Device state detection method, computer device, and storage medium

Abstract

The application provides a device state detection method, a computer device and a storage medium, the method comprising: acquiring sensor parameters collected by a sensor of a device, and acquiring sound data recorded in a running process of the device; determining deviation information for representing performance changes of the device according to the sensor parameters and the sound data; and determining a running state of the device according to the deviation information. The application aims to improve the detection accuracy of the device state.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more particularly to a device status detection method, computer device, and storage medium. Background Technology

[0002] During the use of electronic and mechanical equipment, a series of functional degradations are inevitable. Some functional degradations do not cause equipment failure, but others can lead to malfunctions, significantly reducing production efficiency and causing economic losses or safety accidents. Therefore, how to accurately detect equipment status and prevent overall equipment failure or malfunction has become a problem worthy of research and discussion in the field of reliability. Summary of the Invention

[0003] The main objective of this application is to provide a device status detection method, a computer device, and a storage medium, with the aim of improving the accuracy of device status detection.

[0004] In a first aspect, this application provides a method for detecting the condition of a device, comprising:

[0005] Acquire sensor parameters collected by the device's sensors, and acquire sound data recorded during the device's operation;

[0006] Based on the sensor parameters and the sound data, deviation information used to characterize changes in the device performance is determined;

[0007] The operating status of the equipment is determined based on the deviation information.

[0008] Secondly, this application also provides a computer device, the computer device including a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of the device state detection method as described above.

[0009] Thirdly, this application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of the device state detection method described above.

[0010] This application provides a device status detection method, a computer device, and a storage medium. The method acquires sensor parameters collected by the device's sensors and obtains sound data recorded during device operation. Based on the sensor parameters and sound data, deviation information characterizing changes in device performance is determined. Based on the deviation information, the operating status of the device is determined. This application's embodiments can accurately detect changes in device performance using sensor parameters and sound data, thereby determining the device's operating status and improving the accuracy of device status detection. Furthermore, the operating status of the device can be identified before a malfunction occurs, enabling preventative measures to be taken to avoid overall device failure or malfunction. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the steps of a device status detection method provided in an embodiment of this application.

[0013] Figure 2 A flowchart illustrating the steps of another device status detection method provided in this application embodiment;

[0014] Figure 3 A flowchart illustrating the steps of another device status detection method provided in this application embodiment;

[0015] Figure 4 This is a schematic diagram of the structure of the device status detection model provided in the embodiments of this application;

[0016] Figure 5 A schematic diagram of a scenario for implementing the device status detection method provided in this embodiment;

[0017] Figure 6 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0018] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. 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, 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] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0021] This application provides a device status detection method, a computer device, and a storage medium. The device status detection method can be applied to a terminal device or a server. The terminal device can be an electronic device such as a mobile phone, tablet computer, laptop computer, desktop computer, personal digital assistant, or wearable device. The server can be a single server or a server cluster composed of multiple servers.

[0022] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0023] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of a device status detection method provided in an embodiment of this application.

[0024] like Figure 1 As shown, the device status detection method includes steps S101 to S103.

[0025] Step S101: Obtain the sensor parameters collected by the device's sensors and obtain the sound data recorded during the device's operation.

[0026] In this step, the equipment includes electronic devices and mechanical devices. For example, the equipment can be electronic devices such as mobile phones, tablets, laptops, translators, desktop computers, learning machines, personal digital assistants, and wearable devices, or it can be mechanical devices such as robots, engineering machinery, agricultural machinery, instruments, and electrical machinery.

[0027] In this step, there may be one or more sensors. For example, the sensors include at least one of the following: temperature sensor, humidity sensor, voltmeter, ammeter, pressure sensor, speed sensor, acceleration sensor, radiation sensor, position sensor, etc.

[0028] In this step, sensor parameters can be a data sequence collected by a single sensor at multiple sampling times, or a data set collected by multiple sensors at multiple sampling times. Sensor parameters can include operating parameters of the device and data from the surrounding environment. For example, sensor parameters may include at least one of the following: temperature, humidity, current, voltage, pressure, speed, acceleration, radiation, and position.

[0029] In this step, the sound data refers to the audio data generated and recorded during the operation of the device. This sound data can be collected by devices such as microphones. It should be noted that the device will emit noise of a certain volume during operation, and the recorded sound data can characterize the operating status of the device.

[0030] In this step, the sensor parameters and sound data can be collected under the same operating conditions, such as when the equipment is rotating forward or backward. The collection time periods for the sensor parameters and sound data can also be the same. For example, while collecting sensor parameters through each sensor, the sound emitted by the equipment during operation can also be collected simultaneously to obtain sensor parameters and sound data with relatively consistent time dimensions.

[0031] Step S102: Determine the deviation information used to characterize changes in device performance based on sensor parameters and sound data.

[0032] It's important to note that equipment undergoes a series of degradation processes from its initial malfunction to complete failure or malfunction. Therefore, if this degradation can be detected effectively, maintenance personnel can perform further inspection and maintenance, preventing the equipment or the entire system from failing or malfunctioning. Thus, equipment failure only becomes possible after a long period of degradation, reaching a certain level of deterioration. For a device composed of multiple components, the failure of one component does not necessarily mean that the entire device will immediately fail.

[0033] In this step, the equipment's performance changes due to functional degradation or deterioration during use, or functional optimization or improvement resulting from maintenance. This step utilizes sensor parameters and sound data to accurately detect these performance changes, thereby determining the equipment's operating status and improving the accuracy of equipment status monitoring.

[0034] In this step, the deviation information is used to characterize changes in device performance. These changes refer to the difference between the current device performance, determined by sensor parameters and sound data, and a preset performance. The preset performance could be, for example, the device's performance at the time of manufacture, or it could be the performance after adjustments; this embodiment does not specifically limit this.

[0035] In this step, the deviation information can be determined based on the sensor parameters, sound data and the sensor parameters and sound data corresponding to the preset performance. Of course, it can also be determined in other ways, such as by processing the sensor parameters and sound data through the device status detection model and then outputting the result. This embodiment does not make specific limitations on this.

[0036] Step S103: Determine the operating status of the equipment based on the deviation information.

[0037] In this step, after identifying the deviation information used to characterize changes in equipment performance, the operating status of the equipment can be accurately determined based on this deviation information. This step allows for the identification of the equipment's operating status before a failure occurs, enabling preventative measures to be taken to avoid overall equipment failure or malfunction.

[0038] In this step, the equipment's operating status includes normal and abnormal states. A normal state means the equipment operates without any abnormalities, while an abnormal state means the equipment operates with abnormalities. Abnormal states can include component abnormalities, data abnormalities, and environmental abnormalities. Component abnormalities refer to one or more components malfunctioning; data abnormalities refer to abnormal operating parameters collected by sensors during equipment operation; and environmental abnormalities refer to abnormal ambient environmental data collected by sensors during equipment operation.

[0039] The device status detection method provided in the above embodiments acquires sensor parameters collected by the device's sensors and obtains sound data recorded during device operation; based on the sensor parameters and sound data, it determines deviation information used to characterize changes in device performance; and based on the deviation information, it determines the device's operating status. By using sensor parameters and sound data, it can accurately detect changes in device performance, thereby determining the device's operating status, which greatly improves the accuracy of device status detection. Furthermore, it can identify the device's operating status before a failure occurs, enabling preventative measures to be taken to avoid overall device failure or malfunction.

[0040] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the steps of another device status detection method provided in an embodiment of this application.

[0041] like Figure 2 As shown, the device status detection method includes steps S201 to S205.

[0042] Step S201: Obtain the sensor parameters collected by the device's sensors and obtain the sound data recorded during the device's operation.

[0043] In one embodiment, the receiving device sends sensor parameters and sound data, which may be collected by various sensors and microphones in the device.

[0044] In one embodiment, sensor parameters and sound data are retrieved from a database, which may be pre-stored. The database includes cloud databases and local databases; this embodiment does not specifically limit the types of databases used.

[0045] Step S202: Determine the first degree of matching between the sensor parameters and the preset sensor parameters.

[0046] The preset sensor parameters can be one or more, and the first matching degree can be a parameter such as similarity or variance. For example, the preset sensor parameters can be sensor parameters that have been calibrated at the factory, or the preset sensor parameters can include sensor parameters corresponding to different performance characteristics of the device. This embodiment does not specifically limit this.

[0047] In one embodiment, a preset sensor parameter is used to characterize the sensor parameters corresponding to the device in normal or abnormal states. The similarity between the sensor parameters and the preset sensor parameter is calculated as a first matching degree.

[0048] In one embodiment, multiple preset sensor parameters are used to characterize sensor parameters corresponding to different performance characteristics of the device. The similarity between the sensor parameter and the multiple preset sensor parameters is calculated to obtain multiple similarity scores; these multiple similarities are determined as a first matching degree. Alternatively, the variance between the sensor parameter and the multiple preset sensor parameters is calculated, and this variance is determined as the first matching degree.

[0049] Step S203: Determine the second matching degree between the sound data and the preset sound data.

[0050] The preset sound data can be one or more, and the second matching degree can be a parameter such as similarity or variance. For example, the preset sound data can be the sound data that was calibrated when the device left the factory, or the preset sound data can include the sound data corresponding to different performance levels of the device. This embodiment does not make specific limitations in this regard.

[0051] In one embodiment, a preset sound data set is provided, which represents the sound data corresponding to the device in normal or abnormal states. The similarity between the current sound data and the preset sound data is calculated as a second matching degree.

[0052] In one embodiment, multiple preset sound data are used to characterize sound data corresponding to different device performance. The similarity between the sound data and the multiple preset sound data is calculated to obtain multiple similarity scores; these multiple similarity scores are then determined as a second matching degree. Alternatively, the variance between the sound data and the multiple preset sound data is calculated, and this variance is determined as the second matching degree.

[0053] Step S204: Determine the deviation information used to characterize the changes in equipment performance based on the first matching degree and the second matching degree.

[0054] It should be noted that the first and second matching degrees can be used to determine the degree of deviation between the sensor parameters and the preset sensor parameters, as well as the degree of deviation between the sound data and the preset sound data. Since the preset sensor parameters and preset sound data can characterize the operating performance of the device, the deviation information can be accurately determined based on the first and second matching degrees.

[0055] In one embodiment, a target matching degree is determined based on a first matching degree and a second matching degree; the target matching degree is then used to determine deviation information used to characterize changes in device performance. The first and second matching degrees are data such as similarity or variance, and the target matching degree can be determined using methods such as the average method or the weighted average method. There is a correspondence between the target matching degree and the deviation information used to characterize changes in device performance, or the difference between the target matching degree and a preset matching degree can be used as deviation information. Therefore, deviation information can be accurately determined based on the target matching degree.

[0056] For example, both the first matching degree and the second matching degree are used to characterize the similarity between the current operating state and the abnormal state of the device. The average of the first matching degree and the second matching degree is calculated to obtain the target matching degree; the difference between the target matching degree and the preset matching degree is calculated to obtain the deviation information used to characterize changes in device performance. The preset matching degree can be set according to actual conditions; for example, the preset matching degree can be used to characterize the similarity between the current operating state of the device in a normal state and an abnormal state.

[0057] In one embodiment, a first difference between a first matching degree and a first preset matching degree is calculated, and a second difference between a second matching degree and a second preset matching degree is calculated. The first and second differences are used to determine deviation information characterizing changes in device performance. The first and second preset matching degrees can be set according to actual conditions. The first and second differences in the deviation information can effectively distinguish the specific type of abnormal state when the device's operating state is abnormal, thereby improving the accuracy of detecting abnormal device states.

[0058] Step S205: Determine the operating status of the equipment based on the deviation information.

[0059] The operating status of equipment includes normal and abnormal states. Abnormal states can include component malfunctions, data anomalies, and environmental anomalies. It's important to note that deviation information allows for accurate determination of the equipment's operating status. Therefore, the operating status can be identified before a malfunction occurs, enabling preventative measures to be taken to avoid overall equipment failure or malfunction, thus improving equipment safety.

[0060] For example, the aforementioned first matching degree and second matching degree are both used to characterize the similarity or variance between the current operating state and the abnormal state of the device, and the deviation information is the difference between the target matching degree and the preset matching degree; if the difference is greater than or equal to the preset difference, the operating state of the device is determined to be an abnormal state; if the difference is less than the preset difference, the operating state of the device is determined to be a normal state.

[0061] For example, the deviation information includes the aforementioned first difference and second difference. If the first difference is greater than or equal to a first preset difference, or the second difference is greater than or equal to a second preset difference, then the device's operating state is determined to be abnormal. If the difference is less than the first preset difference and the second difference is less than the second preset difference, then the device's operating state is determined to be normal. In some examples, if the device's operating state is abnormal, the specific type of abnormal state, such as component abnormality, data abnormality, or environmental abnormality, can be determined based on the first and second differences, thereby improving the accuracy of detecting abnormal device states.

[0062] The device status detection method provided in the above embodiments determines a first matching degree between sensor parameters and preset sensor parameters, and a second matching degree between sound data and preset sound data. This allows for the determination of deviation information used to characterize changes in device performance based on the first and second matching degrees. The deviation information is then used to determine the operating status of the device, improving the accuracy of device status detection. This facilitates timely and appropriate handling of the device based on its status, preventing device failure or malfunction, and greatly enhancing device reliability.

[0063] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the steps of another device status detection method provided in this application embodiment.

[0064] like Figure 3 As shown, the device status detection method includes steps S301 to S304.

[0065] Step S301: Obtain the sensor parameters collected by the device's sensors and obtain the sound data recorded during the device's operation.

[0066] The sensor parameters and sound data can be collected under the same operating conditions, such as when the equipment is rotating forward or backward, to ensure the accuracy of subsequent data processing calculations.

[0067] The sensor parameters and sound data can be collected during the same time period. For example, while collecting sensor parameters from various sensors, the sound emitted by the device during operation can also be collected simultaneously, resulting in sensor parameters and sound data collected during the same time period. Using sensor parameters and sound data collected during the same time period for subsequent data processing helps improve the accuracy of detecting the device's status.

[0068] Step S302: Input the sensor parameters and sound data into the trained device state detection model for processing to obtain the reconstruction error.

[0069] The trained device status detection model can be a convolutional neural network, a recurrent neural network, or a convolutional recurrent neural network. Reconstruction error refers to the mean square error between the model's output value and the original input. By inputting sensor parameters and sound data into the trained device status detection model and processing them, the reconstruction error is obtained, thereby enabling multimodal device status detection and improving the detection performance.

[0070] In one embodiment, the training process of the trained device state detection model includes: acquiring multiple training samples, which include sample sensor parameters and sample sound data; iteratively training the device state detection model using the multiple training samples until the device state detection model converges, thus obtaining the trained device state detection model. The convergence conditions of the device state detection model can be set according to actual conditions, such as the number of iterations reaching a preset number, the iteration duration exceeding a preset duration, or the model loss value being less than a preset loss value.

[0071] In one embodiment, acquiring multiple training samples includes: cleaning the sensor parameter sets collected by multiple sensors; performing dimensionless processing on the multiple sensor parameter sets after data cleaning; generating a correlation coefficient matrix of the multiple sensor parameter sets after dimensionless processing; selecting a target sensor parameter set from the multiple sensor parameter sets after data cleaning based on multiple correlation coefficients in the correlation coefficient matrix, and determining the sensor parameters in the target sensor parameter set as sample sensor parameters; and performing high-pass filtering on the sound data recorded during device operation to obtain sample sound data.

[0072] In this training sample, the sensor parameters can be multiple sensor parameters collected by one or more sensors, and the collection period of these sensor parameters can match the sound data recorded during device operation. It should be noted that the multiple training samples are obtained through preprocessing, including sensor data cleaning, sensor feature dimensionality reduction, and audio frequency band selection. Sensor feature dimensionality reduction involves removing sensor parameter pairs with correlation coefficients greater than a preset coefficient, thereby eliminating highly correlated sensor parameters. The preset coefficient is, for example, 0.9. Audio frequency band selection involves selecting high-frequency bands with more important information and removing low-frequency bands with more irrelevant noise. Experiments have shown that most of the important information in the sound generated by the device is concentrated in the high-frequency range, while the low-frequency range contains more irrelevant noise. Therefore, the original sound data can be passed through a high-pass filter to retain the sample sound data in the high-frequency range. These preprocessing methods remove redundant dimensions in the features, retain important features, and make the subsequent modeling representation of the data distribution more accurate, achieving better results in the device state detection task.

[0073] In one embodiment, data cleaning is performed on the sensor parameter sets collected by multiple sensors, including: calculating the relative change amplitude between every two adjacent sensor parameters in each sensor parameter set; determining a target sensor parameter from the sensor parameter set based on the relative change amplitude between every two adjacent sensor parameters, wherein the relative change amplitude of the target sensor parameter is greater than or equal to a preset change amplitude; and removing the target sensor parameter from the sensor parameter set to obtain a data-cleaned sensor parameter set. The relative change amplitude is, for example, 20%. By removing bad points whose relative change amplitude exceeds the preset change amplitude, short-term drastic fluctuations in sensor parameters in each sensor parameter set are reduced, ensuring the short-term stability of sensor parameter changes, which is beneficial to improving the robustness and reliability of the equipment condition detection model.

[0074] In one embodiment, the equipment condition detection model includes a feature processing module, an anomaly correction module, and an error calculation module. The model is iteratively trained using multiple training samples, including: inputting training samples, including sample sensor parameters and sample sound data, into the feature processing module for feature extraction and concatenation to obtain a concatenated vector; inputting the concatenated vector into the anomaly correction module for reconstruction processing to obtain a reconstructed vector; inputting the concatenated vector and the reconstructed vector into the error calculation module for error analysis to obtain the mean square error between the concatenated vector and the reconstructed vector; determining the model parameters of the equipment condition detection model based on the mean square error; and iteratively training the equipment condition detection model based on the model parameters.

[0075] For example, the input to the device status detection model is training samples, which include sample sensor parameters and sample sound data. A feature processing module extracts features from the sample sensor parameters and sample sound data in the training samples to obtain corresponding sensor feature parameters and acoustic feature parameters, which are then concatenated. The acoustic feature parameters, for example, are acoustic features obtained through filter bank processing. The feature processing module can be composed of a Long Short-Term Memory (LSTM) network, such as a two-layer LSTM. Assuming the sensor feature parameters are T*F1 dimensional, and the time dimension of the acoustic feature parameters is aligned with the sensor feature parameters, then the acoustic feature parameters are T*F2 dimensional, resulting in a T*(F1+F2) dimensional concatenated vector.

[0076] In one embodiment, the anomaly correction module includes a feature reconstruction layer filled with multiple positive samples, which are correctly labeled training samples. Before each round of training of the device state detection model, the method further includes: clearing multiple first positive samples in the feature reconstruction layer; obtaining multiple second positive samples and filling the feature reconstruction layer with the multiple second positive samples.

[0077] For example, the feature reconstruction layer is an N*T*F dimensional matrix M, where N represents the number of positive samples, T represents the time dimension, and F represents the number of nodes. At the beginning of each iteration during training, the positive samples in the feature reconstruction layer are cleared, and then a new round of positive samples is used to fill them in. For example, N normal training samples (accurately labeled) are used as input and sequentially processed by LSTM1 and LSTM2 in the feature processing module to obtain a T*F dimensional vector (the second positive sample), which is then filled into the matrix M of the feature reconstruction layer until the filling is complete. This makes the representation in the feature reconstruction layer more accurate.

[0078] It should be noted that by adding a feature reconstruction layer to the device state detection model and periodically clearing and updating the positive samples in the feature reconstruction layer during the training iteration, the element information of the accurately labeled training samples is recorded, and the elements are made more accurate through continuous updates.

[0079] In one embodiment, the anomaly correction module includes a feature reconstruction layer and an attention layer. The concatenated vector is input to the attention layer to calculate the attention coefficients, obtaining the attention coefficients corresponding to the concatenated vector. The concatenated vector and its corresponding attention coefficients are then input to the feature reconstruction layer to reconstruct the feature vector, obtaining the reconstructed vector.

[0080] It's important to note that the attention coefficients calculated through the attention layer are used to reconstruct the spliced ​​vector using positive samples from the feature reconstruction layer. If the input training sample is an anomalous sample mislabeled as normal, the reconstructed features, which closely resemble normal samples, will be reconstructed based on the positive samples. The device status detection model trained in this way, given a test sample (sensor parameters and sound data), will only use a limited number of positive samples recorded in the feature reconstruction layer to reconstruct the test sample. As a result, the network reconstruction will be closer to normal samples; during testing, the reconstruction error for normal samples is small, while the reconstruction error for anomalous samples is large. Therefore, the reconstruction error can be used as a standard for detecting whether the device status is abnormal.

[0081] For example, the formula for calculating the attention coefficient is: Where d is the cosine distance. Z represents the concatenated vector, Zt represents the concatenated vector at the t-th sampling time point, and m it represents the training sample at the t-th sampling time point in the i-th feature reconstruction layer, and N represents the number of training samples in the feature reconstruction layer.

[0082] For example, the formula for reconstructing the feature vector is: Where Zt is the reconstructed vector. It should be noted that the more similar the concatenated vector Z is to the training samples in the feature reconstruction layer, the higher the attention coefficient w corresponding to that concatenated vector. it The larger the value, the higher the attention coefficient w. it The value of is in the range of 0 to 1, so the closer the reconstructed vector is to the concatenated vector, the better.

[0083] For example, the formula for calculating the mean square error between the concatenated vector and the reconstructed vector is as follows: Here, MSELoss represents the mean squared error. The concatenated vector and the reconstructed vector are input into the error calculation module for error analysis, i.e., Z′. T With Z T The MSE distance is calculated and used as a loss function to update the model parameters of the device condition detection model.

[0084] In one embodiment, the trained device state detection model includes a feature processing module, an anomaly correction module, and an error calculation module; sensor parameters and sound data are input into the feature processing module for feature extraction and splicing to obtain a first feature vector; the first feature vector is input into the anomaly correction module for reconstruction processing to obtain a second feature vector.

[0085] The anomaly correction module includes an attention layer and a feature reconstruction layer. The first feature vector is input to the attention layer to calculate the attention coefficient, thereby obtaining the attention coefficient corresponding to the first feature vector. The first feature vector and the attention coefficient corresponding to the first feature vector are input to the feature reconstruction layer to reconstruct the feature vector, thereby obtaining the second feature vector. The first feature vector and the second feature vector are input to the error calculation module to perform error analysis, thereby obtaining the reconstruction error.

[0086] It should be noted that the methods for obtaining the first and second feature vectors can refer to the corresponding embodiments of the concatenated vectors and reconstructed vectors described above. The implementation method for inputting the first and second feature vectors into the error calculation module for error analysis to obtain the reconstruction error can refer to the corresponding embodiment of calculating the mean square error between the concatenated vectors and reconstructed vectors described above; this embodiment will not be repeated here.

[0087] For example, such as Figure 4 As shown, the device status detection model includes a feature processing module 10, an anomaly correction module 20, and an error calculation module 30. Both the feature processing module 10 and the error calculation module 30 include two-layer LSTM, and the anomaly correction module 20 includes an attention layer and a feature reconstruction layer. The model input consists of sensor parameters and sound data, and the model output is the reconstruction error. Through the feature processing module 10, the anomaly correction module 20, and the error calculation module 30, the reconstruction error of the first and second feature vectors can be accurately output.

[0088] Step S303: Use the reconstruction error as deviation information to characterize changes in equipment performance.

[0089] Step S304: Determine the operating status of the equipment based on the deviation information.

[0090] The operating status of the equipment includes normal status and abnormal status. Abnormal status of the equipment can include component abnormal status, data abnormal status, environmental abnormal status, etc.

[0091] In one embodiment, the deviation information is, for example, the reconstruction error between the first feature vector and the second feature vector. If the reconstruction error is greater than or equal to a preset error threshold thresh, the input test sample (sensor parameters and sound data) is considered abnormal, i.e., the device's operating state is abnormal; conversely, if the reconstruction error is less than the preset error threshold thresh, the device's operating state is considered normal. After the abnormal test sample passes through the anomaly correction module, it can only reconstruct features close to those of a normal sample based on positive samples, resulting in a larger MSE distance and making it easier to detect.

[0092] Please refer to Figure 5 , Figure 5A schematic diagram of a scenario for implementing the device status detection method provided in this embodiment.

[0093] like Figure 5 As shown, each sensor in device 100 acquires sensor parameters, and device 100 also acquires sound data recorded during its operation. Device 100 sends the acquired sensor parameters and sound data to server 200. Server 200 stores a trained device state detection model. Server 200 processes the sensor parameters and sound data by inputting them into the trained device state detection model to obtain reconstruction error. Server 200 uses the reconstruction error as deviation information to characterize changes in device performance and uses the deviation information to determine the operating state of device 100.

[0094] It should be noted that current equipment status detection models are primarily linear, thus limiting their ability to represent data distributions and making them inadequate for accurate equipment status detection in complex business scenarios. Furthermore, current equipment status detection models rely heavily on the accuracy of the training data annotations. If a small amount of abnormal data is mixed into the training data, the model may incorrectly adapt to the abnormal data, treating similar abnormal data as normal data. This significantly reduces its sensitivity to abnormal data, leading to missed alarms in actual system use.

[0095] The device status detection method provided in the above embodiments processes sensor parameters and sound data into a trained device status detection model to obtain reconstruction error. The reconstruction error is used as deviation information to characterize changes in device performance, and the deviation information is used to determine the operating status of the device. This improves the accuracy of device status detection, facilitates timely and appropriate handling of the device based on its status, avoids device failure or malfunction, and greatly improves device reliability.

[0096] Furthermore, the equipment status detection method provided in the above embodiments utilizes the characteristics of the anomaly correction module in the equipment status detection model to effectively eliminate the adverse effects of a small amount of incorrectly labeled data on the model, and avoids the model from overfitting abnormal data, such as multimodal sensor parameter features and sound features, to detect equipment anomalies, thereby greatly improving the detection effect.

[0097] Please see Figure 6 , Figure 6 This is a schematic block diagram illustrating the structure of a computer device provided in an embodiment of this application. The computer device may be a server or a terminal device.

[0098] like Figure 6As shown, the computer device 400 includes a processor 402 and a memory 403 connected via a system bus 401, wherein the memory 403 may include a non-volatile storage medium and internal memory.

[0099] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions that, when executed, cause the processor 402 to perform any device status detection method.

[0100] The processor 402 provides computing and control capabilities to support the operation of the entire computer device 400.

[0101] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor 402, the processor 402 can execute any device status detection method.

[0102] The computer device 400 may also include a network interface for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 400 to which the present application is applied. The specific computer device 400 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0103] It should be understood that processor 402 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic components, discrete gate or transistor logic components, discrete hardware components, etc. Among these, the general-purpose processor can be a microprocessor or any conventional processor.

[0104] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0105] Acquire sensor parameters collected by the device's sensors, and acquire sound data recorded during the device's operation;

[0106] Based on the sensor parameters and the sound data, deviation information used to characterize changes in the device performance is determined;

[0107] The operating status of the equipment is determined based on the deviation information.

[0108] In one embodiment, when the processor determines deviation information characterizing changes in device performance based on the sensor parameters and the sound data, it performs the following:

[0109] Determine the first degree of matching between the sensor parameters and the preset sensor parameters;

[0110] Determine the second matching degree between the sound data and the preset sound data;

[0111] The deviation information is determined based on the first matching degree and the second matching degree.

[0112] In one embodiment, when the processor determines the deviation information characterizing the device performance change based on the sensor parameters and the sound data, it is configured to:

[0113] The sensor parameters and the sound data are input into a trained device state detection model for processing to obtain the reconstruction error.

[0114] The reconstruction error is used as the deviation information.

[0115] In one embodiment, the trained device state detection model includes a feature processing module, an anomaly correction module, and an error calculation module; the processor, when processing the sensor parameters and the sound data by inputting them into the trained device state detection model to obtain the reconstruction error, is configured to:

[0116] The sensor parameters and the sound data are input into the feature processing module for feature extraction and concatenation to obtain a first feature vector.

[0117] The first feature vector is input into the anomaly correction module for reconstruction processing to obtain the second feature vector;

[0118] The first feature vector and the second feature vector are input into the error calculation module for error analysis to obtain the reconstruction error.

[0119] In one embodiment, when the processor implements the anomaly correction module, which includes an attention layer and a feature reconstruction layer, and inputs the first feature vector into the anomaly correction module for reconstruction processing to obtain the second feature vector, it is used to implement:

[0120] The first feature vector is input into the attention layer to calculate the attention coefficient, thereby obtaining the attention coefficient corresponding to the first feature vector;

[0121] The first feature vector and the attention coefficient corresponding to the first feature vector are input into the feature reconstruction layer to reconstruct the feature vector and obtain the second feature vector.

[0122] In one embodiment, when implementing the training process of the trained device state detection model, the processor is configured to:

[0123] Acquire multiple training samples, which include sample sensor parameters and sample sound data;

[0124] The device status detection model is iteratively trained using multiple training samples until it converges, resulting in a well-trained device status detection model.

[0125] In one embodiment, when acquiring multiple training samples, the processor is configured to:

[0126] Data cleaning is performed on the sensor parameter sets collected by multiple sensors.

[0127] The multiple sensor parameter sets that have undergone data cleaning are then processed to be dimensionless;

[0128] Generate a correlation coefficient matrix for the multiple sensor parameter sets after dimensionless processing;

[0129] Based on multiple correlation coefficients in the correlation coefficient matrix, a target sensor parameter set is selected from multiple sets of sensor parameters after data cleaning; the sensor parameters in the target sensor parameter set are determined as sample sensor parameters; and

[0130] The sound data recorded during the operation of the device is subjected to high-pass filtering to obtain sample sound data.

[0131] In one embodiment, the processor, in implementing the device state detection model, includes a feature processing module, an anomaly correction module, and an error calculation module; when iteratively training the device state detection model using multiple training samples, it is used to achieve:

[0132] The training samples, including the sample sensor parameters and sample sound data, are input into the feature processing module for feature extraction and concatenation to obtain a concatenated vector.

[0133] The spliced ​​vector is input into the anomaly correction module for reconstruction processing to obtain the reconstructed vector;

[0134] The spliced ​​vector and the reconstructed vector are input into the error calculation module for error analysis to obtain the mean square error between the spliced ​​vector and the reconstructed vector;

[0135] The model parameters of the equipment condition detection model are determined based on the mean square error, and the equipment condition detection model is iteratively trained based on the model parameters.

[0136] In one embodiment, the anomaly correction module includes a feature reconstruction layer filled with multiple positive samples, which are correctly labeled training samples; the processor, before performing each iteration of training on the device state detection model, is also configured to:

[0137] Remove multiple first positive sample instances from the feature reconstruction layer;

[0138] Obtain multiple second positive sample samples and fill the feature reconstruction layer with the multiple second positive sample samples.

[0139] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the computer device described above can be referred to the corresponding process in the aforementioned device status detection method embodiments, and will not be repeated here.

[0140] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, and the method implemented when the program instructions are executed can refer to various embodiments of the device status detection method of this application.

[0141] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0142] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0143] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. It should be noted that, herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0144] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. The above descriptions are merely specific implementations of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting equipment status, characterized in that, include: Acquire sensor parameters collected by the device's sensors, and acquire sound data recorded during the device's operation; Based on the sensor parameters and the sound data, deviation information used to characterize changes in the device performance is determined; The operating status of the equipment is determined based on the deviation information; The step of determining the deviation information characterizing the performance change of the device based on the sensor parameters and the sound data includes: The sensor parameters and the sound data are input into a trained device state detection model for processing to obtain the reconstruction error; the reconstruction error is used as the deviation information. The trained device state detection model includes a feature processing module, an anomaly correction module, and an error calculation module; the process of inputting the sensor parameters and the sound data into the trained device state detection model for processing to obtain the reconstruction error includes: The sensor parameters and the sound data are input into the feature processing module for feature extraction and concatenation to obtain a first feature vector; the first feature vector is input into the anomaly correction module for reconstruction processing to obtain a second feature vector; the first feature vector and the second feature vector are input into the error calculation module for error analysis to obtain the reconstruction error.

2. The equipment status detection method as described in claim 1, characterized in that, The step of determining deviation information characterizing changes in device performance based on the sensor parameters and the sound data includes: Determine the first degree of matching between the sensor parameters and the preset sensor parameters; Determine the second matching degree between the sound data and the preset sound data; The deviation information is determined based on the first matching degree and the second matching degree.

3. The equipment status detection method as described in claim 1, characterized in that, The anomaly correction module includes an attention layer and a feature reconstruction layer. The step of inputting the first feature vector into the anomaly correction module for reconstruction processing to obtain a second feature vector includes: The first feature vector is input into the attention layer to calculate the attention coefficient, thereby obtaining the attention coefficient corresponding to the first feature vector; The first feature vector and the attention coefficient corresponding to the first feature vector are input into the feature reconstruction layer to reconstruct the feature vector and obtain the second feature vector.

4. The equipment status detection method as described in claim 1, characterized in that, The training process of the trained device state detection model includes: Acquire multiple training samples, which include sample sensor parameters and sample sound data; The device status detection model is iteratively trained using multiple training samples until it converges, resulting in a well-trained device status detection model.

5. The equipment status detection method as described in claim 4, characterized in that, The acquisition of multiple training samples includes: Data cleaning is performed on the sensor parameter sets collected by multiple sensors. The multiple sensor parameter sets that have undergone data cleaning are then processed to be dimensionless; Generate a correlation coefficient matrix for the multiple sensor parameter sets after dimensionless processing; Based on multiple correlation coefficients in the correlation coefficient matrix, a target sensor parameter set is selected from multiple sets of sensor parameters after data cleaning; the sensor parameters in the target sensor parameter set are determined as sample sensor parameters; and The sound data recorded during the operation of the device is subjected to high-pass filtering to obtain sample sound data.

6. The equipment status detection method as described in claim 5, characterized in that, The equipment status detection model includes a feature processing module, an anomaly correction module, and an error calculation module; The iterative training of the device status detection model using multiple training samples includes: The training samples, including the sample sensor parameters and sample sound data, are input into the feature processing module for feature extraction and concatenation to obtain a concatenated vector. The spliced ​​vector is input into the anomaly correction module for reconstruction processing to obtain the reconstructed vector; The spliced ​​vector and the reconstructed vector are input into the error calculation module for error analysis to obtain the mean square error between the spliced ​​vector and the reconstructed vector; The model parameters of the equipment condition detection model are determined based on the mean square error, and the equipment condition detection model is iteratively trained based on the model parameters.

7. The equipment status detection method as described in claim 6, characterized in that, The anomaly correction module includes a feature reconstruction layer, which is filled with multiple positive samples, and the positive samples are correctly labeled training samples. Before each iteration of training the device state detection model, the method further includes: Remove multiple first positive sample instances from the feature reconstruction layer; Obtain multiple second positive sample samples and fill the feature reconstruction layer with the multiple second positive sample samples.

8. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of the device state detection method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the device state detection method as described in any one of claims 1 to 7.

Images