Methods, devices, equipment, storage media, and software products for water pump fault monitoring

Abstract

This application relates to a method, apparatus, device, storage medium, and program product for monitoring water pump faults. The method includes: acquiring sound signals generated during water pump operation; preprocessing the acquired sound signals to obtain preprocessed sound signals; extracting voiceprint features from the preprocessed sound signals to obtain first feature data; inputting the first feature data into a trained autoencoder neural network to output reconstructed second feature data; the trained autoencoder neural network reconstructs the input first feature data by pre-learning features from the feature data corresponding to the sound signals during normal water pump operation; thus, training can be performed using only data from normal water pump operation, reducing reliance on fault data. Based on the difference between the first and second feature data, it is determined whether a water pump fault has occurred. This enables early warning of water pump faults, improving water pump reliability and safety.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for monitoring water pump faults. Background Technology

[0002] Water pumps inevitably experience various malfunctions during long-term operation. If these malfunctions are not detected and addressed in their early stages, they can not only lead to a decline in pump performance but also pose safety hazards and cause economic losses.

[0003] Traditional technologies for monitoring water pump faults primarily rely on conventional signal analysis theories, such as Fourier transform, wavelet transform, and time-frequency domain analysis. However, using traditional signal analysis methods for water pump fault detection often depends on manual experience in data processing and threshold setting. This requires a high level of expertise and is highly subjective, making it impossible to ensure the reliability and safety of the water pump. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for monitoring water pump failures, which can achieve early warning of water pump failures and improve the reliability and safety of water pumps, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for monitoring water pump failures, the method comprising:

[0006] Collect the sound signals generated by the water pump during operation;

[0007] The collected sound signals are preprocessed to obtain preprocessed sound signals;

[0008] The preprocessed audio signal is subjected to voiceprint feature extraction to obtain the first feature data;

[0009] The first feature data is input into the trained autoencoder neural network, and the reconstructed second feature data is output. The trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0010] Based on the differences between the first feature data and the second feature data, it is determined whether the water pump has malfunctioned.

[0011] In one embodiment, the preprocessing of the acquired sound signal to obtain a preprocessed sound signal includes:

[0012] The audio signal in the form of a data frame is standardized to obtain a preprocessed audio signal; the standardization process includes dividing the amplitude of the audio signal within the length of the data frame by the maximum absolute value of the audio signal within the length of the data frame.

[0013] In one embodiment, the step of extracting voiceprint features from the preprocessed audio signal to obtain first feature data includes:

[0014] The preprocessed audio signal is decomposed by wavelet packet decomposition to obtain the energy elements of each frequency band after multi-level wavelet packet decomposition.

[0015] Construct wavelet packet frequency band energy feature vectors based on energy elements of each frequency band;

[0016] The wavelet packet frequency band energy feature vector is subjected to energy normalization processing to obtain the first feature data.

[0017] In one embodiment, determining whether the water pump has malfunctioned based on the difference between the first feature data and the second feature data includes:

[0018] Determine the threshold for judging normal operation of the water pump;

[0019] Determine the mean square error between the first feature data and the second feature data;

[0020] If the mean square error is greater than the discrimination threshold, then the water pump is determined to be faulty;

[0021] If the mean square error is not greater than the discrimination threshold, then the water pump is judged to be operating normally.

[0022] In one embodiment, before inputting the first feature data into the trained autoencoder neural network and outputting the reconstructed second feature data, the method further includes:

[0023] Based on the dimension of the first feature data, an initial autoencoder neural network is constructed; the initial autoencoder neural network includes: an input layer, a hidden layer, and an output layer;

[0024] Select the sound signal corresponding to the normal operation of the water pump to construct feature data samples;

[0025] The initial autoencoder neural network is trained using the feature data samples.

[0026] During the backpropagation process of the initial autoencoder neural network, the network parameters are optimized by minimizing the loss function value to obtain the trained autoencoder neural network.

[0027] In one embodiment, the loss function is a mean squared error function, used to constrain the mean of the squared errors between each data point in the input layer and each data point in the output layer of the initial autoencoder neural network.

[0028] Secondly, this application also provides a water pump fault monitoring device, the device comprising:

[0029] The acquisition module is used to acquire the sound signals generated by the water pump during operation;

[0030] The preprocessing module is used to preprocess the acquired sound signal to obtain a preprocessed sound signal;

[0031] The extraction module is used to extract voiceprint features from the preprocessed sound signal to obtain first feature data;

[0032] The processing module is used to input the first feature data into the trained autoencoder neural network and output the reconstructed second feature data; the trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0033] The judgment module is used to determine whether the water pump has malfunctioned based on the difference between the first feature data and the second feature data.

[0034] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0035] Collect the sound signals generated by the water pump during operation;

[0036] The collected sound signals are preprocessed to obtain preprocessed sound signals;

[0037] The preprocessed audio signal is subjected to voiceprint feature extraction to obtain the first feature data;

[0038] The first feature data is input into the trained autoencoder neural network, and the reconstructed second feature data is output. The trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0039] Based on the differences between the first feature data and the second feature data, it is determined whether the water pump has malfunctioned.

[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0041] Collect the sound signals generated by the water pump during operation;

[0042] The collected sound signals are preprocessed to obtain preprocessed sound signals;

[0043] The preprocessed audio signal is subjected to voiceprint feature extraction to obtain the first feature data;

[0044] The first feature data is input into the trained autoencoder neural network, and the reconstructed second feature data is output. The trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0045] Based on the differences between the first feature data and the second feature data, it is determined whether the water pump has malfunctioned.

[0046] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0047] Collect the sound signals generated by the water pump during operation;

[0048] The collected sound signals are preprocessed to obtain preprocessed sound signals;

[0049] The preprocessed audio signal is subjected to voiceprint feature extraction to obtain the first feature data;

[0050] The first feature data is input into the trained autoencoder neural network, and the reconstructed second feature data is output. The trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0051] Based on the differences between the first feature data and the second feature data, it is determined whether the water pump has malfunctioned.

[0052] The aforementioned water pump fault monitoring method, device, computer equipment, computer-readable storage medium, and computer program product collect sound signals generated during water pump operation; preprocess the collected sound signals to obtain preprocessed sound signals; thereby reducing the quality difference of water pump sound signals caused by changes in the position of the sound signal collector, thus improving the accuracy of subsequent fault judgment. Voiceprint feature extraction is performed on the preprocessed sound signals to obtain first feature data; thereby obtaining feature information in the sound signals, facilitating subsequent analysis of differences between features. The first feature data is input into a trained autoencoder neural network, outputting reconstructed second feature data; the trained autoencoder neural network reconstructs the input first feature data by pre-learning the features in the feature data corresponding to the sound signals during normal water pump operation; thus, it eliminates the need for repeated training of the autoencoder neural network using fault data, only requiring training with data from normal water pump operation, reducing reliance on fault data. Based on the differences between the first and second feature data, it is determined whether a water pump fault has occurred. This enables early warning of water pump faults, improving water pump reliability and safety. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating a water pump fault monitoring method in one embodiment;

[0055] Figure 2 This is a flowchart illustrating a pump fault monitoring method in another embodiment;

[0056] Figure 3 This is a schematic diagram of the structure of an autoencoder neural network in one embodiment;

[0057] Figure 4 This is a structural block diagram of a water pump fault monitoring device in one embodiment;

[0058] Figure 5 This is a structural block diagram of a water pump fault monitoring device in another embodiment;

[0059] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] Currently, pump fault monitoring technologies can be broadly categorized into two types. One type is based on traditional signal analysis theory, primarily including Fourier transform, wavelet transform, and time-frequency domain analysis. The other type is data-driven fault monitoring, mainly including machine learning, deep learning, and pattern recognition techniques. Both methods can monitor pump faults, but both have drawbacks. Traditional signal analysis methods often rely on human experience for data processing and threshold setting, requiring a high level of expertise. Data-driven methods do not rely heavily on human experience, but they are overly dependent on data quality, typically requiring a large amount of fault data for training the diagnostic model. However, in practical industrial applications, training data that clearly identifies fault types is often extremely difficult to obtain. Therefore, how to achieve simple and effective, accurate, and rapid pump fault monitoring remains a key research challenge.

[0062] To address the problems existing in the prior art, this application aims to provide a water pump fault monitoring method. This method can standardize the sound signal through preprocessing, reducing quality differences in the collected water pump sound signal caused by changes in microphone position. It performs multi-scale analysis of the signal in the time and frequency domains through wavelet packet decomposition to obtain deep features containing time-frequency information, thereby improving the accuracy of model recognition. Furthermore, it constructs an autoencoder neural network for fault analysis, eliminating the need for training with fault data and using only data from normal pump operation, thus reducing reliance on fault data. The method in this embodiment combines the advantages of traditional signal analysis theory and data-driven methods, enabling early warning of water pump faults and improving the reliability and safety of the water pump.

[0063] In one exemplary embodiment, such as Figure 1 As shown, a method for monitoring water pump faults is provided, which may include steps 101 to 105. Wherein:

[0064] Step 101: Collect the sound signal generated by the water pump during operation.

[0065] In this embodiment, at least one microphone can be installed near the water pump to collect the sound signals generated during pump operation. Optionally, when multiple microphones are installed, the sound signals collected by different microphones can be marked. For example, the sound signal collected by microphone 01 is marked with 01, and the sound signal collected by microphone 02 is marked with 02. This facilitates grouping during subsequent processing, that is, the feature data extracted from sound signals with the same mark are grouped together, and fault diagnosis is performed based on signals from the same group, thereby minimizing errors introduced by different microphone installation locations.

[0066] Optionally, the sound signal collected by the microphone can be transmitted to the processor (processing unit) in the form of data frames, and the processor can continue to process steps 102 to 104.

[0067] Step 102: Preprocess the collected sound signal to obtain the preprocessed sound signal.

[0068] In this embodiment, the sound signals collected by the microphones near the water pump may be subject to interference due to their varying installation locations. Therefore, the collected sound signals are first preprocessed to eliminate this interference as much as possible.

[0069] For example, a sound signal in the form of a data frame is normalized to obtain a preprocessed sound signal; wherein, the normalization process includes: dividing the amplitude of the sound signal within the length of the data frame by the maximum value of the absolute value of the sound signal within the length of the data frame.

[0070] Step 103: Extract voiceprint features from the preprocessed audio signal to obtain the first feature data.

[0071] In this embodiment, wavelet packet decomposition can be used to extract speaker features from the preprocessed audio signal. Wavelet packet decomposition can decompose not only the low-frequency components but also the high-frequency components, thus providing more spectral matching possibilities. Through multi-level decomposition, wavelet packets can adapt to the analysis needs of complex signals, providing a powerful tool for signal processing. Wavelet packet decomposition improves upon multi-resolution analysis, enabling more refined signal analysis. It divides the frequency band into multiple levels and selects appropriate frequency bands based on the characteristics of the analyzed signal to match the signal spectrum, thus obtaining a more refined signal decomposition than binary discrete wavelet transform.

[0072] For example, the preprocessed sound signal is decomposed into wavelet packets to obtain energy elements of each frequency band after multi-level wavelet packet decomposition; wavelet packet frequency band energy feature vectors are constructed based on the energy elements of each frequency band; and the wavelet packet frequency band energy feature vectors are normalized to obtain the first feature data.

[0073] Optionally, wavelet packet decomposition provides approximate and detail components of the signal at different frequencies and time scales. Processing these frequency band components yields features containing time-frequency information of the signal at different scales. The processing steps are as follows:

[0074] Step 1031: Decompose the signal into multiple layers based on approximate and detail components. The sum of the energy of each component after decomposition is the same as before decomposition. The energy formula for N-layer wavelet packet decomposition is:

[0075]

[0076] In the formula, The energy of the processed sound signal, Let be the energy of the j-th frequency band component after N-level wavelet packet decomposition. It can be represented as:

[0077]

[0078] In the formula: N is the number of decomposition layers, The reconstructed signal of the j-th frequency band component in the N-th layer of wavelet packet decomposition. for energy, for The discrete point amplitude, where n is the number of discrete points.

[0079] Step 1032: Construct feature vectors using the energy of each frequency band as elements. :

[0080]

[0081] Energy normalization of the above equation yields the wavelet packet band energy eigenvector. :

[0082]

[0083] For example, if the number of layers in the wavelet packet decomposition is set to 5, then the feature vector for:

[0084]

[0085] Wavelet packet frequency band energy feature vector for:

[0086]

[0087] Step 104: Input the first feature data into the trained autoencoder neural network and output the reconstructed second feature data.

[0088] The trained autoencoder neural network reconstructs the first input feature data by learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0089] In this embodiment, the autoencoder neural network is an unsupervised deep learning model designed to learn common features from a large amount of sample data. The autoencoder neural network approximates the output as the input, learning towards the goal of reconstructing the input, thus learning the key features for reconstructing the input data. The autoencoder neural network learns features from the wavelet packet frequency band energy feature vectors of normal data samples. However, the wavelet packet frequency band energy feature vectors of abnormal data differ significantly from those of normal data, resulting in a larger error after reconstruction by the autoencoder neural network, thus enabling identification. By designing an autoencoder neural network, the wavelet packet frequency band energy feature vectors extracted from normal sounds during water pump operation can be used as training samples to train the model, resulting in a trained autoencoder neural network.

[0090] Step 105: Based on the difference between the first feature data and the second feature data, determine whether the water pump has malfunctioned.

[0091] In this embodiment, since the trained autoencoder neural network can output reconstructed second feature data, and the second feature data is reconstructed based on the features in the feature data corresponding to the normal sound signal, it is possible to determine whether the water pump has malfunctioned based on the difference between the first feature data and the second feature data.

[0092] For example, a threshold for judging normal operation of the water pump is determined; the mean square error between the first feature data and the second feature data is determined; if the mean square error is greater than the threshold, the water pump is judged to be faulty; if the mean square error is not greater than the threshold, the water pump is judged to be operating normally.

[0093] For example, during the training of an autoencoder neural network, the mean square error of the input and output can be used as a criterion for judging normal samples. Optionally, considering the fluctuations in sound during device operation, the mean square error obtained after the autoencoder neural network training is completed can be increased by three times the variance as the discrimination threshold P.

[0094] Optionally, if the mean square error between the first feature data and the second feature data is greater than the discrimination threshold P, then the first feature data is determined to be fault data; if the mean square error between the first feature data and the second feature data is less than or equal to the discrimination threshold P, then the first feature data is determined to be normal data. That is, when the first feature data is normal data, it indicates that the water pump is operating normally.

[0095] In the aforementioned water pump fault monitoring method, sound signals generated during water pump operation are collected. These signals are then preprocessed to obtain preprocessed sound signals, reducing the quality differences in the water pump sound signals caused by changes in the location of the sound signal collector and improving the accuracy of subsequent fault diagnosis. Voiceprint features are extracted from the preprocessed sound signals to obtain first feature data, thus obtaining feature information from the sound signals for subsequent analysis of differences between features. The first feature data is input into a trained autoencoder neural network, which outputs reconstructed second feature data. The trained autoencoder neural network reconstructs the input first feature data by pre-learning the features in the feature data corresponding to the sound signals during normal water pump operation. This eliminates the need for repeated training of the autoencoder neural network using fault data; training only requires data from normal water pump operation, reducing reliance on fault data. Based on the differences between the first and second feature data, a fault is determined in the water pump. This enables early warning of water pump faults, improving water pump reliability and safety.

[0096] In another exemplary embodiment, such as Figure 2 As shown, a method for monitoring water pump faults is provided, which may include steps 201 to 209. Wherein:

[0097] Step 201: Collect the sound signal generated by the water pump during operation.

[0098] Step 202: Preprocess the collected sound signal to obtain the preprocessed sound signal.

[0099] Step 203: Extract voiceprint features from the preprocessed audio signal to obtain the first feature data.

[0100] In this embodiment, please refer to the detailed implementation process and technical effects of steps 201 to 203. Figure 1 The relevant descriptions of steps 101 to 103 in the method embodiment shown will not be repeated here.

[0101] Step 204: Construct an initial autoencoder neural network based on the dimension of the first feature data.

[0102] The initial autoencoder neural network consists of an input layer, a hidden layer, and an output layer.

[0103] For example, assuming the wavelet packet decomposition layer number is set to 5, the resulting wavelet packet frequency band energy feature vector is 32-dimensional data. Therefore, the initial autoencoder neural network can be designed with input and output layer parameters of 32, and the parameters of the three hidden layers can be designed to be 16, 8, and 16 respectively. For example, Figure 3 The structure of the autoencoder neural network is given, such as Figure 3 As shown in the figure, x1~x 32 This represents the input data, x'1~x' 32 Indicates the output data. h1 1 ~h 16 1 h1 represents the parameters of the first hidden layer. 2 ~ h8 2 h1 represents the parameters of the second hidden layer. 3 ~h 16 3 This represents the parameters of the third hidden layer.

[0104] Step 205: Select the sound signal corresponding to the normal operation of the water pump to construct feature data samples.

[0105] In this embodiment, there is no need to collect sound signals when the water pump malfunctions to build training samples. Instead, feature data samples are directly constructed using the sound signals corresponding to the normal operation of the water pump.

[0106] Optionally, the sound signal corresponding to the normal operation of the water pump is preprocessed, and then the preprocessed sound signal is subjected to voiceprint feature extraction. The extracted feature data is then used to construct a feature data sample.

[0107] Step 206: Train the initial autoencoder neural network using feature data samples.

[0108] In this embodiment, the following can be used: The constraint is to use a loss function to make the output and input as similar as possible. For input data, This is the output result.

[0109] Step 207: During the backpropagation of the initial autoencoder neural network, the network parameters are optimized by minimizing the loss function value to obtain the trained autoencoder neural network.

[0110] In this embodiment, during the backpropagation process of the initial autoencoder neural network, the network parameters are optimized by minimizing the loss function value, thereby obtaining the autoencoder neural network model.

[0111] For example, the loss function uses the mean-square error (MSE) function to constrain the mean of the squared errors between each data point in the input layer and each data point in the output layer of the initial autoencoder neural network. Where:

[0112]

[0113] In the formula, For the sample size, To input the raw data, To output reconstructed data.

[0114] Step 208: Input the first feature data into the trained autoencoder neural network and output the reconstructed second feature data.

[0115] The trained autoencoder neural network reconstructs the first input feature data by learning the features in the feature data corresponding to the sound signal when the water pump is running normally.

[0116] Step 209: Based on the difference between the first feature data and the second feature data, determine whether the water pump has malfunctioned.

[0117] In this embodiment, please refer to the detailed implementation process and technical effects of steps 208 to 209. Figure 1 The relevant descriptions of steps 104 to 105 in the method embodiment shown will not be repeated here.

[0118] In this embodiment, wavelet packet decomposition can be used to perform multi-scale analysis of the signal in both the time and frequency domains to obtain deep features containing the signal's time-frequency information, thereby improving the accuracy of data model recognition. The input feature data is reconstructed using a trained autoencoder neural network, and the pump's fault condition is determined based on the difference between the reconstructed feature data and the input feature data. Since the autoencoder neural network does not require training with fault data, it can reconstruct the input data using only data from when the pump is operating normally. This reduces reliance on fault data, enabling real-time monitoring of pump faults, timely fault detection, and improved pump reliability and safety.

[0119] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0120] Based on the same inventive concept, this application also provides a pump fault monitoring device for implementing the pump fault monitoring method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the pump fault monitoring device provided below can be found in the limitations of the pump fault monitoring method described above, and will not be repeated here.

[0121] In one exemplary embodiment, such as Figure 4 As shown, a water pump fault monitoring device is provided, including: a data acquisition module 401, a preprocessing module 402, an extraction module 403, a processing module 404, and a judgment module 405, wherein:

[0122] Acquisition module 401 is used to acquire sound signals generated during the operation of the water pump;

[0123] The preprocessing module 402 is used to preprocess the acquired sound signal to obtain the preprocessed sound signal;

[0124] Extraction module 403 is used to extract voiceprint features from the preprocessed sound signal to obtain first feature data;

[0125] Processing module 404 is used to input the first feature data into the trained autoencoder neural network and output the reconstructed second feature data; the trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally;

[0126] The judgment module 405 is used to determine whether the water pump has malfunctioned based on the difference between the first feature data and the second feature data.

[0127] For example, the preprocessing module 402 is specifically used to: perform standardization processing on the audio signal in the form of a data frame to obtain a preprocessed audio signal; wherein, the standardization processing includes: dividing the amplitude of the audio signal within the length of the data frame by the maximum value of the absolute value of the audio signal within the length of the data frame.

[0128] For example, the extraction module 403 is specifically used to: perform wavelet packet decomposition on the preprocessed sound signal to obtain energy elements of each frequency band after multi-level wavelet packet decomposition; construct wavelet packet frequency band energy feature vectors based on the energy elements of each frequency band; and perform energy normalization processing on the wavelet packet frequency band energy feature vectors to obtain the first feature data.

[0129] For example, the judgment module 405 is specifically used to: determine the discrimination threshold for normal operation of the water pump; determine the mean square error between the first feature data and the second feature data; if the mean square error is greater than the discrimination threshold, then the water pump is judged to be faulty; if the mean square error is not greater than the discrimination threshold, then the water pump is judged to be operating normally.

[0130] In another exemplary embodiment, such as Figure 5 As shown, a water pump fault monitoring device is provided, which can... Figure 4 Based on the device shown, it may also include:

[0131] Autoencoder neural network building module 406 is used to build an initial autoencoder neural network based on the dimension of the first feature data; the initial autoencoder neural network includes an input layer, a hidden layer and an output layer;

[0132] The sample construction module 407 is used to select the sound signal corresponding to the normal operation of the water pump to construct feature data samples;

[0133] Training module 408 is used to train the initial autoencoder neural network using feature data samples;

[0134] The optimization module 409 is used to optimize the network parameters by minimizing the loss function value during the backpropagation process of the initial autoencoder neural network, so as to obtain a trained autoencoder neural network.

[0135] For example, the loss function uses the mean squared error function to constrain the sum of squares of the errors between each data point in the input layer and each data point in the output layer of the initial autoencoder neural network.

[0136] Each module in the aforementioned water pump fault monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0137] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a water pump fault monitoring method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0138] 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 to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0139] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0140] The system collects the sound signal generated by the water pump during operation; preprocesses the collected sound signal to obtain a preprocessed sound signal; extracts voiceprint features from the preprocessed sound signal to obtain first feature data; inputs the first feature data into a trained autoencoder neural network to output reconstructed second feature data; the trained autoencoder neural network reconstructs the input first feature data by learning the features in the feature data corresponding to the sound signal when the water pump is running normally; and determines whether the water pump has malfunctioned based on the difference between the first and second feature data.

[0141] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0142] The audio signal in the form of a data frame is standardized to obtain a preprocessed audio signal; the standardization process includes dividing the amplitude of the audio signal within the length of the data frame by the maximum absolute value of the audio signal within the length of the data frame.

[0143] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0144] The preprocessed sound signal is decomposed into wavelet packets to obtain the energy elements of each frequency band after multi-level wavelet packet decomposition; wavelet packet frequency band energy feature vectors are constructed based on the energy elements of each frequency band; energy normalization is performed on the wavelet packet frequency band energy feature vectors to obtain the first feature data.

[0145] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0146] Determine the threshold for judging normal operation of the water pump; determine the mean square error between the first feature data and the second feature data; if the mean square error is greater than the threshold, the water pump is judged to be faulty; if the mean square error is not greater than the threshold, the water pump is judged to be operating normally.

[0147] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0148] Based on the dimension of the first feature data, an initial autoencoder neural network is constructed. The initial autoencoder neural network includes an input layer, a hidden layer, and an output layer. The sound signal corresponding to the normal operation of the water pump is selected to construct feature data samples. The initial autoencoder neural network is trained using the feature data samples. During the backpropagation process of the initial autoencoder neural network, the network parameters are optimized by minimizing the loss function value to obtain the trained autoencoder neural network.

[0149] In one embodiment, the loss function is a mean squared error function, which is used to constrain the mean of the sum of squares of the errors between each data point in the input layer and each data point in the output layer of the initial autoencoder neural network.

[0150] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being implemented when executed by a processor. Figure 1 , Figure 2 The steps involved in implementing the method are shown.

[0151] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements... Figure 1 , Figure 2 The steps involved in implementing the method are shown.

[0152] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0153] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0154] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0155] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring water pump faults, characterized in that, The method includes: Collect the sound signals generated by the water pump during operation; The acquired sound signal is preprocessed to obtain a preprocessed sound signal; this includes: standardizing the sound signal in the form of a data frame to obtain a preprocessed sound signal; wherein, the standardization process includes: dividing the amplitude of the sound signal within the length of the data frame by the maximum absolute value of the sound signal within the length of the data frame; The process of extracting voiceprint features from the preprocessed audio signal to obtain first feature data includes: performing wavelet packet decomposition on the preprocessed audio signal to obtain energy elements of each frequency band after multi-level wavelet packet decomposition; constructing wavelet packet frequency band energy feature vectors based on the energy elements of each frequency band; and performing energy normalization processing on the wavelet packet frequency band energy feature vectors to obtain the first feature data. The first feature data is input into the trained autoencoder neural network, and the reconstructed second feature data is output. The trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally. Based on the differences between the first feature data and the second feature data, it is determined whether the water pump has malfunctioned.

2. The method according to claim 1, characterized in that, The step of determining whether the water pump is malfunctioning based on the difference between the first feature data and the second feature data includes: Determine the threshold for judging normal operation of the water pump; Determine the mean square error between the first feature data and the second feature data; If the mean square error is greater than the discrimination threshold, then the water pump is determined to be faulty; If the mean square error is not greater than the discrimination threshold, then the water pump is judged to be operating normally.

3. The method according to claim 1 or 2, characterized in that, Before inputting the first feature data into the trained autoencoder neural network and outputting the reconstructed second feature data, the method further includes: Based on the dimension of the first feature data, an initial autoencoder neural network is constructed; the initial autoencoder neural network includes: an input layer, a hidden layer, and an output layer; Select the sound signal corresponding to the normal operation of the water pump to construct feature data samples; The initial autoencoder neural network is trained using the feature data samples. During the backpropagation process of the initial autoencoder neural network, the network parameters are optimized by minimizing the loss function value to obtain the trained autoencoder neural network.

4. The method according to claim 3, characterized in that, The loss function is a mean squared error function, used to constrain the mean of the squared errors between each data point in the input layer and each data point in the output layer of the initial autoencoder neural network.

5. A water pump fault monitoring device, characterized in that, The device includes: The acquisition module is used to acquire the sound signals generated by the water pump during operation; The preprocessing module is used to preprocess the acquired sound signal to obtain a preprocessed sound signal. Specifically, the preprocessing module is used to: standardize the sound signal in the form of a data frame to obtain a preprocessed sound signal; wherein, the standardization process includes: dividing the amplitude of the sound signal within the length of the data frame by the maximum absolute value of the sound signal within the length of the data frame. The extraction module is used to extract voiceprint features from the preprocessed sound signal to obtain first feature data. Specifically, the extraction module is used to: perform wavelet packet decomposition on the preprocessed sound signal to obtain energy elements of each frequency band after multi-level wavelet packet decomposition; construct wavelet packet frequency band energy feature vectors based on the energy elements of each frequency band; and perform energy normalization processing on the wavelet packet frequency band energy feature vectors to obtain the first feature data. The processing module is used to input the first feature data into the trained autoencoder neural network and output the reconstructed second feature data; the trained autoencoder neural network performs data reconstruction processing on the input first feature data by pre-learning the features in the feature data corresponding to the sound signal when the water pump is running normally. The judgment module is used to determine whether the water pump has malfunctioned based on the difference between the first feature data and the second feature data.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images