Air energy equipment detection method, equipment and medium based on vibration signal

Abstract

The embodiment of the present application discloses a method, device and medium for detecting air energy equipment based on vibration signals, which belongs to the technical field of equipment fault detection and solves the problem that traditional air energy equipment detection methods rely on manual analysis of vibration data, which is time-consuming and labor-intensive, and has certain errors and delays. It includes obtaining vibration signals corresponding to different air energy equipment to be tested; denoising the vibration signals by fusing adaptive filters with wavelet transforms; encoding and decoding the denoised vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy equipment to be tested; performing correlation information analysis on the feature data by presetting a multi-head self-attention mechanism model to determine different operating states corresponding to the air energy equipment to be tested; constructing a device association graph based on the air energy equipment state, so as to determine cross-device maintenance information corresponding to multiple air energy equipment to be tested based on the device association graph.

Description

Technical Field

[0001] The present application relates to the technical field of equipment fault detection, and in particular to a method, equipment and medium for detecting air energy equipment based on vibration signals. Background Art

[0002] In the current field of industrial equipment maintenance and management, air energy equipment, as a key energy conversion device, has a significant impact on the overall system performance and energy consumption due to its operational stability and efficiency. To promptly identify potential equipment failures and implement maintenance, real-time monitoring and diagnosis of air energy equipment's operational status is crucial.

[0003] Vibration data is a crucial source of information reflecting the operating status of equipment. The accuracy and timeliness of its processing and analysis directly impact the accuracy and timeliness of fault diagnosis. Traditional methods for inspecting air-cooled equipment rely heavily on manual analysis of vibration data, a process typically encompassing data acquisition, signal processing, and feature extraction. These steps largely rely on the operator's experience and skill level, and are susceptible to subjective factors. Furthermore, this method is time-consuming and labor-intensive, subject to errors and delays. Consequently, with the increasing number of equipment and rising monitoring requirements, the efficiency and accuracy of manual analysis methods are no longer sufficient to meet practical needs. Summary of the Invention

[0004] The embodiments of the present application provide an air energy equipment detection method, equipment and medium based on vibration signals, which are used to solve the following technical problems: Traditional air energy equipment detection methods mostly rely on manual analysis of vibration data, which is time-consuming and labor-intensive, and has certain errors and delays. With the increase in the number of equipment and the improvement of monitoring requirements, the efficiency and accuracy of manual analysis methods have been difficult to meet actual needs.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] The embodiment of the present application provides an air energy device detection method based on vibration signals. The method comprises the following steps: obtaining vibration signals corresponding to different air energy devices to be tested by setting vibration sensors at different positions of the air energy devices to be tested; performing noise reduction processing on the vibration signals by fusing an adaptive filter with a wavelet transform; encoding and decoding the noise-reduced vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy devices to be tested; performing correlation information analysis on the feature data by presetting a multi-head self-attention mechanism model to determine the probability values ​​of the air energy devices to be tested corresponding to the feature data being in different operating states, so as to determine the air energy device states corresponding to the different air energy devices to be tested based on the probability values; and constructing a device association graph based on the air energy device states, so as to determine cross-device maintenance information corresponding to multiple air energy devices to be tested based on the device association graph.

[0007] The embodiment of the present application uses an adaptive filter to dynamically adjust the filter parameters according to the real-time characteristics of the signal, effectively suppressing noise interference; the wavelet transform can perform multi-resolution analysis on the signal, finely remove noise in different frequency bands, and retain the key features of the signal. Secondly, the embodiment of the present application uses a multi-layer noise reduction autoencoder to encode and decode the noise-reduced vibration signal layer by layer, which can automatically learn the complex features contained in the vibration signal and effectively extract feature data that can accurately reflect the operating status of the equipment. Combined with the preset multi-head self-attention mechanism model, the probability value of the air energy equipment to be tested in different operating states is accurately determined. In addition, the embodiment of the present application constructs an equipment association diagram based on the equipment status, clearly showing the relationship between different air energy equipment to be tested, which helps maintenance personnel to grasp the equipment operation status as a whole, quickly determine the cross-equipment maintenance information corresponding to multiple air energy equipment to be tested, improve the scientific nature of maintenance decisions and the efficiency of maintenance work, reduce equipment downtime, and reduce maintenance costs.

[0008] In one implementation of the present application, the vibration signal is denoised by fusing an adaptive filter with a wavelet transform, specifically comprising: dynamically adjusting the step size factor based on the initial step size, the maximum step size, and the minimum step size; constructing a weight update function of the adaptive filtering algorithm based on the dynamically adjusted step size factor, the filter coefficient vector, the error, and the vibration signal to iteratively update the filter coefficient; constructing an adaptive filter based on a recursive neural network using the iteratively updated filter coefficients, and adaptively filtering the vibration signal based on the vibration signal at the current moment and the output corresponding to the adaptive filter at the previous moment; performing a wavelet transform on the filtered vibration signal, determining an error value between the wavelet transformed signal and the current output signal of the adaptive filter, and updating and adjusting the coefficients of the adaptive filter according to the error size and error direction; sparsely representing the wavelet transformed vibration signal on an overcomplete dictionary; screening the sparse representation coefficients based on the sparse representation coefficients and a preset coefficient threshold; and reconstructing the vibration signal using the screened coefficients and the overcomplete dictionary to obtain a denoised vibration signal.

[0009] In one implementation of the present application, the denoised vibration signal is encoded and decoded layer by layer based on a multi-layer denoising autoencoder to obtain characteristic data corresponding to different air energy equipment to be tested, specifically including: determining the low-frequency energy band and the high-frequency noise band based on the energy distribution and noise level corresponding to the vibration signal sample; increasing the noise intensity in the low-frequency energy band, and injecting the noise type with the same noise characteristics into the high-frequency noise band; discriminating the encoded features of the original signal and the encoded features after noise pollution through a preset discriminator, and adjusting the parameters of the multi-layer denoising autoencoder based on the discrimination result; decoding the multi-layer denoising autoencoder after the vibration signal input parameter adjustment, determining the first correlation between the encoded features and the decoded features through the global attention mechanism, and determining the second correlation between the frequency bands in the encoded features through the local attention mechanism; assigning weights to each encoded feature based on the first correlation and the second correlation to obtain a reconstructed vibration signal, and obtaining the characteristic data corresponding to different air energy equipment to be tested based on the reconstructed vibration signal.

[0010] In one implementation of the present application, weighting each coding feature based on the first correlation and the second correlation to obtain a reconstructed vibration signal specifically includes: using the function:

[0011]

[0012] The first correlation and the second correlation are weighted to each encoding feature; where α is the adjustment parameter; β is the adjustment parameter; W is the weight; G is the global attention score vector; L is the local attention score matrix; L ij is the local correlation between feature i and feature j; E is the encoded feature vector; n is the dimension; based on the function:

[0013] E i '=W i ·Ei;

[0014] The reconstructed vibration signal is obtained; wherein, E' is the weighted eigenvector; W is the weight; and E is the encoded eigenvector.

[0015] In one implementation of the present application, a multi-head self-attention mechanism model is preset to perform correlation information analysis on the feature data to determine the probability value of the air energy equipment to be tested corresponding to the feature data in different operating states, specifically including: determining the number of self-attention heads based on the dimension and data complexity corresponding to the feature data; in the constructed multi-head self-attention mechanism stacking model structure, processing the output of the next layer based on the output of the previous layer, and performing nonlinear transformation and feature integration on the feature representation output by the multi-head self-attention mechanism stacking model through a preset feedforward neural network; and converting the output of the multi-head self-attention mechanism stacking model into a probability value of the air energy equipment to be tested in different operating states through a preset Softmax activation function.

[0016] In one implementation of the present application, the output of the multi-head self-attention mechanism stacking model is converted into a probability value of the air energy equipment to be tested being in different operating states by presetting a Softmax activation function, specifically including: based on the function:

[0017]

[0018] in,

[0019] Determine the probability value of the air energy equipment under test in different operating states; where P j is the probability value; z j is the output of the multi-head self-attention mechanism stacking model, where n is the number of categories of air energy equipment operating status; the air energy equipment state transfer weight matrix is ​​W = [w ij ] n×n ; u is the characteristic uncertainty coefficient; ρ j is the correction factor; i is the state category i of the air energy equipment; j is the state category j of the air energy equipment; γ is the first adjustment parameter; is the second adjustment parameter; ACF i (τ) is the autocorrelation function; τ is the delay time; (h i -l i ) is the model uncertainty interval.

[0020] In one implementation of the present application, a device association diagram is constructed based on the state of the air energy equipment, so as to determine the cross-device maintenance information corresponding to multiple air energy equipment to be tested based on the device association diagram, specifically including: constructing a device association diagram based on the positions between different air energy equipment to be tested, the similarities between different air energy equipment to be tested, and the production process relationships between different air energy equipment to be tested; marking the air energy equipment in the same state in the device association diagram, and determining the reference air energy equipment in the fault state; determining the fault propagation path corresponding to each reference air energy equipment in the device association diagram, and determining the associated reference air energy equipment based on the fault propagation path; labeling the associated reference air energy equipment caused by the same fault cause in the same way; grouping the reference air energy equipment and the associated reference air energy equipment based on the air energy equipment labeling, so as to construct corresponding maintenance information for the air energy equipment in different groups.

[0021] In one implementation of the present application, the fault propagation path corresponding to each reference air energy device is determined in the device association diagram, and the associated reference air energy device is determined based on the fault propagation path, specifically including: in the historical air energy device fault database, the historical fault air energy device combination corresponding to the reference air energy device is determined, and the air energy device fault history data is determined based on the historical fault air energy device combination; according to the device association diagram and the air energy device fault history data, a Bayesian network model is constructed; wherein, in the Bayesian network model, the air energy device state is used as a node, the association relationship between the air energy devices is used as an edge, and the weight of the edge is used as a conditional probability; the reference air energy device is used as the starting point of the faulty air energy device, and the fault propagation probability between the air energy devices and the reference propagation path are determined through Bayesian network reasoning; the probabilities of different propagation paths are sorted, and the fault propagation path and the associated reference air energy device are determined based on the sorting results.

[0022] An embodiment of the present application provides an air energy equipment detection device based on vibration signals, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: obtain vibration signals corresponding to different air energy equipment to be tested through vibration sensors set at different positions of the air energy equipment to be tested; perform noise reduction processing on the vibration signals through fusion of an adaptive filter and a wavelet transform; encode and decode the noise-reduced vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy equipment to be tested; perform correlation information analysis on the feature data through a preset multi-head self-attention mechanism model to determine the probability value of the air energy equipment to be tested corresponding to the feature data being in different operating states, so as to determine the air energy equipment states corresponding to the different air energy equipment to be tested based on the probability value; construct a device association graph based on the air energy equipment states, so as to determine cross-device maintenance information corresponding to multiple air energy equipment to be tested based on the device association graph.

[0023] A non-volatile computer storage medium provided by an embodiment of the present application stores computer-executable instructions, wherein the computer-executable instructions are configured to: obtain vibration signals corresponding to different air energy devices to be tested by means of vibration sensors arranged at different positions of the air energy devices to be tested; perform noise reduction processing on the vibration signals by fusing an adaptive filter with a wavelet transform; encode and decode the noise-reduced vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy devices to be tested; perform correlation information analysis on the feature data by means of a preset multi-head self-attention mechanism model to determine the probability values ​​of the air energy devices to be tested corresponding to the feature data being in different operating states, so as to determine the air energy device states corresponding to the different air energy devices to be tested based on the probability values; construct a device association graph based on the air energy device states, so as to determine cross-device maintenance information corresponding to multiple air energy devices to be tested based on the device association graph.

[0024] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: the embodiments of the present application dynamically adjust the filter parameters according to the real-time characteristics of the signal through an adaptive filter, effectively suppressing noise interference; the wavelet transform can perform multi-resolution analysis on the signal, finely removing noise in different frequency bands while retaining the key features of the signal. Secondly, the embodiments of the present application use a multi-layer noise reduction autoencoder to encode and decode the noise-reduced vibration signal layer by layer, automatically learning the complex features contained in the vibration signal and effectively extracting feature data that can accurately reflect the operating status of the equipment. Combined with a preset multi-head self-attention mechanism model, the probability value of the air energy equipment under test being in different operating states is accurately determined. In addition, the embodiments of the present application construct a device association diagram based on the device status, clearly displaying the relationship between different air energy equipment under test, helping maintenance personnel to grasp the equipment operation status from a holistic perspective and quickly determine the cross-device maintenance information corresponding to multiple air energy equipment under test, improving the scientific nature of maintenance decisions and the efficiency of maintenance work, reducing equipment downtime and reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0026] Figure 1 A flow chart of a method for detecting air energy equipment based on vibration signals provided in an embodiment of the present application;

[0027] Figure 2 A schematic structural diagram of an air energy equipment detection device based on vibration signals provided in an embodiment of the present application.

[0028] Reference numerals:

[0029] 200: Air energy equipment detection device based on vibration signal, 201: Processor, 202: Memory. DETAILED DESCRIPTION

[0030] The embodiments of the present application provide a method, device, and medium for detecting air energy equipment based on vibration signals.

[0031] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0032] The technical solutions proposed in the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0033] Figure 1 This is a flow chart of a method for detecting air energy equipment based on vibration signals provided in an embodiment of the present application. Figure 1 As shown, the air energy equipment detection method based on vibration signals includes the following steps:

[0034] Step 101: Acquire vibration signals corresponding to different air energy devices to be tested by using vibration sensors arranged at different positions of the air energy devices to be tested.

[0035] In one implementation of this application, multiple vibration sensors are installed at different locations on the air-energy equipment to be tested. These locations may include key components, vibration-sensitive areas, or areas requiring monitoring. For example, vibration sensors can be placed on key components such as the compressor, fan, condenser, and water pump. The vibration signals collected by the vibration sensors are converted into electrical signals and transmitted to a data analysis and processing module.

[0036] Step 102: noise reduction processing is performed on the vibration signal by fusing the adaptive filter with the wavelet transform.

[0037] In one implementation of the present application, the step size factor is dynamically adjusted based on the initial step size, the maximum step size, and the minimum step size. Based on the dynamically adjusted step size factor, the filter coefficient vector, the error, and the vibration signal, a weight update function of the adaptive filtering algorithm is constructed to iteratively update the filter coefficients. An adaptive filter based on a recursive neural network is constructed using the iteratively updated filter coefficients, and the vibration signal is adaptively filtered based on the vibration signal at the current moment and the output corresponding to the adaptive filter at the previous moment. The filtered vibration signal is subjected to a wavelet transform, and the error value between the wavelet transformed signal and the current output signal of the adaptive filter is determined. The coefficients of the adaptive filter are updated and adjusted based on the error size and error direction. The wavelet transformed vibration signal is sparsely represented on an overcomplete dictionary. The sparse representation coefficients are screened based on the sparse representation coefficients and a preset coefficient threshold. The vibration signal is reconstructed using the screened coefficients and the overcomplete dictionary to obtain a denoised vibration signal.

[0038] Specifically, the step size factor plays a vital role in the adaptive filtering algorithm, which determines the speed and stability of the filter coefficient update. Let the initial step size be μ0 and the maximum step size be μ max , the minimum step size is μ min The step size factor is changed according to the variance of the signal or the change of the error. For example, when the error is large, the step size factor can be increased to speed up the convergence speed; when the error gradually decreases and tends to be stable, the step size factor can be reduced to reduce the steady-state error. The dynamic adjustment formula can be expressed as:

[0039]

[0040] Where μ(n) is the step size factor at time n, e(n) is the error at time n, and N is the number of samples in a certain time window before the current time. The step size factor is dynamically adjusted between the minimum and maximum values ​​based on the ratio of the current error to the maximum error in the time window.

[0041] Assume the filter coefficient vector is w(n), the input vibration signal is x(n), and the expected response is d(n), then the filter output is y(n) = w T (n)·x(n), error e(n)=d(n)-y(n). The weight update function based on the least mean square algorithm is:

[0042] w(n+1)=w(n)+2μ(n)e(n)x(n);

[0043] Where 2μ(n)e(n)x(n) represents the direction and magnitude of the filter coefficients adjusted based on the current step size factor, error, and input signal. By continuously iterating this update formula, the filter coefficients gradually converge to the optimal value, enabling the filter to better filter the vibration signal.

[0044] Furthermore, recursive neural networks have a memory function and can handle long-term dependencies in time series data. In the adaptive filter of the embodiment of the present application, in addition to inputting the vibration signal x(n) at the current moment, the output y(n-1) of the adaptive filter at the previous moment is also considered. The hidden layer of the recursive neural network processes this input information and outputs a predicted value for the signal at the current moment. The error between the predicted value and the expected response is used to further adjust the filter coefficients, thereby achieving adaptive filtering of the vibration signal.

[0045] Furthermore, the vibration signal after adaptive filtering is subjected to a wavelet transform, and then the coefficients of the adaptive filter are further adjusted by comparing the error between the wavelet transformed signal and the current output signal of the adaptive filter. Specifically, the wavelet transform can decompose the signal into sub-band signals of different frequency bands, analyze the signal at different time and frequency resolutions, determine the signal after the wavelet transform and the current output signal of the adaptive filter, calculate the error between the two, and adjust the coefficients of the adaptive filter based on the size and direction of the error using a weight update function similar to the aforementioned one. For example, if the error is large and positive, it means that the output of the adaptive filter is too small, and the correlation coefficient needs to be increased to improve the output; conversely, if the error is negative and large, the correlation coefficient needs to be reduced. In this way, the performance of the adaptive filter is continuously optimized.

[0046] Furthermore, the vibration signal after wavelet transformation is sparsely represented on an overcomplete dictionary, and then screened according to the sparse representation coefficients and the preset coefficient threshold, and finally reconstructed to obtain the vibration signal after noise reduction. Specifically, on the overcomplete dictionary, the vibration signal can be represented as a linear combination of a few atoms, and the sparse representation of the signal can be obtained by solving the optimization problem. For example, the orthogonal matching pursuit algorithm is used to solve the sparse representation coefficients. A preset coefficient threshold is set, and the sparse representation coefficient vector is screened, and the coefficients with absolute values ​​greater than the preset coefficient threshold are retained, and the remaining coefficients are set to zero to obtain the screened coefficient vector. Then, the vibration signal is reconstructed using the screened coefficients and the overcomplete dictionary. In this way, the noise and redundant components in the signal are removed to obtain the vibration signal after noise reduction.

[0047] Step 103: Encode and decode the noise-reduced vibration signal layer by layer based on a multi-layer noise reduction autoencoder to obtain characteristic data corresponding to different air energy equipment to be tested.

[0048] In one implementation of the present application, based on the energy distribution and noise level corresponding to the vibration signal sample, the low-energy frequency band and the high-noise frequency band are determined. The noise intensity is increased in the low-energy frequency band, and the noise type with the same noise characteristics is injected into the high-noise frequency band. The features after encoding the original signal and the encoded features after noise pollution are discriminated by a preset discriminator, so as to adjust the parameters of the multi-layer denoising autoencoder based on the discrimination result. The multi-layer denoising autoencoder after the vibration signal input parameter adjustment is decoded, and the first correlation between the encoded features and the decoded features is determined by the global attention mechanism, and the second correlation between the frequency bands in the encoded features is determined by the local attention mechanism. Based on the first correlation and the second correlation, each encoded feature is weighted to obtain a reconstructed vibration signal, so as to obtain the characteristic data corresponding to different air energy equipment to be tested based on the reconstructed vibration signal.

[0049] Specifically, by analyzing vibration signal samples and calculating the energy distribution of the signal within different frequency ranges, we can identify low-frequency bands where energy is primarily concentrated. For example, we perform a Fourier transform on the collected vibration signal of an air-conditioning device, converting it to the frequency domain and calculating the power spectrum density values ​​at different frequency points. By observing the power spectrum density curve, we can identify frequency ranges with relatively high and concentrated power values ​​in the low-frequency portion as low-energy bands. In the high-frequency portion, frequency ranges with large amplitude fluctuations that exceed the normal signal range can be identified as high-frequency noise bands.

[0050] Furthermore, in the determined low-energy frequency band, in order to enhance the adaptability of the multi-layer denoising autoencoder to low-frequency noise, the noise intensity is appropriately increased. This can be done by generating a noise signal (such as Gaussian white noise) with similar characteristics to the original low-frequency band noise, adjusting its amplitude as needed, and then superimposing it on the low-energy frequency band portion of the original vibration signal. In the high-frequency noise band, a noise type with the same characteristics as the noise in that frequency band is injected. If it is known that the noise in the high-frequency noise band is high-frequency electromagnetic noise, a noise signal with similar spectral characteristics can be generated by simulating the generation mechanism of high-frequency electromagnetic noise, and then injected into the high-frequency noise band of the original vibration signal.

[0051] Furthermore, a pre-set discriminator is used to distinguish between the encoded features of the original signal and the encoded features contaminated by noise. The discriminator is typically built on a neural network, such as a convolutional neural network or a recurrent neural network. It extracts and classifies the input encoded features, determining whether they originate from the original signal or the noise-contaminated signal. Based on the discriminant results, a backpropagation algorithm is used to adjust the parameters of the multi-layer denoising autoencoder. If the discriminator mistakenly identifies the noise-contaminated encoded features as the original signal, this indicates that the multi-layer denoising autoencoder's noise suppression capabilities are insufficient, and its parameters need to be adjusted to enhance the noise reduction effect.

[0052] Furthermore, the vibration signal is input into a multi-layer denoising autoencoder with adjusted parameters for decoding. The global attention mechanism traverses all encoded features and the feature representations generated at each moment in the decoding process when calculating the first correlation between the encoded and decoded features. The similarity between the encoded and decoded feature vectors in different dimensions is evaluated through dot product operations or cosine similarity calculations. The local attention mechanism focuses on the relationship between the frequency bands in the encoded features. Because vibration signals have different frequency components, features in different frequency bands may be correlated. The local attention mechanism calculates the second correlation by analyzing the relationship between feature vectors of adjacent frequency bands or frequency bands with specific frequency intervals.

[0053] Furthermore, for each coding feature, the weight is comprehensively determined based on its first correlation with the decoding feature and the second correlation between the frequency bands. First, based on the first correlation, the similarity values ​​of each coding feature and all decoding features are weighted and summed to obtain the weight based on the first correlation. Then, the weight is adjusted according to the second correlation. If the frequency band where the coding feature is located has a strong correlation with other important frequency bands, its weight is appropriately increased; otherwise, the weight is appropriately reduced. After determining the weight of each coding feature, each coding feature is multiplied by its corresponding weight through the decoding process, and then a reconstructed signal is generated through the operation of the decoding layer. The reconstructed vibration signal contains key information about the operating status of the equipment. For different air-energy devices to be tested, further feature extraction operations are performed on the reconstructed signal, such as time domain analysis to calculate the mean, variance, peak index, etc.; frequency domain analysis to extract frequency components, power spectrum density, etc.; or time-frequency domain joint analysis, such as wavelet transform coefficients, etc., to obtain the corresponding feature data.

[0054] Specifically, when weighting each coding feature based on the first correlation and the second correlation to obtain the reconstructed vibration signal, the embodiment of the present application uses the function:

[0055]

[0056] The first correlation and the second correlation are weighted to each encoding feature; where α is the adjustment parameter; β is the adjustment parameter; W is the weight; G is the global attention score vector; L is the local attention score matrix; L ij is the local correlation between feature i and feature j; E is the encoded feature vector; n is the dimension;

[0057] Function-based:

[0058] E i '=W i ·Ei;

[0059] The reconstructed vibration signal is obtained; wherein, E' is the weighted eigenvector; W is the weight; and E is the encoded eigenvector.

[0060] Step 104: By presetting a multi-head self-attention mechanism model, the feature data is analyzed for correlation information to determine the probability values ​​of the air energy equipment to be tested corresponding to the feature data being in different operating states, so as to determine the air energy equipment states corresponding to different air energy equipment to be tested based on the probability values.

[0061] In one implementation of the present application, the number of self-attention heads is determined based on the dimension and data complexity corresponding to the feature data. In the constructed multi-head self-attention mechanism stacking model structure, the output of the next layer is processed based on the output of the previous layer, and the feature representation of the output of the multi-head self-attention mechanism stacking model is nonlinearly transformed and integrated through a preset feedforward neural network. By presetting the Softmax activation function, the output of the multi-head self-attention mechanism stacking model is converted into a probability value of the air energy equipment to be tested being in different operating states.

[0062] Specifically, the dimensionality and complexity of feature data are key factors in determining the number of self-attention heads. When feature data has a high dimensionality, it means the data contains rich information, requiring more self-attention heads to capture this information from different angles. For example, if the feature data covers multiple dimensions such as the vibration frequency, temperature, and pressure of air-conditioning equipment, the high dimensionality makes it difficult for a single self-attention head to fully capture the relationships between all dimensions. In this case, increasing the number of self-attention heads, with each head focusing on the feature associations of different dimensions or combinations of dimensions, enables a more comprehensive analysis of the data.

[0063] Regarding data complexity, if the feature data exhibits complex distribution patterns or contains multiple potential feature interactions, more self-attention heads are needed to mine these complex relationships. For example, when air-energy equipment operates under complex conditions, its vibration signal characteristics may interact with multiple operating parameters, and this influence is not simply linear. As data complexity increases, more self-attention heads can analyze data from different subspaces through parallel computing, thereby better handling such complex situations.

[0064] Furthermore, in the constructed multi-head self-attention stacked model structure, each layer processes based on the output of the previous layer. The feature representation output by the previous layer contains the preliminary analysis results of the input data. The next layer uses the multi-head self-attention mechanism to further extract features and mine relationships on these feature representations. Each self-attention head performs a weighted summation of the input features in different subspaces, focusing on different aspects of feature correlation. For example, one self-attention head may focus on the long-range dependencies between vibration frequency features, while another head focuses on the correlation between vibration frequency and device operating time.

[0065] Furthermore, by stacking multiple layers, the model can gradually delve deeper into complex feature relationships within the data. The output of each layer further abstracts and integrates features based on the previous layer, resulting in a more comprehensive and in-depth understanding of the data. For example, the bottom layer may primarily capture local features and simple correlations in the data. As the number of layers increases, the upper layers gradually focus on global features and complex cross-dimensional correlations. For example, in a three-layer stacked model with a multi-head self-attention mechanism, the first layer receives preprocessed feature data from air conditioning equipment, processes this data through 15 self-attention heads, and outputs a set of feature representations that initially capture the correlations between the feature data, such as the different frequency bands of the vibration signal. The second layer takes the output of the first layer as input and processes it again through 15 self-attention heads. At this point, the model may capture the correlation between the vibration signal features and the energy consumption characteristics of the equipment. The third layer, based on the output of the second layer, further explores the potential connections between these features and the equipment's failure history, outputting a more advanced feature representation that provides more valuable information for subsequent analysis.

[0066] Furthermore, feedforward neural networks typically include multiple fully connected layers. Although the feature representations output by the multi-head self-attention mechanism stacked model already contain rich feature association information, they still require further nonlinear transformation and feature integration. Feedforward neural networks perform nonlinear transformations on the input feature representations through fully connected layers, such as using the ReLU activation function. This increases the model's nonlinear expression capabilities and enables it to learn more complex functional relationships. Simultaneously, the connection weights between different neurons perform a weighted summation of the features to achieve feature integration. In this way, the feature information extracted by the multi-head self-attention mechanism is mapped to a new space, making the feature representation more discriminative and providing more effective data for the subsequent determination of the probability value of the device's operating status.

[0067] Furthermore, the Softmax activation function converts the output of the multi-head self-attention stack model, processed by the feedforward neural network, into probabilities of the air conditioning equipment under test being in different operating states. The Softmax function converts each element in the input vector into a probability value, with the sum of all probabilities being 1. In this way, the model output is converted into a probability distribution, with each probability value representing the likelihood of the equipment being in the corresponding operating state.

[0068] Specifically, the output of the multi-head self-attention mechanism stacking model is converted into the probability value of the air energy equipment to be tested being in different operating states by presetting the Softmax activation function, which can be:

[0069] Function-based:

[0070]

[0071] in,

[0072] Determining the probability values ​​of the air energy equipment to be tested being in different operating states;

[0073] Among them, P j is the probability value; zj is the output of the multi-head self-attention mechanism stacking model, where n is the number of categories of air energy equipment operating status; the air energy equipment state transfer weight matrix is ​​W = [w ij ] n×n ; u is the characteristic uncertainty coefficient; ρ j is the correction factor; i is the state category i of the air energy equipment; j is the state category j of the air energy equipment; γ is the first adjustment parameter; is the second adjustment parameter; ACF i (τ) is the autocorrelation function; τ is the delay time; (h i -l i ) is the model uncertainty interval.

[0074] Step 105: construct a device association diagram based on the state of the air energy equipment, so as to determine cross-device maintenance information corresponding to the plurality of air energy equipment to be tested based on the device association diagram.

[0075] In one implementation of the present application, an equipment association diagram is constructed based on the positions of different air energy devices to be tested, the similarities between different air energy devices to be tested, and the production process relationships between different air energy devices to be tested. Air energy devices in the same state are marked in the equipment association diagram, and reference air energy devices in a faulty state are determined. In the equipment association diagram, the fault propagation paths corresponding to each reference air energy device are determined, and based on the fault propagation paths, the associated reference air energy devices are determined. The associated reference air energy devices caused by the same fault cause are marked in the same way. The reference air energy devices and the associated reference air energy devices are grouped based on the air energy device markings to construct corresponding maintenance information for the air energy devices in different groups.

[0076] Specifically, the location coordinates of different air-energy devices to be tested are obtained. Similarity is evaluated from multiple dimensions of the devices, including device model, operating parameters, and fault history. For device models, devices of the same model have similar structures and performance, and the types and probabilities of faults may also be similar. In terms of operating parameters, the cooling / heating power, compressor speed, refrigerant pressure, and other parameters of the devices are compared, and the similarity is determined by calculating the cosine similarity of the parameter vectors. Furthermore, the role of air-energy devices in the production process and their upstream and downstream relationships are analyzed. For example, in some industrial production scenarios, air-energy devices may be used to provide hot or cold water of a specific temperature for the production process. If the hot water produced by device A is directly supplied to device B for use in the production process, then there is a strong correlation between device A and device B. By constructing a directed graph of the production process, the direction of material and energy flow between devices is clarified, and the correlation relationship and weight are determined. For devices at key production process nodes, their correlation weight with upstream and downstream devices is higher.

[0077] Furthermore, the operating status of air-energy devices is monitored in real time, categorizing them into normal operation, minor faults, and major faults. Devices in the same state are represented by nodes of the same color or shape in the device association diagram. When an air-energy device fails, the device association diagram selects devices that are similar to the failed device in terms of model, operating parameters, and fault history, and are in a faulty state, as reference devices.

[0078] Furthermore, within the device association diagram, the fault propagation path is determined. Based on the determined fault propagation path, other faulty devices or devices at risk of failure along the path are identified and used as associated reference devices. Associated reference devices with the same fault cause are labeled identically, including information such as the fault cause and fault type. For example, if multiple air-energy units experience refrigerant leakage due to aging refrigerant pipes, these devices are collectively labeled as the "Refrigerant Pipe Aging Leakage Fault Group." The labeling details the fault cause and potential impact, allowing maintenance personnel to quickly identify common faults. Based on the device labels, reference devices and associated reference devices are grouped. Grouping can be done by fault type, such as grouping all devices with compressor failures in one group and devices with electrical failures in another. Alternatively, grouping can be done by fault cause, such as grouping devices with environmental factors in one group. When grouping, factors such as device location and similarity are considered, and devices that facilitate centralized maintenance are grouped together whenever possible. Develop specific maintenance information for each group of air-energy units. For compressor failures, maintenance information includes compressor model, replacement procedures, maintenance tool list, and estimated repair time. For electrical failures, maintenance information covers electrical circuit inspection methods, faulty component replacement instructions, and safety precautions. Furthermore, maintenance personnel and resources are rationally deployed based on equipment grouping to improve maintenance efficiency.

[0079] In one implementation of the present application, in a historical air energy device failure database, a combination of historical faulty air energy devices corresponding to a reference air energy device is determined, so as to determine the historical data of air energy device failures based on the combination of historical faulty air energy devices. A Bayesian network model is constructed based on the device association graph and the historical data of air energy device failures; wherein, in the Bayesian network model, the air energy device status is used as a node, the association relationship between air energy devices is used as an edge, and the weight of the edge is used as a conditional probability. The reference air energy device is used as the starting point of the faulty air energy device, and the propagation probability of the fault between air energy devices and the reference propagation path are determined through Bayesian network reasoning. The probabilities of different propagation paths are sorted, and the fault propagation path and the associated reference air energy device are determined based on the sorting results.

[0080] Specifically, within the historical air-energy unit failure database, for a reference air-energy unit, the system searches for fields such as the device model, fault type, and fault occurrence time to identify historically failed air-energy units with similar fault characteristics, thereby forming a historically failed air-energy unit combination. For this historically failed air-energy unit combination, detailed historical failure data is extracted. This data includes the equipment's operating parameters before the failure, such as cooling / heating power, refrigerant pressure, and compressor speed; environmental parameters at the time of the failure, such as room temperature and humidity; the frequency of the failure; and repair records, such as the repair time, repair measures, and replaced parts.

[0081] Furthermore, a Bayesian network model is constructed based on the device association graph. Air-energy equipment states such as normal operation, minor faults, and major faults are used as nodes. Relationships between devices, such as connections based on proximity, similar models, and upstream and downstream production processes, are used as edges. Edge weights are determined according to the method previously determined when constructing the device association graph. For example, the connection weights for devices with similar locations may be calculated based on Euclidean distance, and the connection weights for devices with the same model may be based on structural and performance similarity assessments. These weights will serve as conditional probabilities in the Bayesian network. Using the determined historical data on air-energy equipment failures, the conditional probabilities between nodes are calculated using statistical methods. For example, in the statistical historical data, the number of times when device A is in a minor fault state, the number of times the associated device B is also in a minor fault state is divided by the total number of times device A is in a minor fault state to obtain the conditional probability that device B is in a minor fault state when device A is in a minor fault state. By calculating and setting the conditional probabilities between all node pairs, a complete Bayesian network model is constructed.

[0082] Furthermore, the reference air-energy device is used as the starting point for the faulty air-energy device, and inference is performed within the constructed Bayesian network model using a Bayesian network inference algorithm, such as the joint tree algorithm. The algorithm calculates the probability of the fault propagating between air-energy devices based on the fault state of the starting device and the conditional probabilities between nodes. The probabilities of different propagation paths are ranked from high to low. Based on the sorting results, the path with the highest probability is selected as the most likely fault propagation path. At the same time, other devices on the path (such as device B) and other devices associated with the starting device through this path are considered associated reference air-energy devices.

[0083] Figure 2 This is a schematic diagram of the structure of an air energy device detection device based on vibration signals provided in an embodiment of the present application. Figure 2 As shown, an air energy device detection device 200 based on a vibration signal includes: at least one processor 201; and a memory 202 communicatively connected to the at least one processor 201; wherein the memory 202 stores instructions that can be executed by the at least one processor 201, and the instructions are executed by the at least one processor 201 so that the at least one processor 201 can: obtain vibration signals corresponding to different air energy devices to be tested by vibration sensors set at different positions of the air energy devices to be tested; perform noise reduction processing on the vibration signals by fusing an adaptive filter with a wavelet transform; encode and decode the noise-reduced vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy devices to be tested; perform correlation information analysis on the feature data by presetting a multi-head self-attention mechanism model, determine the probability value of the air energy devices to be tested corresponding to the feature data being in different operating states, and determine the air energy device states corresponding to the different air energy devices to be tested based on the probability value; construct a device association graph based on the air energy device states, and determine cross-device maintenance information corresponding to multiple air energy devices to be tested based on the device association graph.

[0084] A non-volatile computer storage medium provided by an embodiment of the present application stores computer-executable instructions, wherein the computer-executable instructions are configured to: obtain vibration signals corresponding to different air energy devices to be tested by means of vibration sensors arranged at different positions of the air energy devices to be tested; perform noise reduction processing on the vibration signals by fusing an adaptive filter with a wavelet transform; encode and decode the noise-reduced vibration signals layer by layer based on a multi-layer noise reduction autoencoder to obtain feature data corresponding to different air energy devices to be tested; perform correlation information analysis on the feature data by means of a preset multi-head self-attention mechanism model to determine the probability values ​​of the air energy devices to be tested corresponding to the feature data being in different operating states, so as to determine the air energy device states corresponding to the different air energy devices to be tested based on the probability values; construct a device association graph based on the air energy device states, so as to determine cross-device maintenance information corresponding to multiple air energy devices to be tested based on the device association graph.

[0085] The various embodiments in this application are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the device, air energy equipment, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant parts, refer to the descriptions of the method embodiments.

[0086] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. It will be apparent to those skilled in the art that various modifications and variations may be made to the embodiments of the present application. However, such modifications or substitutions do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for detecting air energy equipment based on vibration signals, characterized in that: The method comprises: Vibration sensors are arranged at different positions of the air energy equipment to be tested to obtain vibration signals corresponding to different air energy equipment to be tested; Performing noise reduction processing on the vibration signal by fusing an adaptive filter with a wavelet transform; The noise-reduced vibration signal is encoded and decoded layer by layer based on a multi-layer noise reduction autoencoder to obtain characteristic data corresponding to the different air energy equipment to be tested; By presetting a multi-head self-attention mechanism model, the feature data is analyzed for correlation information to determine the probability values ​​of the air energy equipment to be tested corresponding to the feature data being in different operating states, so as to determine the air energy equipment states corresponding to the different air energy equipment to be tested based on the probability values; Based on the state of the air energy equipment, a device association diagram is constructed to determine cross-device maintenance information corresponding to a plurality of air energy equipment to be tested based on the device association diagram.

2. The air energy equipment detection method based on vibration signals according to claim 1, characterized in that: The noise reduction process of the vibration signal is performed by fusing the adaptive filter with the wavelet transform, specifically comprising: Dynamically adjust the step size factor based on the initial step size, maximum step size, and minimum step size; Constructing a weight update function of an adaptive filtering algorithm based on the dynamically adjusted step size factor, the filter coefficient vector, the error, and the vibration signal to iteratively update the filter coefficients; Constructing an adaptive filter based on a recursive neural network using the iteratively updated filter coefficients, and adaptively filtering the vibration signal based on the vibration signal at the current moment and the output corresponding to the adaptive filter at the previous moment; Performing a wavelet transform on the filtered vibration signal, determining an error between the wavelet transformed signal and a current output signal of the adaptive filter, and updating and adjusting the coefficients of the adaptive filter according to the error magnitude and error direction; sparsely representing the vibration signal after wavelet transformation on an overcomplete dictionary; Filtering the sparse representation coefficients based on the sparse representation coefficients and a preset coefficient threshold; The vibration signal is reconstructed using the filtered coefficients and the overcomplete dictionary to obtain a noise-reduced vibration signal.

3. The air energy equipment detection method based on vibration signals according to claim 1, characterized in that: The multi-layer noise reduction autoencoder is used to encode and decode the noise-reduced vibration signal layer by layer to obtain characteristic data corresponding to the different air energy equipment to be tested, specifically including: Based on the energy distribution and noise level corresponding to the vibration signal samples, the low-energy frequency band and the high-noise frequency band are determined; Increasing the noise intensity in the low-energy frequency band, and injecting a noise type with the same noise characteristics as the noise in the high-noise frequency band; The features after encoding the original signal and the features after being contaminated by noise are discriminated by a preset discriminator, so as to adjust the parameters of the multi-layer denoising autoencoder based on the discrimination result; Decoding the multi-layer denoising autoencoder after adjusting the vibration signal input parameters, determining a first correlation between the encoding features and the decoding features through a global attention mechanism, and determining a second correlation between frequency bands in the encoding features through a local attention mechanism; Based on the first correlation and the second correlation, weights are assigned to the coding features to obtain a reconstructed vibration signal, so as to obtain characteristic data corresponding to the different air energy equipment to be tested based on the reconstructed vibration signal.

4. The air energy equipment detection method based on vibration signals according to claim 1, characterized in that: The step of assigning weights to the coding features based on the first correlation and the second correlation to obtain a reconstructed vibration signal specifically includes: Through the function: The first correlation and the second correlation are weighted to each of the coding features; wherein α is an adjustment parameter; β is an adjustment parameter; W is a weight; G is a global attention score vector; L is a local attention score matrix; L ij is the local correlation between feature i and feature j; E is the encoded feature vector; n is the dimension; Function-based: HAVE BEEN i '=W i ·Oi; The reconstructed vibration signal is obtained; wherein, E' is the weighted eigenvector; W is the weight; and E is the encoded eigenvector.

5. The air energy equipment detection method based on vibration signals according to claim 1, characterized in that: The presetting of the multi-head self-attention mechanism model to perform correlation information analysis on the feature data and determine the probability values ​​of the air energy equipment to be tested corresponding to the feature data being in different operating states specifically includes: Determine the number of attention heads based on the dimension and data complexity corresponding to the feature data; In the constructed multi-head self-attention mechanism stacking model structure, the output of the next layer is processed based on the output of the previous layer, and the feature representation of the output of the multi-head self-attention mechanism stacking model is nonlinearly transformed and integrated through a preset feedforward neural network; By presetting a Softmax activation function, the output of the multi-head self-attention mechanism stacking model is converted into a probability value of the air energy equipment to be tested being in different operating states.

6. The air energy equipment detection method based on vibration signals according to claim 5, characterized in that: The method of converting the output of the multi-head self-attention mechanism stacking model into a probability value of the air energy equipment to be tested being in different operating states by presetting a Softmax activation function specifically includes: Function-based: in, Determining the probability values ​​of the air energy equipment to be tested being in different operating states; Among them, P j is the probability value; zj is the output of the multi-head self-attention mechanism stacking model, where n is the number of categories of air energy equipment operating status; the air energy equipment state transfer weight matrix is ​​W = [w ij ] n×n ; u is the characteristic uncertainty coefficient; ρ j is the correction factor; i is the state category i of the air energy equipment; j is the state category j of the air energy equipment; γ is the first adjustment parameter; is the second adjustment parameter; ACF i (τ) is the autocorrelation function; τ is the delay time; (h i -l i ) is the model uncertainty interval.

7. The air energy equipment detection method based on vibration signals according to claim 1, characterized in that: The step of constructing a device association diagram based on the state of the air energy equipment to determine cross-device maintenance information corresponding to a plurality of air energy equipment to be tested based on the device association diagram specifically includes: Constructing a device association diagram based on the positions of different air energy devices to be tested, the similarities between the different air energy devices to be tested, and the production process relationships between the different air energy devices to be tested; Marking the air energy devices in the same state in the device association diagram and determining a reference air energy device in a fault state; Determining the fault propagation paths corresponding to the reference air energy devices in the device association diagram, and determining the associated reference air energy devices based on the fault propagation paths; The same marking is applied to the associated reference air energy equipment caused by the same fault cause; The reference air energy devices and the associated reference air energy devices are grouped based on the air energy device labels, so as to construct corresponding maintenance information for the air energy devices in different groups.

8. The air energy equipment detection method based on vibration signals according to claim 7, characterized in that: Determining the fault propagation paths corresponding to the reference air energy devices in the device association diagram, and determining the associated reference air energy devices based on the fault propagation paths, specifically includes: In a historical air energy equipment failure database, determining a historical fault air energy equipment combination corresponding to the reference air energy equipment, and determining air energy equipment failure history data based on the historical fault air energy equipment combination; A Bayesian network model is constructed based on the device association graph and the air energy device fault history data; wherein the Bayesian network model uses the air energy device status as a node, the association relationship between the air energy devices as an edge, and the edge weight as a conditional probability; Taking the reference air energy device as the starting point of the faulty air energy device, the propagation probability of the fault between the air energy devices and the reference propagation path are determined through Bayesian network reasoning; The probabilities of different propagation paths are sorted, and the fault propagation path and the associated reference air energy device are determined based on the sorting result.

9. An air energy equipment detection device based on vibration signals, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 8.

Images