Fan fault detecting and positioning method and system based on microphone array

Abstract

The invention discloses a fan fault detection and positioning method and system based on a microphone array, and the method comprises the steps: obtaining a sound signal, and carrying out the first preprocessing of the sound signal, and obtaining a first sound signal; obtaining the position of a fault sound source by calculating the time difference of the first sound signals received by the microphones; optimizing the position of the fault sound source by using a first optimization method to obtain a position coordinate of the fault sound source; classifying fault types based on the fault sound source position coordinates to obtain a classification result; acquiring real-time monitoring data, and performing modeling in combination with the classification result to obtain a first neural network model; according to the method, historical operation data and historical fault information are acquired, the first neural network model is trained, the trained first neural network model is verified, a parameter prediction value is output, the parameter prediction value is judged, a fault result is obtained, and the deployment and extension cost is reduced through modular design and software and hardware optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of fan fault detection and location, and in particular to a method and system for fan fault detection and location based on a microphone array. Background Art

[0002] Traditional fan fault detection technologies, such as acoustic probes and vibration analysis, although they can detect fault signals, have great limitations in terms of location accuracy and real-time performance. The emerging trend analysis method based on SCADA system data introduces big data analysis and prediction models, but requires high computing resources and has certain difficulties in applying to edge devices.

[0003] The present invention provides a low-cost and high-efficiency solution for fan fault location and trend analysis by combining microphone array acoustic location technology, delay algorithm and machine learning classification model, filling the gap in multi-functional integration of the existing technology. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method and system for fan fault detection and location based on a microphone array to solve the problems that traditional fan fault detection technologies have great limitations in terms of location accuracy and real-time performance, and emerging technologies require high computing resources and have certain difficulties in applying to edge devices.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for fan fault detection and location based on a microphone array, including:

[0008] Obtain a sound signal, perform a first preprocessing on the sound signal to obtain a first sound signal;

[0009] Calculate the time difference of the first sound signal received by each microphone to obtain the position of the fault sound source;

[0010] Optimize the position of the fault sound source by using a first optimization method to obtain the coordinates of the fault sound source position;

[0011] Classify the fault type based on the coordinates of the fault sound source position to obtain a classification result;

[0012] Obtain real-time monitoring data, and build a model in combination with the classification result to obtain a first neural network model;

[0013] Obtain historical operation data and historical fault information, train the first neural network model, verify the trained first neural network model, output parameter prediction values, and judge the parameter prediction values to obtain fault results.

[0014] As a preferred solution of the fan fault detection and location method based on a microphone array according to the present invention, wherein: the first preprocessing of the sound signal includes:

[0015] Decompose the sound signal into sub-bands by octave analysis and introduce a signal processing method for optimization;

[0016] Adjust the number and range of sub-bands according to the signal characteristics, and combine multi-scale fusion to extract the sub-band sound energy characteristics;

[0017] Normalize the sub-band sound energy characteristics to obtain the first sound signal.

[0018] As a preferred solution of the fan fault detection and location method based on a microphone array according to the present invention, wherein: calculating the time difference of the first sound signal received by each microphone includes:

[0019] Select a reference microphone as a reference point;

[0020] Perform cross-correlation operation on the signal collected by the reference microphone and the signals of other microphones;

[0021] Compare the similarity between the signals, and the time difference corresponding to the maximum similarity is the delay value, that is, the time delay of other microphones relative to the reference microphone.

[0022] As a preferred solution of the fan fault detection and location method based on a microphone array according to the present invention, wherein: further includes:

[0023] Establish a geometric model according to the microphone positions, time delays, and sound propagation speed;

[0024] Utilize the geometric relationship to convert the delay value into the actual distance from the sound source to each microphone.

[0025] As a preferred solution of the fan fault detection and location method based on a microphone array according to the present invention, wherein: optimizing the fault sound source position by using a first optimization method includes:

[0026] Calculate the delay values from the sound source to each microphone and compare them with the known delay values to obtain an objective function;

[0027] Set an initial estimated position;

[0028] The objective function is optimized using the least squares method to obtain the sound source position that best matches the delay value, i.e., the coordinate of the fault sound source position.

[0029] As a preferred solution of the fan fault detection and localization method based on a microphone array according to the present invention, wherein: combining the classification result for modeling includes:

[0030] Form a two-dimensional matrix of real-time monitoring data and classification results as input data, and construct a first neural network model;

[0031] Obtain historical operation data and fault information marked in the historical data as a training set, and optimize the model weights through supervised learning to obtain the trained first neural network model;

[0032] Use the historical operation data, the fault information marked in the historical data, the classification result, and the sub-band sound energy characteristics as the input of the trained first neural network model to obtain parameter prediction values.

[0033] As a preferred solution of the fan fault detection and localization method based on a microphone array according to the present invention, wherein: judging the parameter prediction value includes:

[0034] Determine the first threshold of the parameter according to the historical operation data and the fault information marked in the historical data. If the parameter is greater than the first threshold, the trained first neural network outputs a fault warning.

[0035] In a second aspect, the present invention provides a fan fault detection and localization system based on a microphone array, including:

[0036] A preprocessing module for acquiring a sound signal, performing a first preprocessing on the sound signal to obtain a first sound signal;

[0037] A calculation module for obtaining the fault sound source position by calculating the time difference of the first sound signal received by each microphone;

[0038] A positioning module for optimizing the fault sound source position using a first optimization method to obtain the coordinate of the fault sound source position;

[0039] A classification module for classifying the fault type based on the coordinate of the fault sound source position to obtain a classification result;

[0040] A model establishment module for acquiring real-time monitoring data, combining the classification result for modeling to obtain a first neural network model;

[0041] A judgment module, configured to obtain historical operation data and historical fault information, train the first neural network model, verify the trained first neural network model, output a parameter prediction value, and judge the parameter prediction value to obtain a fault result.

[0042] In a third aspect, the present invention provides a computing device, including:

[0043] A memory and a processor;

[0044] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the fan fault detection and location method based on a microphone array are implemented.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the fan fault detection and location method based on a microphone array are implemented.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows: The system of the present invention integrates functions of signal acquisition, feature extraction, fault location, classification and recognition, and trend prediction, covering the entire life cycle of fan operation and maintenance. It realizes stable location and classification capabilities in a complex noise environment. Through modular design and software and hardware optimization, the deployment and expansion costs are reduced, and it is applicable to a wide range of industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a schematic diagram of the overall process logic of the fan fault detection and location method based on a microphone array according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] Embodiment 1

[0051] Referring toFigure 1 , which is an embodiment of the present invention, provides a method for detecting and locating fan faults based on a microphone array, including:

[0052] S100: Obtain a sound signal, perform a first preprocessing on the sound signal to obtain a first sound signal;

[0053] S200: Calculate the time difference of the first sound signals received by each microphone to obtain the position of the fault sound source;

[0054] S300: Optimize the position of the fault sound source using a first optimization method to obtain the coordinates of the fault sound source position;

[0055] S400: Classify the fault types based on the coordinates of the fault sound source position to obtain a classification result;

[0056] Specifically, use support vector machine classification to accurately classify the fault types through training sample data to obtain a classification result, improving the recognition efficiency and accuracy;

[0057] Reduce the dimension through PCA (Principal Component Analysis) or LDA (Linear Discriminant Analysis) to improve the data discrimination, making different fault categories more obvious.

[0058] S500: Obtain real-time monitoring data, combine with the classification result for modeling to obtain a first neural network model;

[0059] S600: Obtain historical operation data and historical fault information, train the first neural network model, verify the trained first neural network model, output parameter prediction values, and judge the parameter prediction values to obtain a fault result.

[0060] It should be noted that by collecting acoustic signals through a microphone array and using a delay algorithm to calculate the time delay of the sound signals, the position of the fault sound source can be accurately located. The acoustic signals are decomposed using octave analysis, and the FFT is combined to optimize the filtering process, dynamically adjusting the sub-band division to ensure high-precision analysis of abnormal signals in the high-frequency band. Accurately classifying the fault types through support vector machines can efficiently identify different types of fan faults.

[0061] In the embodiment of the present application, the above step S100 includes the following sub-steps A1 - A3;

[0062] In A1: Decompose the sound signal into sub-bands using octave analysis and introduce signal processing methods for optimization;

[0063] In A2: Adjust the number and range of sub-bands according to the signal characteristics, combine multi-scale fusion, and extract the sub-band sound energy characteristics;

[0064] In A3: Normalize the sub-band sound energy characteristics to obtain the first sound signal.

[0065] In an alternative embodiment, the signal processing method can be wavelet transform. Select an appropriate wavelet basis for multi-scale decomposition, decompose the signal into approximate coefficients and detail coefficients of different scales, select the detail coefficients of the key frequency bands, calculate the energy characteristics of each frequency band, so as to obtain optimization. Wavelet transform is suitable for analyzing non-stationary signals, can perform local analysis between the time domain and the frequency domain, improve the flexibility of signal decomposition, further improve the frequency resolution through wavelet packet decomposition, and enhance the ability to capture fan fault signals.

[0066] In an alternative embodiment, the signal processing method can be Hilbert-Huang transform. Decompose the signal into several intrinsic mode functions, and each mode function represents the characteristic components of different frequencies and time scales in the signal. Perform Hilbert transform on each mode function, calculate the instantaneous frequency and instantaneous amplitude, generate the Hilbert spectrum, and analyze the change of the signal frequency over time, so as to obtain optimization. Use empirical mode decomposition to decompose the signal and calculate the instantaneous frequency, which is suitable for the analysis of complex dynamic signals, can be used to analyze non-linear and non-stationary signals, and improve the ability to identify fault noise.

[0067] In an alternative embodiment, the signal processing method can also be sparse representation. Construct a predefined dictionary, and the atoms in the dictionary can be pre-trained fault feature templates or obtained through adaptive learning. Use the sparse coding algorithm to perform sparse representation on the collected signal, and represent the signal as a linear combination of the atoms in the dictionary; extract the key features of the signal through dictionary learning, effectively remove redundant information, improve the classification accuracy, and are suitable for the enhancement and denoising of weak fault signals, and improve the detection sensitivity.

[0068] In the embodiment of the present application, the signal processing method is fast Fourier transform;

[0069] Specifically, adopt a flexible array layout (such as 3×3 or 5×5 matrix arrangement), adjust the microphone density and distribution according to the device size, each microphone collects the sound signal in real time and transmits it to the signal processing unit for preprocessing; decompose the sound signal into sub-bands through octave analysis, extract the sub-band sound energy characteristics. Octave analysis divides the signal into several frequency intervals on a logarithmic scale, and the ratio between the center frequencies is usually a fixed value, such as 1 / 3 octave (the center frequency spacing is one-third of it), and the energy of each octave sub-band can be calculated through a band-pass filter bank.

[0070] Use fast Fourier transform to optimize the filtering process, perform fast Fourier transform on the preprocessed signal, and convert it from the time domain to the frequency domain. Fast Fourier transform decomposes the signal into the amplitude and phase information of different frequency components.

[0071] Dynamically adjust the number and range of sub - frequency bands according to the characteristics of the signal. For example, if there are high - frequency anomalies in the signal, the number of sub - frequency bands in the high - frequency band can be increased to focus on high - frequency characteristics;

[0072] Combine 1 / 3 octave and full - band energy analysis to ensure consideration of both details and overall characteristics. Use the normalization method to eliminate dimensional differences and provide high - quality input for subsequent model processing;

[0073] It should be noted that conventional octave analysis uses a filter bank to decompose the signal one by one, but there may be problems such as too narrow bandwidth and large computational amount in the analysis of complex signals or high - frequency bands. Using the fast Fourier transform to optimize the filtering process can accelerate signal decomposition, segment and aggregate the frequency spectrum range into octave sub - frequency bands, and reduce the redundancy of filter bank calculations.

[0074] In the embodiment of the present application, the above - mentioned step S200 includes the following sub - steps B1 - B5;

[0075] In B1: Select the reference microphone as the reference point;

[0076] In B2: Perform cross - correlation operation on the signal collected by the reference microphone and the signals of other microphones;

[0077] In B3: Compare the similarity between the signals. The time difference corresponding to the maximum similarity is the delay value, that is, the time delay of other microphones relative to the reference microphone;

[0078] In B4: Establish a geometric model according to the microphone positions, time delay, and sound propagation speed;

[0079] In B5: Use geometric relationships to convert the delay value into the actual distance from the sound source to each microphone.

[0080] Specifically, the pre - processed sound signal is input into the delay algorithm module. The delay algorithm determines the position of the fault sound source by calculating the time difference of the sound signals received by each microphone, specifically:

[0081] Select the reference microphone as the reference point. Calculate the time delay of the sound signals received by other microphones relative to the reference microphone. Perform cross - correlation operation on the signal collected by the reference microphone and the signals of other microphones, and compare the similarity between the two signals. The time difference corresponding to the maximum similarity is the delay value. According to the known microphone positions, time delay, and sound propagation speed (such as 343.2 m / s), establish a geometric model. The time delay indicates the distance difference from the sound source to each microphone. Use the geometric relationship formula to convert the delay value into the actual distance from the sound source to each microphone.

[0082] It should be noted that through cross - correlation operation and geometric model, the distances from the sound source to each microphone can be accurately calculated, thereby achieving high - precision fault sound source localization. The calculation efficiency is high, and it can quickly process signals and determine the fault location. Through cross - correlation operation, the time delay of the signal can also be effectively extracted, improving the reliability of localization.

[0083] In the embodiment of the present application, the above - mentioned step S300 includes the following sub - steps C1 - C4;

[0084] In C1: Calculate the delay values from the sound source to each microphone, and compare them with the known delay values to obtain the objective function;

[0085] In C2: Set the initial estimated position;

[0086] In C3: Use the least - squares method to optimize the objective function to obtain the sound source position that best matches the delay value, that is, the fault sound source position coordinates.

[0087] Specifically, after determining the position of the fault sound source through the delay algorithm, the position information is transmitted to the fault location module. The fault location module combines the structure diagram of the fan to accurately locate the fault point;

[0088] The least - squares method is used to optimize the fault sound source localization, and the optimization process is as follows:

[0089] Construct the objective function. Let the position of each microphone in the microphone array be M j (x j ,y j, z j ), the sound source position is S(x, y, z), the speed of sound is c, and the measured time delay is t i , then the distance from the sound source to the microphone satisfies: d i =c×t i , and the error function is expressed as the sum of the squares of the difference between the measured distance and the calculated distance, which is expressed as:

[0090]

[0091] Calculate the gradient and optimize the objective function, calculate the partial derivatives, which are expressed as:

[0092]

[0093] Adopt the gradient descent method or the least - squares fitting method for iterative optimization, continuously adjust x, y, z until the error is minimized;

[0094] Utilize the geometric layout relationship of the microphones, select a reasonable initial estimated position to accelerate the convergence speed, and adopt the Levenberg - Marquardt (LM) algorithm to avoid the problem of local optimal solutions and improve the calculation accuracy.

[0095] It should be noted that by optimizing the objective function through the least squares method, the distances from the sound source to each microphone can be accurately calculated, thereby achieving high-precision fault sound source localization. A reasonable initial estimation position and an efficient optimization algorithm can accelerate the convergence speed, improve the real-time performance of the system response, and finally obtain the fault sound source coordinates that best match the actual measurement data, improving the accuracy and robustness of sound source localization.

[0096] In the embodiment of the present application, the above step S500 includes the following sub-steps D1-D3;

[0097] In D1: The real-time monitoring data and the classification result are combined into a two-dimensional matrix as input data, and a first neural network model is constructed;

[0098] In D2: The historical operation data and the fault information marked in the historical data are obtained as the training set, and the model weights are optimized through supervised learning to obtain the trained first neural network model;

[0099] In D3: The historical operation data, the fault information marked in the historical data, the classification result, and the sub-band sound energy characteristics are used as the input of the trained first neural network model to obtain the parameter prediction value.

[0100] Specifically, the real-time monitoring data includes vibration signals (such as bearing vibration), acoustic fingerprint signals, temperature signals (such as lubricating oil temperature, cooling air temperature), rotational speed signals, torque, power, wind speed and other related data. The preprocessing of the data includes removing outliers (such as detecting through a sliding window mean or Mahalanobis distance) to ensure data stability, normalizing the data to reduce the impact of different data dimensions on modeling, and selecting the most important feature parameters (such as the power spectral density of the vibration signal) through stepwise regression or PCA methods.

[0101] The input data for LSTM trend prediction can be a combination of multiple sources, including the classification result and the original monitoring data. The classification result is used as part of the LSTM input to help identify whether the current working condition is likely to develop into a fault trend, and the time series data collected by the SCADA system alone is used for trend modeling;

[0102] The first neural network model is an LSTM model. The input of the LSTM model is the preprocessed time series data, including the classification result and the real-time monitoring data. The data structure is usually represented as a two-dimensional matrix:

[0103]

[0104] Among them, t is the time step and n is the feature dimension;

[0105] The LSTM layer structure includes an input gate: selectively memorizing input data; a forget gate: discarding unnecessary historical information; and an output gate: generating a prediction result for the current time step.

[0106] Using historical operation data and known fault annotations as the training set, optimize the model weights through supervised learning. The loss function is the mean squared error expressed as:

[0107]

[0108] where y i is the true value, is the model prediction value;

[0109] The LSTM model outputs parameter prediction values for a future period of time, such as vibration amplitude, temperature trend, etc.

[0110] It should be noted that the LSTM model can capture long-term dependencies in time series data, perform high-precision prediction on the change trend of fan operation parameters. By combining historical operation data and real-time monitoring data, the model can more accurately identify potential fault trends.

[0111] In the embodiment of the present application, the above step S600 includes the following sub-step E1;

[0112] In E1: Determine the first threshold of the parameter according to the historical operation data and the fault information annotated in the historical data. If the parameter is greater than the first threshold, the trained first neural network outputs a fault warning;

[0113] Specifically, determine the safety threshold of the key parameter according to the historical fault data. For example, when the vibration signal exceeds a certain amplitude (such as 10 mm / s), the temperature rises to a dangerous value (such as exceeding 80 °C), the model outputs a fault warning;

[0114] Utilize the Lora protocol and the 5G network to build a remote real-time monitoring platform, integrate the positioning result with the fan structure diagram to generate a visual fault point positioning map, support the operation and maintenance team to make efficient decisions, and intuitively present the fault signal characteristics through spectrograms and time-frequency diagrams to assist engineers in analyzing the operation status of the fan.

[0115] It should be noted that by setting the safety threshold of key parameters, the model can send early warning signals in time before a fault occurs, helping the operation and maintenance team take measures in advance to avoid downtime losses caused by sudden equipment failures. The remote real-time monitoring platform combines visualization technologies (such as fault point location maps, spectrograms, and time-frequency diagrams) to provide intuitive fault information for the operation and maintenance personnel, reducing the fault troubleshooting time and labor costs. The operation and maintenance team can quickly locate the fault point and formulate a targeted maintenance plan based on the visualization information. The integration of the fault location result with the fan structure diagram enables the operation and maintenance personnel to clearly understand the fault location and the affected range, providing a scientific basis for rapid decision-making.

[0116] The above is a schematic solution of a fan fault detection and location method based on a microphone array in this embodiment. It should be noted that the technical solution of the fan fault detection and location system based on the microphone array belongs to the same concept as the technical solution of the above-mentioned fan fault detection and location method based on the microphone array. For the details not described in detail in the technical solution of the fan fault detection and location system based on the microphone array in this embodiment, reference can be made to the description of the technical solution of the above-mentioned fan fault detection and location method based on the microphone array.

[0117] The fan fault detection and location system based on the microphone array in this embodiment includes:

[0118] A preprocessing module, configured to obtain a sound signal, perform a first preprocessing on the sound signal to obtain a first sound signal;

[0119] A calculation module, configured to obtain the fault sound source location by calculating the time difference of the first sound signals received by each microphone;

[0120] A location module, configured to optimize the fault sound source location by using a first optimization method to obtain the fault sound source location coordinates;

[0121] A classification module, configured to classify the fault type based on the fault sound source location coordinates to obtain a classification result;

[0122] A model establishment module, configured to obtain real-time monitoring data, and perform modeling in combination with the classification result to obtain a first neural network model;

[0123] A judgment module, configured to obtain historical operation data and historical fault information, train the first neural network model, verify the trained first neural network model, output a parameter prediction value, and judge the parameter prediction value to obtain a fault result.

[0124] This embodiment also provides a computing device, applicable to the situation of fan fault detection and location based on a microphone array, including:

[0125] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for detecting and locating fan faults based on a microphone array as proposed in the above embodiments.

[0126] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for detecting and locating fan faults based on a microphone array as proposed in the above embodiments.

[0127] The storage medium proposed in this embodiment and the method for detecting and locating fan faults based on a microphone array proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0128] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware, and of course, it can also be implemented by hardware. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk, or optical disc of a computer, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present invention.

Claims

1. A method for detecting and locating fan faults based on a microphone array, characterized in that: include: Acquire a sound signal, and perform a first preprocessing on the sound signal to obtain a first sound signal; The location of the fault sound source is obtained by calculating the time difference of the first sound signal received by each microphone; Optimizing the position of the fault sound source using a first optimization method to obtain the coordinates of the fault sound source position; Based on the coordinates of the fault sound source position, the fault type is classified to obtain a classification result; Acquire real-time monitoring data, and perform modeling based on the classification results to obtain a first neural network model; Historical operation data and historical fault information are obtained, the first neural network model is trained, and the trained first neural network model is verified, parameter prediction values ​​are output, and the parameter prediction values ​​are judged to obtain fault results.

2. The method for detecting and locating fan faults based on a microphone array according to claim 1, characterized in that: Performing a first preprocessing on the sound signal includes: Octave analysis is used to decompose the sound signal into sub-bands, and signal processing methods are introduced for optimization; Adjust the number and range of sub-bands according to signal characteristics, and extract sub-band acoustic energy characteristics by combining multi-scale fusion; The sub-band acoustic energy characteristics are normalized to obtain a first sound signal.

3. The method for detecting and locating fan faults based on a microphone array according to claim 2, characterized in that: Calculating the time difference of the first sound signal received by each microphone includes: Select a reference microphone as the reference point; Perform cross-correlation operation on the signal collected by the reference microphone and the signals of other microphones; The similarities between the signals are compared, and the time difference corresponding to the maximum similarity is the delay value, that is, the time delay of other microphones relative to the reference microphone.

4. The method for detecting and locating fan faults based on a microphone array according to claim 3, characterized in that: Also includes: Build a geometric model based on microphone positions, time delays, and sound propagation speed; Using geometric relationships, the delay value is converted into the actual distance from the sound source to each microphone.

5. The method for detecting and locating fan faults based on a microphone array according to claim 3 or 4, characterized in that: Optimizing the fault sound source position using the first optimization method includes: Calculate the delay value from the sound source to each microphone and compare it with the known delay value to obtain the target function; Set the initial estimated position; The objective function is optimized using the least squares method to obtain the sound source position that best matches the delay value, that is, the coordinates of the fault sound source position.

6. The method for detecting and locating fan faults based on a microphone array according to claim 5, characterized in that: Modeling based on the classification results includes: The real-time monitoring data and the classification results are combined into a two-dimensional matrix as input data to build a first neural network model; Obtain historical operation data and fault information marked in the historical data as a training set, optimize the model weights through supervised learning, and obtain the first neural network model after training; The historical operation data, the fault information marked in the historical data, the classification results and the sub-band acoustic energy characteristics are used as the input of the trained first neural network model to obtain the parameter prediction value.

7. The method for detecting and locating fan faults based on a microphone array according to claim 6, characterized in that: Determining the parameter prediction value includes: A first threshold value of the parameter is determined according to the historical operation data and the fault information marked in the historical data. If the parameter is greater than the first threshold value, the trained first neural network outputs a fault warning.

8. A system using the method for detecting and locating fan faults based on a microphone array as described in any one of claims 1 to 7, characterized in that: include: A preprocessing module, used to obtain a sound signal, and perform a first preprocessing on the sound signal to obtain a first sound signal; A calculation module, used for obtaining the location of the fault sound source by calculating the time difference of the first sound signal received by each microphone; A positioning module, used to optimize the position of the fault sound source by using a first optimization method to obtain the position coordinates of the fault sound source; A classification module, used to classify the fault type based on the position coordinates of the fault sound source to obtain a classification result; A model building module, used to obtain real-time monitoring data, and to build a model based on the classification results to obtain a first neural network model; The judgment module is used to obtain historical operation data and historical fault information, train the first neural network model, verify the trained first neural network model, output parameter prediction values, and judge the parameter prediction values ​​to obtain fault results.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the fan fault detection and positioning method based on a microphone array as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the fan fault detection and positioning method based on a microphone array as described in any one of claims 1 to 7.

Images