Method and device for determining mechanical equipment failure

Abstract

The present invention discloses a method and device for determining mechanical equipment faults. The invention includes: obtaining vibration data corresponding to a target test position of the equipment and a preset neural network model; inputting the vibration data into the preset neural network model and obtaining an output result of the preset neural network model; determining a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; in the case where the target label is a fault label, determining that a fault occurs at the target test position, and conversely, determining that no fault occurs at the target test position. Through the present invention, the technical problem in the related art of lacking a detection means for high-precision mechanical equipment faults is solved.

Description

Technical Field

[0001] The present invention relates to the field of mechanical equipment detection, and more particularly, to a method and device for determining mechanical equipment faults. Background Art

[0002] In related technologies, the detection of mechanical equipment status and fault diagnosis are currently popular research directions. With the accumulation and storage of various types of data such as mechanical equipment operation data, device fault prediction methods based on big data are emerging in an endless stream, and various neural network models have emerged. Most of the research is based on laboratory environments and cannot simulate and construct a real on-site working condition environment affected by multiple factors, resulting in difficulty in industrial implementation of the model. Moreover, fault feature extraction is a difficult point, which directly affects the accuracy of device fault prediction.

[0003] Vibration signals, as commonly used fault status response data for reciprocating or rotating mechanical equipment, are widely used in the industry. Ideally, vibration signals can respond in a timely manner when a fault occurs. However, due to various conditions such as on-site noise interference and multi-condition operation on-site, the fault response of vibration signals is not obvious, or it is difficult to extract fault features. Existing research mostly extracts fault features through various methods such as time-domain statistical index extraction, frequency-domain statistical index extraction, time-frequency domain joint distribution matrix extraction, and signal decomposition based on vibration signals, and combines various neural networks for fault prediction. However, the general prediction accuracy is not high, and the fault features are only reflected in a certain dimension, and it is impossible to comprehensively and timely capture the short-term instantaneous features and long-term decay degradation features of faults.

[0004] In view of the above problems in related technologies, no effective solution has been proposed yet. Summary of the Invention

[0005] The main object of the present invention is to provide a method and device for determining mechanical equipment faults, so as to solve the technical problem of the lack of high-precision detection means for mechanical equipment faults in related technologies.

[0006] To achieve the above object, according to one aspect of the present invention, a method for determining mechanical equipment faults is provided. The method includes: obtaining vibration data corresponding to a target test position of the device and a preset neural network model; inputting the vibration data into the preset neural network model and obtaining an output result of the preset neural network model; determining a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; in the case where the target label is a fault label, determining that a fault occurs at the target test position, and conversely, determining that no fault occurs at the target test position.

[0007] Further, before obtaining the preset neural network model, the method further includes: constructing an initial preset neural network model, where the initial preset neural network model includes multiple convolutional layers, multiple pooling layers, and multiple fully connected layers, and each neural network layer is connected by a ReLU activation function, and a LogSoftmax function is connected to the fully connected layer of the initial preset neural network model; determining a training data set for training the initial preset neural network model, and training the initial preset neural network model with the training data set to obtain the preset neural network model.

[0008] Further, determining a training data set for training the initial preset neural network model includes: obtaining source data, where the source data is data collected by vibration sensors arranged on mechanical equipment, and the source data is any one of the following types: fault type, non-fault type; based on the source data, constructing a 3D time-frequency joint distribution cube including preset dimensions, where the preset dimensions at least include the following dimensions: long time domain dimension, short time domain dimension, and frequency domain dimension. Performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices; dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model.

[0009] Further, before dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model, the method further includes: determining the label type corresponding to the LST-FD joint distribution matrix according to the type of source data corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: fault label, non-fault label.

[0010] Further, constructing multiple 3D time-frequency joint distribution cubes including preset dimensions based on the source data includes: determining a first time granularity, and performing data segmentation on the source data with the first time granularity to obtain a fault sample data set and a non-fault sample data set, where the fault sample data set includes multiple fault data subsets, and the non-fault sample data set includes multiple non-fault data subsets, and the first time granularity at least includes the total time corresponding to multiple rotation cycles of the mechanical equipment; determining a second time granularity; dividing each fault data subset according to the second time granularity to obtain multiple first short time cycle data; dividing each non-fault data subset according to the second time granularity to obtain multiple second short time cycle data; processing the multiple first short time cycle data corresponding to the fault sample data set and the multiple second short time cycle data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube.

[0011] Further, process multiple first short-time period data corresponding to the fault sample data set and multiple second short-time period data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube, including: performing STFT transform processing on the multiple first short-time period data to obtain multiple first time-domain joint distribution matrices; performing STFT transform processing on the multiple second short-time period data to obtain multiple second time-domain joint distribution matrices; stacking the multiple first time-domain joint distribution matrices and the multiple second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and the preset order; constructing a 3D time-frequency joint distribution cube based on the time-domain joint distribution matrix set.

[0012] Further, perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain an LST-FD joint distribution matrix set, including: based on the long time domain dimension, perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube through formula one to obtain an LST-FD joint distribution matrix, where formula one is:

[0013] where z(k, t, f re ) is the time-frequency joint distribution matrix under the long time dimension of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, and F re _d is the frequency time-domain average difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension.

[0014] Further, the preset neural network model is a 2D-CNN neural network model.

[0015] To achieve the above object, according to another aspect of the present invention, a device for determining mechanical equipment faults is provided. The device includes: a first acquisition unit, configured to acquire vibration data corresponding to a target test position of the device and a preset neural network model; a first input unit, configured to input the vibration data into the preset neural network model and acquire the output result of the preset neural network model; a first determination unit, configured to determine a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; a second determination unit, configured to determine that a fault occurs at the target test position when the target label is a fault label, and vice versa, determine that no fault occurs at the target test position.

[0016] To achieve the above object, according to another aspect of the present invention, there is provided a computer-readable storage medium, which includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute the method for determining a mechanical equipment fault according to any one of claims 1 to 7.

[0017] To achieve the above object, according to another aspect of the present invention, there is provided a processor, characterized in that the processor is used to run a program. When the program runs, it executes the method for determining a mechanical equipment fault according to any one of claims 1 to 7.

[0018] By the present invention, the following steps are adopted: obtaining vibration data corresponding to a target test position of a device and a preset neural network model; inputting the vibration data into the preset neural network model and obtaining an output result of the preset neural network model; determining a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; in the case where the target label is a fault label, determining that a fault occurs at the target test position, and conversely, determining that no fault occurs at the target test position. This solves the technical problem in the related art of lacking a high-precision detection means for mechanical equipment faults, and further achieves the technical effect of capturing and reflecting the response of vibration signals to faults from multiple dimensions, being able to capture the instantaneous characteristics of faults and also reflect the long-term degradation trend characteristics of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0020] Figure 1 is a flowchart of a method for determining a mechanical equipment fault according to an embodiment of the present invention; and

[0021] Figure 2 is a flowchart corresponding to another method for determining a mechanical equipment fault provided by the present application;

[0022] Figure 3 is a schematic diagram of a device for determining a mechanical equipment fault according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in combination with the embodiments.

[0024] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to implement the embodiments of the present invention described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0026] According to an embodiment of the present invention, a method for determining a mechanical equipment failure is provided.

[0027] Figure 1 is a flowchart of a method for determining a mechanical equipment failure provided according to an embodiment of the present invention. As Figure 1 shown, the invention includes the following steps:

[0028] Step S101, obtain vibration data corresponding to a target test position of the device, and a preset neural network model;

[0029] Step S102, input the vibration data into the preset neural network model, and obtain the output result of the preset neural network model;

[0030] Step S103, determine the target label included in the output result, where the target label is any one of the following: a failure label, a non-failure label;

[0031] Step S104, when the target label is a failure label, determine that the target test position has a failure; otherwise, determine that the target test position has no failure.

[0032] As described above, the present application provides a method for determining a mechanical equipment failure. By inputting the vibration data of the part to be tested of the device into a preset neural network model, it is determined whether the part to be tested has a failure based on the label output by the neural network.

[0033] Through the above method, the technical problem of the lack of detection means for high-precision mechanical equipment failures in the related art is solved, and thus the technical effect of capturing and reflecting the response of vibration signals to failures from multiple dimensions is achieved, that is, the instantaneous characteristics of failures can be captured and the long-term deterioration trend characteristics of failures can also be reflected.

[0034] Specifically, the vibration signal acquisition of the part to be tested in this application is mainly achieved by installing vibration sensors at the target detection positions of mechanical equipment, and collecting and transmitting vibration data for storage through data acquisition software.

[0035] For example: during the process of predicting the bearing failure at the input end of the reducer of the on-vehicle plunger pump of a fracturing truck, a single-direction acceleration vibration sensor is installed in the horizontal direction on the input side of the reducer (near the position of the input-side bearing) to collect vibration data. This sensor is marked as AI1-32, the sampling frequency of the sensor is 51.2KHZ, and the rotational speed of the motor at the power end of the reducer is known.

[0036] In an optional embodiment, in a method for determining mechanical equipment failures provided in an embodiment of the present invention, before obtaining a preset neural network model, the method further includes: constructing an initial preset neural network model, where the initial preset neural network model includes multiple convolutional layers, multiple pooling layers, and multiple fully connected layers, and each neural network layer is connected by a ReLU activation function, and a LogSoftmax function is connected to the fully connected layer connected to the initial preset neural network model; determining a training data set for training the initial preset neural network model, and training the initial preset neural network model through the training data set to obtain a preset neural network model.

[0037] In an optional embodiment, the preset neural network model is a 2D-CNN neural network model. In the above method, first, a 2D-CNN neural network model needs to be built. The network structure of the 2D-CNN neural network model includes a total of 7 layers, including 3 convolutional layers, 2 pooling layers, and 2 fully connected layers. Neurons are connected by a ReLU activation function, and finally, the fully connected layer is connected to a LogSoftmax function to output the model prediction result.

[0038] It should be noted that, in an optional embodiment, the loss function of this network model selects the CrossEntropyLoss function to calculate the error between the input data and the prediction result. At the same time, the optimizer corresponding to this model selects the Adam function to optimize the connection weights of the model neurons.

[0039] As described above, this application provides a method for determining the training data set of a preset neural network model, which specifically includes the following steps:

[0040] S201: Obtain source data, where the source data is the data collected by vibration sensors set on mechanical equipment, and the source data is any one of the following types: fault type, non-fault type;

[0041] S202: Based on the source data, construct a 3D time-frequency joint distribution cube including preset dimensions, where the preset dimensions at least include the following dimensions: long time domain dimension, short time domain dimension, and frequency domain dimension;

[0042] S203: Perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices;

[0043] S204: Divide the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing a preset neural network model.

[0044] Specifically, the present application provides a mechanical equipment fault prediction method based on an LST-FD (Long short time-Frequency difference) matrix. The LST-FD matrix starts from two time granularity dimensions of short time (instantaneous) and long time, constructs a 3D time-frequency joint distribution cube under the long and short double time dimensions, and calculates the time domain average difference of the frequency signal under the long time dimension to perform dimensionality reduction feature extraction to obtain the LST-FD matrix, which is used as the feature matrix for fault classification prediction of the preset neural network model to perform equipment fault classification prediction. The LST-FD matrix adds another long time domain dimension based on the time-frequency joint distribution matrix, and performs feature dimensionality reduction extraction under this time domain dimension. On the one hand, it amplifies the time-frequency fluctuation characteristics when the fault signal occurs, and on the other hand, it can simultaneously capture two types of fault signals: the short-time instantaneous feature response and the long-term decay feature wave - motion of the equipment at the initial stage of fault occurrence. Compared with the traditional time-frequency diagram or other feature indicators as the model input, the prediction accuracy is relatively more accurate.

[0045] In an optional embodiment, in a method for determining mechanical equipment failures provided by an embodiment of the present invention, based on source data, a plurality of 3D time-frequency joint distribution cubes including preset dimensions are constructed, including: determining a first time granularity, and performing data segmentation on the source data through the first time granularity to obtain a failure sample data set and a non-failure sample data set, where the failure sample data set contains a plurality of failure data subsets, the non-failure sample data set contains a plurality of non-failure data subsets, and the first time granularity includes at least the total time corresponding to a plurality of rotation integral periods of the mechanical equipment; determining a second time granularity; according to the second time granularity, dividing each failure data subset to obtain a plurality of first short-time period data; according to the second time granularity, dividing each non-failure data subset to obtain a plurality of second short-time period data; processing the plurality of first short-time period data corresponding to the failure sample data set and the plurality of second short-time period data corresponding to the non-failure sample data set to obtain a 3D time-frequency joint distribution cube.

[0046] Specifically, the embodiments of the present application provide two time granularities, including a first time granularity and a second time granularity.

[0047] In a specific embodiment, according to the rotational speed of the equipment power end, the source data is segmented into sample data according to the time granularity t1 (the first time granularity) (the length of the time granularity t1 includes at least the time lengths of 10 rotation integral periods of the equipment), so as to obtain the segmented failure data sample data set X_ fault ={x_ (f1,) x_ (f2,) …,x_ fi} and the normal data sample data set X_ normal ={x_ (n1,) x_ (n2,) …,x_ ni}, where x_ (f1,) etc. are failure data subsets, and x_ (n1,) is a non-failure data subset.

[0048] Further, determine the second time granularity t1. Preferably, the second time granularity is determined according to the equipment rotation period, generally covering the time length of 1 rotation integral period of the equipment.

[0049] According to the second time granularity, the above-mentioned respective failure data subsets and non-failure data subsets are respectively divided into data to obtain a plurality of first short-time period data and second short-time period data. For example: x_ (f2,) the failure data subset, according to the time granularity t2, is divided into x_ (f2,) ={x_ (t11,) x_ (t12,) …,x_t1k}, where \(k = t2 / t1\), and the above processing is performed on all other sample data.

[0050] In an alternative embodiment, in a method for determining mechanical equipment faults provided by an embodiment of the present invention, multiple first short-time period data corresponding to a fault sample data set and multiple second short-time period data corresponding to a non-fault sample data set are processed to obtain a 3D time-frequency joint distribution cube, including: performing STFT transformation on the multiple first short-time period data to obtain multiple first time-domain joint distribution matrices; performing STFT transformation on the multiple second short-time period data to obtain multiple second time-domain joint distribution matrices; stacking the multiple first time-domain joint distribution matrices and the multiple second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and a preset order; constructing a 3D time-frequency joint distribution cube based on the time-domain joint distribution matrix set.

[0051] Further, perform STFT transformation on the multiple first short-time period data and the multiple second short-time period data obtained by dividing through the second time granularity. Among them, the window function selects a hann window, the window function length is 256, and the window function overlap is 50%. After STFT transformation, the time-frequency joint distribution matrix \(z\) is obtained (t,fre) , where \(t\) is the time length, \(f\) re is the frequency range, and the median value of \(z\) is the frequency amplitude. Example: \(x\_\) (f2) = \(\{x\_\) (t11,) \(x\_\) (t12,) …, \(x\_\) t1k \}\), after STFT transformation, \(k\) time-frequency joint distribution matrices \(\{z\_\) (x_t11,) \(z\_\) (x_t12,) \(z\_\) (x_t13,) …, \(z\_\) x_t1k \}\) are obtained. That is, each short-time period data in \(X\_\) (fault) = \(\{x\_\) (f1,) \(x\_\) (f2,) …, \(x\_\) fi \}\) and \(X\_\) (normal) = \(\{x\_\) (n1,) \(x\_\) (n2,) …, \(x\_\) ni \}\) generates \(k\) time-frequency joint distribution matrices.

[0052] Further, \(X\_\) (fault) = \(\{x\_\) (f1,) \(x\_\) (f2,) …, \(x\_\) fi \}\) and \(X\_\) (normal) = \(\{x\_\) (n1,) \(x\_\) (n2,) …, \(x\_\) niIn the two datasets, each dataset contains k time-frequency joint distribution matrices. With k (i.e., the time length of t2) as the third dimension, stack the k time-frequency joint distribution matrices in the order of t2 time sequence to generate a 3D time-frequency joint distribution cube construction. That is, each sample data in the above two datasets corresponds to a 3D time-frequency joint distribution cube (k, t, fre). In this 3D time-frequency joint distribution cube, the first dimension is the long time domain dimension of the vibration signal (including at least 10 complete rotation cycles of the device), the second dimension is the short time domain dimension of the vibration signal (covering one rotation cycle of the device), and the third dimension is the frequency domain dimension of the vibration signal.

[0053] In an optional embodiment, in a method for determining mechanical equipment faults provided by an embodiment of the present invention, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices includes: based on the long time domain dimension, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube through Formula 1 to obtain an LST-FD joint distribution matrix, where Formula 1 is:

[0054] where z (k，t，fre) is the time-frequency joint distribution matrix under the long time dimension (t1 time dimension) of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, and Fre_ d is the average frequency time domain difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension.

[0055] Furthermore, perform dimensionality reduction feature extraction on the constructed 3D time-frequency joint distribution cube based on the first dimension, that is, the long time domain dimension of the vibration signal. That is, calculate the average frequency time domain difference Fre_d in the order of time along the first dimension for the amplitudes corresponding to the frequency values at different short time points on the second and third dimensions.

[0056] Perform dimensionality reduction feature extraction on each 3D time-frequency joint distribution cube according to the above formula to obtain each LST-FD joint distribution matrix. The shape of this matrix is (t, Fre_d), where t is the short time domain dimension of the vibration signal of the original 3D time-frequency joint distribution cube, and Fre_d is the average frequency time domain difference obtained from the frequency amplitudes of the original 3D time-frequency joint distribution cube based on the first dimension.

[0057] In an optional embodiment, in a method for determining a mechanical equipment fault provided by an embodiment of the present invention, before dividing a plurality of LST-FD joint distribution matrices included in the LST-FD joint distribution matrix set to obtain a training data set and a test data set for testing a preset neural network model, the method further includes: determining a label type corresponding to the LST-FD joint distribution matrix according to the source data type corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: fault label, non-fault label.

[0058] Specifically, after the above processing of the source data, the source data set X fault ={x f1 x f2 ,..., x fi} and X normal ={x n1 , x n2 ,..., x ni} in each sample data (each data subset) is processed into a corresponding LST-FD joint distribution matrix set. And corresponding label data sets of labels 1 and 0 are respectively generated for the LST-FD joint distribution matrices generated by non-fault data and fault data. Among them, 1 represents a device state fault, and 0 represents a normal device state. The LST-FD joint distribution matrix sets corresponding to normal data and fault data and their respective label data are divided according to a ratio of 1:9 to generate a test set and a training set.

[0059] Further, after generating the training set and the test set, the training data set is input into the above neural network model. After the data is calculated in each layer of the network, it is output to the next layer of the neural network layer through the ReLU activation function. The last layer of the LogSoftmax function outputs the calculation result. The calculation result and the real data are input into the CrossEntropyLoss loss function. The loss function calculates the loss value. When the loss value is greater than the set threshold β, the optimization function Adam updates the connection weights of each layer of the network in the gradient direction towards the direction of reducing the loss value according to the backward propagation value of the loss value; when the loss function value is less than the set threshold β, the neural network training ends, and the network structure and information of each level of neurons are saved;

[0060] The test data set is input into the trained neural network model, and the test result and the test accuracy metric AUC value are output. If the AUC value is less than the set threshold α, the model test is completed. If the AUC value is greater than the threshold α, the sample data is resampled and the model training and test are performed again.

[0061] In a specific embodiment provided by the present application, the experimental data selects the horizontal AI1-32 sensor data on the input side of the reduction gearbox of the fracturing truck on-board plunger pump, and the sensor sampling frequency is 51.2 kHz. A total of 30 hours of AI1-32 sensor data during normal operation of the reduction gearbox and 18 hours of reduction gearbox bearing roller fault data are obtained. According to the rotation cycle time length of the equipment crankshaft, the time granularity t2 = 1 s and t1 = 10 s are set respectively.

[0062] Training process:

[0063] And according to steps 1 to 5, 10,800 normal data LST-FD joint distribution matrices and the same number of normal data labels are generated, 6,480 fault data LST-FD joint distribution matrices and the same number of fault data labels are generated. The samples are divided according to a ratio of 1:9. A total of 9,720 normal training data and labels, 1,080 normal test data and labels, 5,832 fault training data and labels, and 648 fault test data and labels are obtained.

[0064] Based on the above training data and test data, model training and model testing are carried out. The relevant parameters are as follows:

[0065] Epoch (Number of model training times) 100 Learning rate 0.0001 Loss value threshold β 0.01 AUC value threshold α 0.90

[0066] The model prediction results and the comparison of the same model directly using the time-frequency diagram or the statistical indicators of the time-domain signal as features for fault classification prediction are as follows:

[0067]

[0068] Therefore, it can be seen from the above specific embodiments that the 2D_CNN model based on LST-FD has the highest prediction accuracy and the largest AUC value. The model prediction accuracy based on the statistical indicators of the time-domain signal is relatively low, and the AUC value just exceeds the set threshold.

[0069] Therefore, a method for determining mechanical equipment faults provided by the present application has the following advantages:

[0070] 1: Innovatively construct the LST-FD joint distribution matrix, and combine it with the preset neural network model to achieve equipment fault prediction with relatively high prediction accuracy;

[0071] 2: Considering from the perspective of the characteristics of equipment fault occurrence, a 3D joint distribution cube including three dimensions of long time-domain dimension, short time-domain dimension, and frequency-domain dimension is constructed. This cube can capture the short-time instantaneous spectrum fault characteristics of the fault and can also capture the long-time degradation trend characteristics of the fault, and can detect incipient faults earlier and monitor the fault degradation trend for a long time;

[0072] 3: Perform dimensionality reduction and principal feature extraction on the 3D joint distribution cube in the long time domain dimension to generate the LST-FD joint distribution matrix, which not only retains the original three-dimensional data features but also reduces the data dimension, improving the model prediction accuracy and operation efficiency.

[0073] This application also provides another method for determining mechanical equipment failures, as Figure 2 shown. It should be noted that Figure 2 in which t2 is the total duration corresponding to multiple complete rotation cycles of the mechanical equipment, and t1 is the time length of one complete rotation cycle of the equipment. Through the Figure 1 method provided, it also solves the technical problem in the related art of lacking high-precision detection means for mechanical equipment failures, and further achieves the technical effect of capturing and reflecting the response of vibration signals to failures from multiple dimensions, being able to capture both the instantaneous fault characteristics and the long-term degradation trend characteristics of faults.

[0074] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0075] A method for determining mechanical equipment failures provided by an embodiment of the present invention includes obtaining vibration data corresponding to a target test position of the equipment and a preset neural network model; inputting the vibration data into the preset neural network model and obtaining the output result of the preset neural network model; determining the target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; in the case where the target label is a fault label, it is determined that a fault occurs at the target test position, and conversely, it is determined that no fault occurs at the target test position, solving the technical problem in the related art of lacking high-precision detection means for mechanical equipment failures. Furthermore, it achieves the technical effect of capturing and reflecting the response of vibration signals to failures from multiple dimensions, being able to capture both the instantaneous fault characteristics and the long-term degradation trend characteristics of faults.

[0076] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0077] An embodiment of the present invention also provides a device for determining mechanical equipment failures. It should be noted that the device for determining mechanical equipment failures in an embodiment of the present invention can be used to execute the method for determining mechanical equipment failures provided by the embodiment of the present invention. The following introduces the device for determining mechanical equipment failures provided by the embodiment of the present invention.

[0078] Figure 3 It is a schematic diagram of a device for determining mechanical equipment failures according to an embodiment of the present invention. As Figure 3 shown, the device includes: an acquisition unit 301, configured to acquire vibration data corresponding to a target test position of the device, and a preset neural network model; an input unit 302, configured to input the vibration data into the preset neural network model and obtain an output result of the preset neural network model; a first determination unit 303, configured to determine a target label included in the output result, where the target label is any one of the following: a failure label, a non-failure label; a second determination unit 304, configured to determine that a failure occurs at the target test position when the target label is a failure label, and conversely, determine that no failure occurs at the target test position.

[0079] In an optional embodiment, a construction unit is configured to construct an initial preset neural network model before acquiring the preset neural network model, where the initial preset neural network model includes multiple convolutional layers, multiple pooling layers, and multiple fully connected layers, and each neural network layer is connected by a ReLU activation function, and a LogSoftmax function is connected to the fully connected layer of the initial preset neural network model; a third determination unit is configured to determine a training data set for training the initial preset neural network model, and train the initial preset neural network model with the training data set to obtain the preset neural network model.

[0080] In an optional embodiment, the third determination unit includes: an acquisition subunit, configured to acquire source data, where the source data is data collected by vibration sensors arranged on the mechanical equipment, and the source data is any one of the following types: a failure type, a non-failure type; a construction subunit, configured to construct a 3D time-frequency joint distribution cube including a preset dimension based on the source data, where the preset dimension includes at least the following dimensions: a long time domain dimension, a short time domain dimension, and a frequency domain dimension. An extraction subunit is configured to perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices; a partitioning subunit is configured to partition multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model.

[0081] In an optional embodiment, the partitioning subunit includes: a first determination module, configured to determine a label type corresponding to the LST-FD joint distribution matrix according to the source data type corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: a failure label, a non-failure label.

[0082] In an alternative embodiment, the construction subunit includes: a second determination module, configured to determine a first time granularity, and perform data segmentation on source data through the first time granularity to obtain a fault sample data set and a non-fault sample data set, where the fault sample data set contains multiple fault data subsets, and the non-fault sample data set contains multiple non-fault data subsets, and the first time granularity at least includes the total time corresponding to multiple rotation integral periods of the mechanical equipment; a third determination module, configured to determine a second time granularity; a first segmentation module, configured to segment each fault data subset according to the second time granularity to obtain multiple first short-time period data; a second segmentation module, configured to segment each non-fault data subset according to the second time granularity to obtain multiple second short-time period data; and a processing module, configured to process the multiple first short-time period data corresponding to the fault sample data set and the multiple second short-time period data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube.

[0083] In an alternative embodiment, the processing module includes: a first processing sub-module, configured to perform STFT transform processing on multiple first short-time period data to obtain multiple first time-domain joint distribution matrices; a second processing sub-module, configured to perform STFT transform processing on multiple second short-time period data to obtain multiple second time-domain joint distribution matrices; a stacking sub-module, configured to stack the multiple first time-domain joint distribution matrices and the multiple second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and a preset order; and a construction sub-module, configured to construct a 3D time-frequency joint distribution cube according to the time-domain joint distribution matrix set.

[0084] In an alternative embodiment, the extraction subunit includes: an extraction module, configured to perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube based on a long time domain dimension through Formula 1 to obtain an LST-FD joint distribution matrix, where Formula 1 is:

[0085] where z(k, t, fre) is the time-frequency joint distribution matrix in the long time domain dimension of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, and Fre_d is the frequency time-domain average difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension.

[0086] A device for determining mechanical equipment faults provided by an embodiment of the present invention includes an acquisition unit 301, configured to acquire vibration data corresponding to a target test position of the equipment and a preset neural network model; an input unit 302, configured to input the vibration data into the preset neural network model and acquire an output result of the preset neural network model; a first determination unit 303, configured to determine a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; a second determination unit 304, configured to, when the target label is a fault label, determine that a fault occurs at the target test position, and vice versa, determine that no fault occurs at the target test position.

[0087] A device for determining mechanical equipment faults includes a processor and a memory. The above-mentioned acquisition unit 201, etc. are all stored in the memory as program units, and the corresponding functions are implemented by the processor executing the above program units stored in the memory.

[0088] The processor includes a kernel, and the kernel is used to retrieve the corresponding program unit from the memory. One or more kernels can be set, and by adjusting the kernel parameters, the technical problem of lacking a high-precision detection means for mechanical equipment faults in the related art is solved.

[0089] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.

[0090] An embodiment of the present invention provides a storage medium, on which a program is stored, and when the program is executed by a processor, a method for determining mechanical equipment faults is implemented.

[0091] An embodiment of the present invention provides a processor, and the processor is used to run a program. When the program runs, a method for determining mechanical equipment faults is executed.

[0092] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored on the memory and executable on the processor. When the processor executes the program, the following steps are implemented: acquiring vibration data corresponding to a target test position of the equipment and a preset neural network model; inputting the vibration data into the preset neural network model and acquiring an output result of the preset neural network model; determining a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; when the target label is a fault label, determining that a fault occurs at the target test position, and vice versa, determining that no fault occurs at the target test position.

[0093] In an alternative embodiment, before obtaining the preset neural network model, the method further includes: constructing an initial preset neural network model, where the initial preset neural network model includes multiple convolutional layers, multiple pooling layers, and multiple fully connected layers, and each neural network layer is connected by a ReLU activation function, and a LogSoftmax function is connected to the fully connected layer of the initial preset neural network model; determining a training data set for training the initial preset neural network model, and training the initial preset neural network model with the training data set to obtain the preset neural network model.

[0094] In an alternative embodiment, determining a training data set for training the initial preset neural network model includes: obtaining source data, where the source data is data collected by vibration sensors arranged on mechanical equipment, and the source data is any one of the following types: fault type, non-fault type; based on the source data, constructing a 3D time-frequency joint distribution cube including preset dimensions, where the preset dimensions at least include the following dimensions: long time domain dimension, short time domain dimension, and frequency domain dimension. Performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices; dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model.

[0095] In an alternative embodiment, before dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model, the method further includes: determining the label type corresponding to the LST-FD joint distribution matrix according to the source data type corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: fault label, non-fault label.

[0096] In an alternative embodiment, based on the source data, multiple 3D time-frequency joint distribution cubes including preset dimensions are constructed, including: determining a first time granularity, and performing data segmentation on the source data through the first time granularity to obtain a fault sample data set and a non-fault sample data set, where the fault sample data set contains multiple fault data subsets, and the non-fault sample data set contains multiple non-fault data subsets, and the first time granularity at least includes the total time corresponding to multiple rotation cycles of the mechanical equipment; determining a second time granularity; according to the second time granularity, each fault data subset is segmented to obtain multiple first short-time cycle data; according to the second time granularity, each non-fault data subset is segmented to obtain multiple second short-time cycle data; processing the multiple first short-time cycle data corresponding to the fault sample data set and the multiple second short-time cycle data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube.

[0097] In an alternative embodiment, processing the multiple first short-time cycle data corresponding to the fault sample data set and the multiple second short-time cycle data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube includes: performing STFT transform processing on the multiple first short-time cycle data to obtain multiple first time-domain joint distribution matrices; performing STFT transform processing on the multiple second short-time cycle data to obtain multiple second time-domain joint distribution matrices; stacking the multiple first time-domain joint distribution matrices and the multiple second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and a preset order; constructing a 3D time-frequency joint distribution cube based on the time-domain joint distribution matrix set.

[0098] In an alternative embodiment, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices includes: performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube based on the long time domain dimension through formula one to obtain an LST-FD joint distribution matrix, where formula one is:

[0099] where z(k, t, fre) is the time-frequency joint distribution matrix in the long time domain dimension of the kth layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, and Fre_d is the frequency time domain average difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension.

[0100] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0101] The present invention also provides a computer program product, which when executed on a data processing device, is adapted to execute a program initialized with the following method steps: obtaining vibration data corresponding to a target test location of the device, and a preset neural network model; inputting the vibration data into the preset neural network model, and obtaining an output result of the preset neural network model; determining a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; in the case where the target label is a fault label, determining that the target test location has a fault, and conversely, determining that the target test location has no fault.

[0102] In an alternative embodiment, before obtaining the preset neural network model, the method further includes: constructing an initial preset neural network model, where the initial preset neural network model includes multiple convolutional layers, multiple pooling layers, and multiple fully connected layers, and each neural network layer is connected by a ReLU activation function, and a LogSoftmax function is connected to the fully connected layer connected in the initial preset neural network model; determining a training data set for training the initial preset neural network model, and training the initial preset neural network model with the training data set to obtain the preset neural network model.

[0103] In an alternative embodiment, determining the training data set for training the initial preset neural network model includes: obtaining source data, where the source data is data collected by vibration sensors provided on a mechanical device, and the source data is any one of the following types: a fault type, a non-fault type; based on the source data, constructing a 3D time-frequency joint distribution cube including a preset dimension, where the preset dimension at least includes the following dimensions: a long time domain dimension, a short time domain dimension, and a frequency domain dimension. Performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices; dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model.

[0104] In an alternative embodiment, before dividing the multiple LST-FD joint distribution matrices included in the set of LST-FD joint distribution matrices to obtain a training data set and a test data set for testing the preset neural network model, the method further includes: determining a label type corresponding to the LST-FD joint distribution matrix according to the type of source data corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: a fault label, a non-fault label.

[0105] In an alternative embodiment, based on the source data, a plurality of 3D time-frequency joint distribution cubes including preset dimensions are constructed, comprising: determining a first time granularity, and performing data segmentation on the source data through the first time granularity to obtain a fault sample data set and a non-fault sample data set, wherein the fault sample data set contains a plurality of fault data subsets, and the non-fault sample data set contains a plurality of non-fault data subsets, and the first time granularity includes at least the total time corresponding to a plurality of rotation integral periods of the mechanical equipment; determining a second time granularity; according to the second time granularity, dividing each fault data subset to obtain a plurality of first short-time period data; according to the second time granularity, dividing each non-fault data subset to obtain a plurality of second short-time period data; processing the plurality of first short-time period data corresponding to the fault sample data set and the plurality of second short-time period data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube.

[0106] In an alternative embodiment, processing the plurality of first short-time period data corresponding to the fault sample data set and the plurality of second short-time period data corresponding to the non-fault sample data set to obtain a 3D time-frequency joint distribution cube includes: performing STFT transform processing on the plurality of first short-time period data to obtain a plurality of first time-domain joint distribution matrices; performing STFT transform processing on the plurality of second short-time period data to obtain a plurality of second time-domain joint distribution matrices; stacking the plurality of first time-domain joint distribution matrices and the plurality of second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and a preset order; constructing a 3D time-frequency joint distribution cube based on the time-domain joint distribution matrix set.

[0107] In an alternative embodiment, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube to obtain a set of LST-FD joint distribution matrices includes: based on the long time domain dimension, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube through formula one to obtain an LST-FD joint distribution matrix, where formula one is:

[0108] where z(k, t, fre) is the time-frequency joint distribution matrix under the long time dimension of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, and Fre_d is the frequency time domain average difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension.

[0109] Those skilled in the art will understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0110] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0111] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0113] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0114] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory. The memory is an example of computer-readable media.

[0115] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0116] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0117] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] The above are merely embodiments of the present invention and are not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A method for determining a mechanical equipment failure, characterized in that, Including: Obtain the vibration data corresponding to the target test position of the device, and a preset neural network model; Input the vibration data into the preset neural network model, and obtain the output result of the preset neural network model; Determine the target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; In the case where the target label is the fault label, determine that the target test position has a fault, otherwise, determine that the target test position does not have the fault; Before obtaining the preset neural network model, the method further includes: Construct an initial preset neural network model; Construct a 3D time-frequency joint distribution cube including preset dimensions based on the source data, where the preset dimensions at least include the following dimensions: a long time domain dimension, a short time domain dimension, and a frequency domain dimension; Based on the long time domain dimension, perform dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube through Formula 1 to obtain an LST-FD joint distribution matrix, and Formula 1 is: ， z(k, t, f re ) is the time-frequency joint distribution matrix in the long time dimension of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, F re _d is the frequency-time domain average difference obtained by frequency replication calculation of the 3D time-frequency joint distribution cube based on the long time domain dimension; Divide the multiple LST-FD joint distribution matrices included in the LST-FD joint distribution matrix set to obtain a training data set and a test data set for testing the preset neural network model, and train the initial preset neural network model through the training data set to obtain the preset neural network model.

2. The method according to claim 1, wherein Before dividing the multiple LST-FD joint distribution matrices included in the LST-FD joint distribution matrix set to obtain the training data set and the test data set for testing the preset neural network model, the method further includes: Determine the label type corresponding to the LST-FD joint distribution matrix according to the source data type corresponding to the LST-FD joint distribution matrix, where the label type is any one of the following: a fault label, a non-fault label.

3. The method according to claim 1, wherein Construct multiple 3D time-frequency joint distribution cubes including preset dimensions based on the source data, including: Determine a first time granularity, and perform data segmentation on the source data through the first time granularity to obtain a fault sample data set and a non-fault sample data set, where the fault sample data set includes multiple fault data subsets, the non-fault sample data set includes multiple non-fault data subsets, and the first time granularity at least includes the total time corresponding to multiple rotation integral periods of the mechanical equipment; Determine a second time granularity; Divide each of the fault data subsets according to the second time granularity to obtain multiple first short time period data; Divide each of the non-fault data subsets according to the second time granularity to obtain multiple second short time period data; Process the multiple first short time period data corresponding to the fault sample data set and the multiple second short time period data corresponding to the non-fault sample data set to obtain the 3D time-frequency joint distribution cube.

4. The method according to claim 3, wherein Processing the multiple pieces of the first short-time period data corresponding to the faulty sample data set and the multiple pieces of the second short-time period data corresponding to the non-faulty sample data set to obtain the 3D time-frequency joint distribution cube, including: Performing STFT transform processing on the multiple pieces of the first short-time period data to obtain multiple first time-domain joint distribution matrices; Performing STFT transform processing on the multiple pieces of the second short-time period data to obtain multiple second time-domain joint distribution matrices; Stacking the multiple first time-domain joint distribution matrices and the multiple second time-domain joint distribution matrices into a time-domain joint distribution matrix set according to the first time granularity and a preset order; Constructing the 3D time-frequency joint distribution cube according to the time-domain joint distribution matrix set.

5. The method according to any one of claims 1 to 4, characterized in that The preset neural network model is a 2D-CNN neural network model.

6. A device for determining mechanical equipment failures, characterized in that, Including: An acquisition unit, configured to acquire vibration data corresponding to a target test position of a device and a preset neural network model; An input unit, configured to input the vibration data into the preset neural network model and acquire an output result of the preset neural network model; A first determination unit, configured to determine a target label included in the output result, where the target label is any one of the following: a fault label, a non-fault label; A second determination unit, configured to determine that the target test position has a fault when the target label is the fault label, and vice versa, determine that the target test position does not have the fault; A construction unit, configured to construct an initial preset neural network model before acquiring the preset neural network model; A third determination unit is configured to construct a 3D time-frequency joint distribution cube including a preset dimension according to source data, where the preset dimension at least includes the following dimensions: a long time-domain dimension, a short time-domain dimension, and a frequency-domain dimension; based on the long time-domain dimension, performing dimensionality reduction feature extraction on the 3D time-frequency joint distribution cube through Formula 1 to obtain an LST-FD joint distribution matrix, and the Formula 1 is: ， z(k, t, f re ) is the time-frequency joint distribution matrix in the long time dimension of the k-th layer in the 3D time-frequency joint distribution cube, t is the short time domain dimension corresponding to the 3D time-frequency joint distribution cube, F re _d is the frequency-time domain average difference obtained by calculating frequency replication of the 3D time-frequency joint distribution cube based on the long time domain dimension; Dividing the multiple LST-FD joint distribution matrices included in the LST-FD joint distribution matrix set to obtain a training data set and a test data set for testing the preset neural network model, and training the initial preset neural network model through the training data set to obtain the preset neural network model.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, where, when the program runs, it controls the device where the computer-readable storage medium is located to execute the method for determining a mechanical equipment fault according to any one of claims 1 to 5.

8. A processor, characterized in that, The processor is configured to run a program, where, when the program runs, it executes the method for determining a mechanical equipment fault according to any one of claims 1 to 5.

Images