.

[0097] If the manual review result is that the detection is correct, the corresponding relationship between the sound signal and the actual operating state can be directly combined into model update data, so as to update the model parameters using the model update data.

[0098] S303: If the manual review result is a detection error, the corresponding relationship between the sound signal and the correct operating state recorded in the manual review result is combined into model update data.

[0099] If the manual review result is a detection error, the manual review result must also include the correct operating status. At this time, the correspondence between the sound signal and the correct operating status recorded in the manual review result can be combined into model update data to update the model parameters using the model update data.

[0100] S304: Use the model update data to update the server detection model.

[0101] After obtaining a sufficient amount of model update data, this embodiment can use the data to retrain the server detection model to further improve the detection reliability of the model.

[0102] Based on the above embodiment, the above server fault detection method is introduced below based on a specific self-checking scenario. In a possible situation, the method may also include:

[0103] S401. When the server executes a self-check process, collect a self-check sound signal generated by the server.

[0104] In this embodiment, when the server is started, it usually performs a series of self-check operations, including memory check, hardware status check, network connection test, etc. At the same time, during the self-check process, the server will generate specific sounds, such as fan startup sound, hard disk startup sound, etc. In order to ensure that the server can complete these self-check tasks within the specified time, the present invention can determine whether the self-check process is carried out normally and whether it is completed within the predetermined time by detecting the self-check sound signal generated during the server self-check process (i.e., the sound signal generated by each hardware component during the self-check stage).

[0105] S402: Input the self-checking sound signal into the server detection model to obtain the actual self-checking status of the server.

[0106] It should be pointed out that the actual self-test status includes a normal self-test status and an abnormal self-test status. The normal self-test status indicates that the self-test is normal, and the abnormal self-test status indicates that the self-test is abnormal.

[0107] S403: Determine whether the actual self-check state is a self-check abnormal state.

[0108] S404: If the actual self-check status is the self-check abnormal status, it is determined that the server has a self-check fault;

[0109] S405: If the actual self-check state is not the self-check abnormal state, determine whether the server's state retention time in the actual self-check state reaches a preset maximum time;

[0110] S406: If the state is maintained for a period of time that reaches the preset maximum period of time, it is determined that the server has a self-check failure;

[0111] S407: If the state maintenance time does not reach the preset maximum time, enter the step of determining whether the actual self-check state is a fault state.

[0112] After analyzing the self-test sound signals collected in real time using the server detection model, it can be determined whether the self-test encounters an abnormality or failure, or whether the self-test process is completed within the specified time. If an abnormal sound signal occurs during the self-test process, or the self-test status remains unchanged for a long time exceeding the specified time limit, the monitoring system can automatically issue an alarm to notify the operation and maintenance personnel. The above method can help operation and maintenance personnel discover problems in the self-test process in advance and reduce the occurrence of server startup failures.

[0113] It should be noted that this embodiment does not limit the specific value of the preset maximum duration, which can be set according to actual application requirements.

[0114] Furthermore, when it is determined that a self-check failure occurs in the server, the server may be controlled to be restarted so that the server re-executes the self-check process to prevent the server from being stuck for a long time.

[0115] Based on this, after determining that the server has a self-check failure, the following steps may also be included:

[0116] Step 41: Control the server to restart so that the server re-executes the self-check process.

[0117] The server fault detection device, electronic device, computer program product, and computer-readable storage medium provided in an embodiment of the present invention are introduced below. The server fault detection device, electronic device, computer program product, and computer-readable storage medium described below can be referenced to each other with the server fault detection method described above.

[0118] Please refer to Figure 4 , Figure 4 A structural block diagram of a server fault detection device provided by an embodiment of the present invention, the device may include:

[0119] The acquisition module 401 is used to acquire the sound signal generated by the server during the operation of the server;

[0120] The operation status detection module 402 is used to input the sound signal into a trained server detection model to obtain the actual operation status of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operation states;

[0121] The fault detection module 403 is used to determine the fault status of the server according to the actual operation status.

[0122] Optionally, the running status detection module 402 may include:

[0123] The encoding submodule is used to encode the sound signal to obtain input data;

[0124] A processing submodule, configured to input the input data into the server detection model, so that the input layer, the hidden layer and the output layer in the server detection model process the input data in sequence to obtain the detection probability of the server corresponding to various preset operating states;

[0125] The actual operation status determination submodule is used to determine the actual operation status of the server according to the detection probability.

[0126] Optionally, the fault detection module 403 may include:

[0127] A recording submodule, used to record changes in the actual operating status of the server;

[0128] The prediction submodule is used to predict the fault state of the server according to the change situation.

[0129] Optionally, the device may further include:

[0130] An acquisition module, used for acquiring standard sound signals generated by the server under a plurality of preset operating states, and dividing the standard sound signals into a training set and a validation set according to a preset data division ratio;

[0131] A training module, used to obtain a neural network model to be trained, randomly initialize the model parameters of the neural network model, and perform multiple rounds of model training processing on the neural network model using the training set and a preset optimization algorithm, until the loss value generated by the neural network model is less than a preset threshold value, and then stop the model training processing; wherein each round of the model training processing includes a forward propagation processing and a back propagation processing, and the loss value is calculated by using the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal;

[0132] A verification module, used to determine the performance indicator value corresponding to the trained neural network model using the verification set;

[0133] A setting module is used to set the neural network model as the server detection model when it is determined that the performance indicator value meets a preset condition.

[0134] Optionally, the device may further include:

[0135] A receiving module, used for receiving a manual review result of the actual operating status;

[0136] A first combining module, for combining the corresponding relationship between the sound signal and the actual operating state into model updating data if the manual review result is that the detection is correct;

[0137] A second combining module is used to combine the corresponding relationship between the sound signal and the correct operating state recorded in the manual review result into model update data if the manual review result is a detection error;

[0138] A model updating module is used to update the server detection model using the model updating data.

[0139] Optionally, the acquisition module 401 may include:

[0140] A collection submodule, used for collecting original sound signals generated by various hardware components in the server;

[0141] The sound processing submodule is used to perform audio enhancement processing, audio noise reduction processing and analog-to-digital conversion processing on the original sound signal in sequence to obtain the sound signal.

[0142] Optionally, the acquisition module 401 may include:

[0143] A self-check sound signal collection submodule, used to collect the self-check sound signal generated by the server when the server executes the self-check process;

[0144] The operation status detection module 402 may include:

[0145] A self-check status detection module, used for inputting the self-check sound signal into the server detection model to obtain the actual self-check status of the server;

[0146] The fault detection module 403 may include:

[0147] A first judgment submodule is used to judge whether the actual self-check state is a self-check abnormal state; if the actual self-check state is the self-check abnormal state, it is determined that the server has a self-check fault;

[0148] The second judgment submodule is used to judge whether the state maintenance time of the server in the actual self-check state reaches a preset maximum time if the actual self-check state is not the self-check abnormal state; if the state maintenance time reaches the preset maximum time, it is determined that the server has a self-check fault; if the state maintenance time does not reach the preset maximum time, enter the step of judging whether the actual self-check state is a fault state.

[0149] Optionally, the device may further include:

[0150] The restart control module is used to control the server to restart so that the server re-executes the self-check process.

[0151] Please refer to Figure 5 , Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. The embodiment of the present invention provides an electronic device 10, including a processor 11 and a memory 12; wherein the memory 42 is used to store a computer program; and the processor 11 is used to execute the server fault detection method provided in the aforementioned embodiment when executing the computer program.

[0152] For the specific process of the above-mentioned server fault detection method, reference may be made to the corresponding contents provided in the aforementioned embodiments, which will not be repeated here.

[0153] Furthermore, the memory 12 as a carrier for storing resources may be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the storage method may be temporary storage or permanent storage.

[0154] In addition, the electronic device 10 also includes a power supply 13, a communication interface 14, an input / output interface 15 and a communication bus 16; wherein the power supply 13 is used to provide working voltage for each hardware device on the electronic device 10; the communication interface 14 can create a data transmission channel between the electronic device 10 and an external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present invention, and is not specifically limited here; the input / output interface 15 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0155] An embodiment of the present invention further provides a computer program product, including a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the server fault detection method as described in the above embodiment is implemented.

[0156] Since the embodiments of the computer program product part correspond to the embodiments of the server fault detection method part, please refer to the description of the embodiments of the server fault detection method part for the embodiments of the computer program product part, which will not be repeated here.

[0157] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the server fault detection method described in the above embodiment is implemented.

[0158] Since the embodiments of the computer-readable storage medium part correspond to the embodiments of the server fault detection method part, please refer to the description of the embodiments of the server fault detection method part for the embodiments of the storage medium part, which will not be repeated here.

[0159] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0160] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0161] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0162] The above is a detailed introduction to a server fault detection method, device, electronic device and storage medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A server failure detection method, characterized in that: include: During the operation of the server, collecting the sound signal generated by the server; Inputting the sound signal into a trained server detection model to obtain the actual operating state of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operating states; The fault status of the server is determined according to the actual operating status.

2. The server failure detection method according to claim 1, characterized in that: The step of inputting the sound signal into a trained server detection model to obtain an actual operating state of the server includes: Encoding the sound signal to obtain input data; Inputting the input data into the server detection model, so that the input layer, the hidden layer and the output layer in the server detection model process the input data in sequence to obtain the detection probability of the server corresponding to various preset operating states; An actual operating state of the server is determined according to the detection probability.

3. The server failure detection method according to claim 1, characterized in that: The training process of the server detection model includes: Acquire standard sound signals generated by the server under multiple preset operating states, and divide the standard sound signals into a training set and a validation set according to a preset data division ratio; Obtain a neural network model to be trained, randomly initialize the model parameters of the neural network model, and perform multiple rounds of model training processing on the neural network model using the training set and a preset optimization algorithm, until the loss value generated by the neural network model is less than a preset threshold value, and then stop the model training processing; wherein each round of the model training processing includes a forward propagation processing and a back propagation processing, and the loss value is calculated by using the predicted operating state output by the neural network model according to the standard sound signal in the training set and the preset operating state corresponding to the same standard sound signal; Determine the performance index value corresponding to the trained neural network model using the validation set; When it is determined that the performance indicator value meets the preset condition, the neural network model is set as the server detection model.

4. The server failure detection method according to claim 1, characterized in that: After obtaining the actual running status of the server, the method further includes: receiving a manual review result of the actual operating status; If the manual review result is that the detection is correct, combining the corresponding relationship between the sound signal and the actual operating state into model update data; If the manual review result is a detection error, the corresponding relationship between the sound signal and the correct operating state recorded in the manual review result is combined into model update data; and the server detection model is updated using the model update data.

5. The server failure detection method according to claim 1, characterized in that: The collecting of the sound signal generated by the server includes: Collecting original sound signals generated by various hardware components in the server; The original sound signal is sequentially subjected to audio enhancement processing, audio noise reduction processing and analog-to-digital conversion processing to obtain the sound signal.

6. The server failure detection method according to any one of claims 1 to 5, characterized in that: During the operation of the server, collecting the sound signal generated by the server includes: When the server executes a self-check process, collecting a self-check sound signal generated by the server; The step of inputting the sound signal into a trained server detection model to obtain an actual operating state of the server includes: Inputting the self-checking sound signal into the server detection model to obtain the actual self-checking status of the server; Determining the fault state of the server according to the actual operating state includes: Determining whether the actual self-test state is a self-test abnormal state; If the actual self-check state is the self-check abnormal state, it is determined that the server has a self-check fault; If the actual self-check state is not the self-check abnormal state, determining whether the server's state retention time in the actual self-check state reaches a preset maximum time; If the state is maintained for a period of time that reaches the preset maximum period of time, it is determined that the server has a self-test failure; If the state retention time does not reach the preset maximum time, the step of determining whether the actual self-check state is a fault state is entered.

7. The server failure detection method according to claim 6, characterized in that: After determining that the server has a self-check failure, the method further includes: The server is controlled to restart so that the server re-executes the self-check process.

8. A server fault detection device, characterized in that: include: A collection module, used to collect sound signals generated by the server during the operation of the server; An operation status detection module, used for inputting the sound signal into a trained server detection model to obtain the actual operation status of the server; wherein the server detection model is a neural network model, and is trained using standard sound signals generated by the server under multiple preset operation statuses; A fault detection module is used to determine the fault status of the server according to the actual operating status.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the server failure detection method according to 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 computer-executable instructions, and when the computer-executable instructions are loaded and executed by the processor, the server fault detection method according to any one of claims 1 to 7 is implemented.

Images