Abrasive particle shape recognition model determination method, abrasive particle shape recognition model application method and abrasive particle shape recognition model determination device

Abstract

The invention discloses a determination method, an application method and a device of an abrasive particle shape recognition model, and relates to the technical field of oil abrasive particle monitoring, and the determination method of the abrasive particle shape recognition model comprises the following steps: obtaining a historical image data set; the historical image data set comprises historical Gramer two-dimensional images and corresponding state labels; inputting the historical Gramer two-dimensional image into a convolutional neural network model to obtain the output of the convolutional neural network model; determining a loss value based on the output of the convolutional neural network model, the state label corresponding to the historical Gramer two-dimensional image and the determined loss function; the loss function is a cross entropy loss function; and optimizing network parameters of the convolutional neural network model based on the loss value to obtain an abrasive particle shape recognition model. The method can quickly and accurately identify the shape of the abrasive particle.

Description

Technical Field

[0001] The present application relates to the technical field of oil wear particle monitoring, and in particular to a method for determining an abrasive particle shape recognition model, an application method, and a device. Background Art

[0002] Studies have shown that approximately 80% of all mechanical failures are caused by poor lubrication of the equipment. Therefore, the quality of the equipment's lubrication environment determines its operating status, giving rise to a range of oil monitoring technologies. Oil monitoring technology evaluates the equipment's operating status by detecting changes in lubricating oil performance and the characteristics of wear particles in the oil during operation, providing early warnings of potential failures and providing information to analyze the cause, type, and location of the failure. The shape characteristics of abrasive particles have long been a research hotspot. This is because different wear mechanisms produce wear particles of varying shapes. Analyzing the shape characteristics of these wear particles can help determine the specific cause of equipment failure and the wear status of the equipment.

[0003] Traditional oil wear particle analysis relies on offline microscopic observation and online ferrography to collect corresponding wear particle images. It then integrates artificial intelligence to accurately segment the wear particles within the image, obtain information such as the wear particle's contour and circumference, and classify the wear particles based on their different shape characteristics. This method suffers from low efficiency and poor real-time performance. Wear particle signals collected by oil wear particle sensors also suffer from shortcomings such as unclear representation of wear particle shape characteristics and difficulty in extracting the corresponding wear particle shape. Existing deep learning-based wear particle classification methods suffer from data redundancy and insufficient feature extraction when directly processing time series signals.

[0004] Therefore, how to quickly and accurately identify the shape of abrasive particles has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this application is to provide a method for determining an abrasive particle shape recognition model, an application method and a device, which can quickly and accurately identify the abrasive particle shape.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for determining an abrasive particle shape recognition model, the method comprising:

[0008] Acquire a historical image dataset; the historical image dataset includes: historical Gram two-dimensional images and corresponding state labels.

[0009] The historical Gram two-dimensional image is input into a convolutional neural network model (CNN) to obtain an output of the convolutional neural network model.

[0010] A loss value is determined based on the output of the convolutional neural network model, the state label corresponding to the historical Gram two-dimensional image, and a determined loss function; the loss function is a cross entropy loss function.

[0011] Based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model.

[0012] Optionally, obtain a historical image dataset, specifically including:

[0013] Obtain historical wear particle signals.

[0014] The historical wear particle signal is subjected to noise reduction processing to obtain a noise-reduced historical wear particle signal.

[0015] Data probability density (DPD) extraction and Gram angular field (GAF) encoding are performed on the historical wear particle signal after the noise reduction process to obtain a historical Gram angular field two-dimensional image.

[0016] Based on the historical Gram two-dimensional images and corresponding state labels, a historical image dataset is obtained.

[0017] Optionally, performing noise reduction processing on the historical wear particle signal to obtain the noise-reduced historical wear particle signal specifically includes:

[0018] The Daubechies 4th order (db4) wavelet is used as a basis function to perform a 4-layer decomposition on the historical wear particle signal to obtain a coefficient set; the coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients.

[0019] The unbiased risk estimation (Rigorous SURE) is used to calculate the threshold of the high-frequency detail coefficient of each layer.

[0020] Based on the threshold of the high-frequency detail coefficient of each layer, the high-frequency detail coefficient of the corresponding layer is shrunk to obtain the high-frequency detail coefficient after shrunk.

[0021] A new coefficient set is constructed based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage process.

[0022] The new coefficient set is reconstructed by using inverse wavelet transform to obtain the historical wear particle signal after noise reduction processing.

[0023] Optionally, performing data probability density extraction and Gram angle field encoding on the historical wear particle signal after the noise reduction process to obtain a historical Gram 2D ​​image specifically includes:

[0024] Based on the historical wear particle signal after the noise reduction process, an amplitude frequency histogram is constructed.

[0025] The amplitude-frequency histogram is normalized to obtain a normalized amplitude-frequency histogram.

[0026] The normalized amplitude frequency histogram is divided into a number of equal-width intervals, and the number of data points in each interval is counted.

[0027] The probability density is calculated based on the number of data points in each interval.

[0028] The probability density is normalized to obtain a normalized probability density.

[0029] The normalized probability density is converted to polar angles.

[0030] A Gram matrix is ​​constructed based on the sum of the cosines of the polar angles.

[0031] Based on the size of the element values ​​in the Gram matrix, red, green and blue colors are assigned to obtain a historical Gram matrix two-dimensional image.

[0032] In a second aspect, the present application provides an application method of an abrasive particle shape recognition model, the application method of the abrasive particle shape recognition model comprising:

[0033] Obtain the wear particle signal to be measured.

[0034] The wear particle signal to be measured is subjected to noise reduction processing to obtain the wear particle signal to be measured after noise reduction processing.

[0035] Data probability density extraction and Gram angle field encoding are performed on the wear particle signal to be measured after the noise reduction process to obtain a Gram two-dimensional image to be measured.

[0036] The Gram two-dimensional image to be measured is input into the wear particle shape recognition model to obtain a state label corresponding to the Gram two-dimensional image to be measured; the wear particle shape recognition model is a model obtained based on any of the above-mentioned wear particle shape recognition model determination methods.

[0037] Optionally, performing noise reduction processing on the wear particle signal to be measured to obtain the wear particle signal to be measured after noise reduction processing specifically includes:

[0038] The Daubechies 4th-order wavelet is used as a basis function to perform a 4-layer decomposition on the wear particle signal to be measured, thereby obtaining a coefficient set. The coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients.

[0039] The threshold of the high-frequency detail coefficient of each layer is calculated using unbiased risk estimation.

[0040] Based on the threshold of the high-frequency detail coefficient of each layer, the high-frequency detail coefficient of the corresponding layer is shrunk to obtain the high-frequency detail coefficient after shrunk.

[0041] A new coefficient set is constructed based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage process.

[0042] The new coefficient set is reconstructed by using inverse wavelet transform to obtain the wear particle signal to be measured after noise reduction processing.

[0043] Optionally, performing data probability density extraction and Gram angle field encoding on the wear particle signal to be measured after the noise reduction process to obtain a Gram two-dimensional image to be measured specifically includes:

[0044] An amplitude frequency histogram is constructed based on the wear particle signal to be measured after the noise reduction processing.

[0045] The amplitude-frequency histogram is normalized to obtain a normalized amplitude-frequency histogram.

[0046] The normalized amplitude frequency histogram is divided into a number of equal-width intervals, and the number of data points in each interval is counted.

[0047] The probability density is calculated based on the number of data points in each interval.

[0048] The probability density is normalized to obtain a normalized probability density.

[0049] The normalized probability density is converted to polar angles.

[0050] A Gram matrix is ​​constructed based on the sum of the cosines of the polar angles.

[0051] Based on the size of the element values ​​in the Gram matrix, red, green and blue colors are assigned to obtain a Gram two-dimensional image to be measured.

[0052] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned methods for determining the abrasive shape recognition model or any of the above-mentioned methods for applying the abrasive shape recognition model.

[0053] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for determining the abrasive shape recognition model or any of the above-mentioned methods for applying the abrasive shape recognition model.

[0054] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for determining the wear particle shape recognition model or any of the above-mentioned methods for applying the wear particle shape recognition model.

[0055] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0056] The present application provides a method for determining, applying and applicating a wear particle shape recognition model. First, a historical image data set is obtained, wherein the historical image data set includes: historical Grammar two-dimensional images and corresponding state labels, which can more reliably identify the wear particle shape states under various different situations; second, the historical Grammar two-dimensional images are input into a convolutional neural network model to obtain the output of the convolutional neural network model; a loss value is determined based on the output of the convolutional neural network model, the state labels corresponding to the historical Grammar two-dimensional images and a determined loss function; the loss function is a cross-entropy loss function; based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model, which can automatically learn the key features of the wear particle shape from the image without manually designing a feature extraction algorithm, thereby improving the performance and efficiency of the model and reducing the influence of manual intervention and subjective factors; at the same time, the cross-entropy loss function can provide a clear direction and optimization goal for model training, so that the model can automatically adjust parameters to minimize the difference between the predicted results and the true labels. This comparison method based on probability distribution can more sensitively capture changes in the accuracy of model predictions, thereby more effectively guiding the model training process and improving the model's classification performance. In addition, by optimizing the network parameters, the model can better fit the training data and improve the recognition accuracy of wear particle shapes in historical Grammar two-dimensional images, thereby more reliably identifying wear particle shape states in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] 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 use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 This is a diagram of the application environment of a method for determining a wear particle shape recognition model in one embodiment of the present application.

[0059] Figure 2 A flowchart of a method for determining an abrasive particle shape recognition model provided in one embodiment of the present application.

[0060] Figure 3 A diagram of a laboratory simulation of abrasive particles in oil passing through an oil abrasive particle sensor device provided in one embodiment of the present application.

[0061] Figure 4 This is a diagram showing the signal of a tiny abrasive particle passing through a sensor and the effect of wavelet decomposition and noise reduction provided in one embodiment of the present application.

[0062] Figure 5 This is an amplitude frequency histogram corresponding to the denoised signal of wear particles of different shapes provided in one embodiment of the present application.

[0063] Figure 6 This is a characteristic diagram of the Gram angle field converted from abrasive particles of different shapes provided in one embodiment of the present application.

[0064] Figure 7 A schematic flow chart of an application method of an abrasive shape recognition model provided in one embodiment of the present application.

[0065] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions 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 in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0067] The present application discloses a wear particle shape recognition method based on DPD-GAF-CNN, specifically a wear particle shape recognition method based on data probability density and Gram angular field coding combined with convolutional neural network, which realizes the recognition of wear particles of different shapes by classifying the DPD-GAF feature map generated by the original signal.

[0068] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0069] The method for determining the wear particle shape recognition model provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the acquired historical image data set to the server 104, and the historical image data set includes: historical Grammar two-dimensional images and corresponding status labels; after the server 104 receives the historical image data set, for the historical image data set, the server 104 inputs the historical Grammar two-dimensional images into the convolutional neural network model to obtain the output of the convolutional neural network model; based on the output of the convolutional neural network model, the status labels corresponding to the historical Grammar two-dimensional images and the determined loss function, the loss value is determined; the loss function is a cross entropy loss function; based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model. The server 104 can feed back the obtained wear particle shape recognition model to the terminal 102. In addition, in some embodiments, the method for determining the wear particle shape recognition model can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly determine the wear particle shape recognition model for the historical image data set, or the server 104 can obtain the historical image data set from the data storage system and determine the wear particle shape recognition model for the historical image data set.

[0070] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, and tablet computers. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or a cloud server.

[0071] In an exemplary embodiment, Figure 2 As shown, a method for determining a wear particle shape recognition model is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps S1 to S4.

[0072] S1: Acquire a historical image dataset; the historical image dataset includes: historical Gram 2D ​​images and corresponding state labels;

[0073] S2: Inputting the historical Gram two-dimensional image into a convolutional neural network model to obtain an output of the convolutional neural network model;

[0074] S3: Determine a loss value based on the output of the convolutional neural network model, the state label corresponding to the historical Gram 2D ​​image, and a determined loss function; the loss function is a cross entropy loss function;

[0075] S4: Based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model.

[0076] By implementing the above steps S1 to S4, a convolutional neural network is used to improve the training speed and accuracy of the wear particle shape recognition model, replacing manual shutdown inspections and subjective experience judgments, and replacing traditional offline wear particle shape recognition with online real-time analysis, which provides a reliable basis for equipment fault warning, reduces equipment maintenance time and costs, improves production efficiency and quality, and promotes the intelligent development of mechanical equipment.

[0077] As an optional implementation, in step S1, obtaining a historical image dataset specifically includes:

[0078] S11: Acquire historical wear particle signals.

[0079] S12: performing noise reduction processing on the historical wear particle signal to obtain a noise-reduced historical wear particle signal.

[0080] S13: performing data probability density extraction and Gram angle field encoding on the historical wear particle signal after the noise reduction process to obtain a historical Gram 2D ​​image.

[0081] S14: Obtain a historical image dataset based on the historical Gram 2D ​​image and the corresponding state label.

[0082] Specifically, in step S11, the induced voltage signal generated when the wear particles pass through the oil wear particle sensor is collected and transmitted to an oscilloscope for display and storage.

[0083] Under the laboratory conditions of this application, Figure 3 The experiment was conducted using the device shown. After abrasive particles were fixed to a rubber hose, the speed-regulating motor was activated, causing the driving pulley to rotate the hose counterclockwise along the tensioning pulley and driven pulley, simulating the movement of abrasive particles in oil. As abrasive particles passed through the oil abrasive particle sensor, corresponding induced voltage signal fluctuations were generated on the oscilloscope. These induced voltage signals were recorded and saved for subsequent analysis.

[0084] As an optional implementation, in step S12, performing noise reduction processing on the historical wear particle signal to obtain the noise-reduced historical wear particle signal specifically includes:

[0085] S121: using Daubechies 4th-order wavelet as basis function, performing 4-layer decomposition on the historical wear particle signal to obtain a coefficient set; the coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients.

[0086] S122: Using unbiased risk estimation, calculate the threshold of the high-frequency detail coefficient of each layer.

[0087] S123: Based on the threshold of each layer of high-frequency detail coefficients, shrinking processing is performed on the high-frequency detail coefficients of the corresponding layer to obtain the high-frequency detail coefficients after shrinking processing.

[0088] S124: Constructing a new coefficient set based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage processing.

[0089] S125: Reconstruct the new coefficient set using inverse wavelet transform to obtain the historical wear particle signal after noise reduction processing.

[0090] Specifically, the original signal is de-noised by using db4 wavelet to decompose the original signal into 4 layers to reduce noise, thereby increasing the signal strength of tiny wear particles while retaining the signal characteristics as much as possible. The noise reduction effect of a certain particle is as follows: Figure 4 As shown, the following steps are included:

[0091] (1) Wavelet decomposition: Daubechies 4th-order wavelet is selected as the basis function, and the original wear particle signal is decomposed into a 4-layer coefficient set: {A4, D4, D3, D2, D1}, where A4 is the low-frequency approximation coefficient; D4, D3, D2, D1 are high-frequency detail coefficients;

[0092] (2) Threshold processing: Use unbiased risk estimation to calculate the threshold λ of each layer detail coefficient j For each layer detail coefficient D j Perform shrinkage processing, keeping the A4 approximation coefficient unchanged:

[0093]

[0094] Among them, σ j is the detail coefficient D of the jth layer j The standard deviation of N is the signal length.

[0095] (3) Signal reconstruction: coefficients based on threshold processing Reconstruct the denoised signal s by inverse wavelet transform denoised (t). For each layer detail coefficient D j Shrinkage treatment can be obtained The calculation formula is:

[0096]

[0097] As an optional implementation, in step S13, data probability density extraction and Gram angle field encoding are performed on the historical wear particle signal after the noise reduction process to obtain a historical Gram 2D ​​image, specifically including:

[0098] S131: constructing an amplitude frequency histogram based on the historical wear particle signal after the noise reduction processing.

[0099] S132: Normalizing the amplitude-frequency histogram to obtain a normalized amplitude-frequency histogram.

[0100] S133: Divide the normalized amplitude frequency histogram into a number of equal-width intervals, and count the number of data points in each interval.

[0101] S134: Calculate the probability density based on the number of data points in each interval.

[0102] S135: Normalizing the probability density to obtain a normalized probability density.

[0103] S136: Convert the normalized probability density into a polar angle.

[0104] S137: Constructing a Gram matrix based on the cosine sum of the polar angle.

[0105] S138: Assign red, green and blue colors based on the values ​​of the elements in the Gram matrix to obtain a historical Gram matrix two-dimensional image.

[0106] Specifically, the image conversion method combining DPD and GAF ​​is used to generate a DPD-GAF feature map of the corresponding signal, including the following steps:

[0107] (1) Data probability density extraction, using the denoised signal s denoised (t) Construct a Figure 5 The amplitude frequency histogram is displayed. The amplitude frequency histogram is normalized and divided into K = 256 equal-width intervals. The number of data points in each interval is counted and the probability density is calculated:

[0108]

[0109] Here, count(k) represents the number of data points in each interval.

[0110] (2) Gram angle field coding, normalize the probability density p(k) and convert it into polar angle θ k, realize polar coordinate mapping, construct a K×K Gram matrix through the cosine sum of the polar angle, and assign red, green and blue colors according to the size change of the matrix element value, forming Figure 6 GAF feature diagram shown.

[0111] The polar angle is calculated as follows:

[0112]

[0113] in, is the result of normalization of probability density p(k).

[0114] The formula for calculating elements in the Gram matrix is:

[0115] GAF(i,j)=cos(θ i +θ j ) i,j=1,2,...,K (5);

[0116] CNN model classification: A CNN model is constructed to classify the DPD-GAF feature map generated based on the sensor signal data and output the wear particle shape classification results.

[0117] (1) Create an image dataset, including the image files generated in step 3 and the state labels of the wear particle shapes associated therewith, and divide them into a ratio of 7:3 for training and testing the classification model.

[0118] (2) Build a CNN model for image classification, including the construction of convolutional layers, pooling layers, and fully connected layers. Use sigmoid as the activation function, the pooling form is maximum pooling, and the cross entropy loss as the loss function. The input of the model is the Gram 2D ​​image, and the output is the state label corresponding to the Gram 2D ​​image.

[0119] (3) Use the training data to train the model, learn the model to the target model and save it; input the Gram two-dimensional image into the trained model, output the wear particle shape recognition result, and realize the real-time monitoring of the wear particle shape during the operation of the equipment.

[0120] The beneficial effects of the present application are: when the power is connected and the switch is turned on, the gear pump starts working, pumping the oil in the gear box into the oil circuit, and through the oil wear particle sensor, the sensor collects the induced voltage signal data of the wear particles passing through the oil, and improves the signal-to-noise ratio of the obtained signal through the signal processing method of wavelet decomposition and noise reduction, and encodes the time series signal into a two-dimensional image through the Gram angle field. The convolutional neural network is used to improve the training speed and accuracy of the wear particle shape recognition model, replacing manual shutdown inspections and subjective experience judgments, and replacing traditional offline wear particle shape recognition with online real-time analysis, providing a reliable basis for equipment fault warning, reducing equipment maintenance time and cost, improving production efficiency and quality, and promoting the intelligent development of mechanical equipment.

[0121] In an exemplary embodiment, Figure 7 As shown, a method for applying the wear particle shape recognition model is provided, and the method for applying the wear particle shape recognition model includes:

[0122] A1: Obtain the wear particle signal to be measured.

[0123] A2: performing noise reduction processing on the wear particle signal to be measured to obtain the wear particle signal to be measured after noise reduction processing.

[0124] A3: performing data probability density extraction and Gram angle field encoding on the wear particle signal to be measured after the noise reduction process to obtain a Gram two-dimensional image to be measured.

[0125] A4: Input the Gram's two-dimensional image to be measured into a wear particle shape recognition model to obtain a state label corresponding to the Gram's two-dimensional image to be measured; the wear particle shape recognition model is a model obtained based on the wear particle shape recognition model determination method described above.

[0126] As an optional implementation manner, in step A2, performing noise reduction processing on the wear particle signal to be measured to obtain the wear particle signal to be measured after noise reduction processing specifically includes:

[0127] A21: Using Daubechies 4th-order wavelet as basis function, the wear particle signal to be measured is decomposed into 4 layers to obtain a coefficient set; the coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients.

[0128] A22: Use unbiased risk estimation to calculate the threshold of the high-frequency detail coefficient of each layer.

[0129] A23: Based on the threshold of the high-frequency detail coefficient of each layer, shrinking processing is performed on the high-frequency detail coefficient of the corresponding layer to obtain the high-frequency detail coefficient after shrinking processing.

[0130] A24: Construct a new coefficient set based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage processing.

[0131] A25: Reconstruct the new coefficient set using inverse wavelet transform to obtain the wear particle signal to be measured after noise reduction processing.

[0132] As an optional implementation, in step A3, data probability density extraction and Gram angle field encoding are performed on the wear particle signal to be measured after the noise reduction process to obtain a Gram two-dimensional image to be measured, which specifically includes:

[0133] A31: constructing an amplitude frequency histogram based on the wear particle signal to be measured after the noise reduction processing.

[0134] A32: Normalize the amplitude-frequency histogram to obtain a normalized amplitude-frequency histogram.

[0135] A33: Divide the normalized amplitude frequency histogram into a number of equal-width intervals, and count the number of data points in each interval.

[0136] A34: Calculate the probability density based on the number of data points in each interval.

[0137] A35: Normalize the probability density to obtain a normalized probability density.

[0138] A36: Convert the normalized probability density into a polar angle.

[0139] A37: Construct a Gram matrix based on the cosine sum of the polar angle.

[0140] A38: Assign red, green and blue colors based on the values ​​of the elements in the Gram matrix to obtain a Gram two-dimensional image to be measured.

[0141] The present application also provides an application scenario, which applies the above-mentioned wear particle shape recognition model application method. Specifically: the wear particle shape recognition model application method provided in this embodiment can be applied in the oil wear particle monitoring scenario. The oil wear particle monitoring scenario includes: a wear particle shape recognition model determination link, a wear particle signal acquisition link, a noise reduction link, a data probability density extraction and Gram angle field encoding link, and an identification link; first, a historical image data set is obtained; the historical image data set includes: a historical Gram two-dimensional image and a corresponding state label; the historical Gram two-dimensional image is input into a convolutional neural network model to obtain the output of the convolutional neural network model; based on the output of the convolutional neural network model, the state label corresponding to the historical Gram two-dimensional image and a determined loss function, a loss value is determined; the loss function is a cross entropy loss function; based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model; then, the wear particle signal to be measured is obtained; the wear particle signal to be measured is subjected to noise reduction processing to obtain the wear particle signal to be measured after noise reduction processing; the wear particle signal to be measured after noise reduction processing is subjected to data probability density extraction and Gram angle field encoding to obtain the Gram two-dimensional image to be measured; the Gram two-dimensional image to be measured is input into the wear particle shape recognition model to obtain the state label corresponding to the Gram two-dimensional image to be measured.

[0142] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store historical image data sets or wear particle signals to be measured. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes a method for determining a wear particle shape recognition model or an application method of a wear particle shape recognition model.

[0143] Those skilled in the art will understand that Figure 8The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0144] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the above method embodiments when executing the computer program.

[0145] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the above-mentioned method embodiments when executed by a processor.

[0146] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the above method embodiments are implemented.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0148] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0149] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0150] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for determining a wear particle shape recognition model, characterized in that: The method for determining the wear particle shape recognition model includes: Acquire a historical image dataset; the historical image dataset includes: historical Gram two-dimensional images and corresponding state labels; Inputting the historical Gram two-dimensional image into a convolutional neural network model to obtain an output of the convolutional neural network model; Determining a loss value based on the output of the convolutional neural network model, the state label corresponding to the historical Gram two-dimensional image, and a determined loss function; the loss function is a cross entropy loss function; Based on the loss value, the network parameters of the convolutional neural network model are optimized to obtain a wear particle shape recognition model.

2. The method for determining the wear particle shape recognition model according to claim 1, wherein: Obtain historical image datasets, including: Obtain historical wear particle signals; Performing noise reduction processing on the historical wear particle signal to obtain a noise-reduced historical wear particle signal; Performing data probability density extraction and Gram angle field encoding on the historical wear particle signal after the noise reduction process to obtain a historical Gram two-dimensional image; Based on the historical Gram two-dimensional images and corresponding state labels, a historical image dataset is obtained.

3. The method for determining the wear particle shape recognition model according to claim 2, wherein: Performing noise reduction processing on the historical wear particle signal to obtain the noise-reduced historical wear particle signal specifically includes: Using Daubechies 4th order wavelet as basis function, the historical wear particle signal is decomposed into 4 layers to obtain a coefficient set; the coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients; Using unbiased risk estimation, the threshold of the high-frequency detail coefficient of each layer is calculated; Based on the threshold of the high-frequency detail coefficient of each layer, the high-frequency detail coefficient of the corresponding layer is shrunk to obtain the high-frequency detail coefficient after shrunk; Constructing a new coefficient set based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage processing; The new coefficient set is reconstructed by using inverse wavelet transform to obtain the historical wear particle signal after noise reduction processing.

4. The method for determining the wear particle shape recognition model according to claim 2, wherein: The historical wear particle signal after the noise reduction process is subjected to data probability density extraction and Gram angle field encoding to obtain a historical Gram two-dimensional image, specifically including: constructing an amplitude frequency histogram based on the historical wear particle signal after the noise reduction process; Normalizing the amplitude-frequency histogram to obtain a normalized amplitude-frequency histogram; Dividing the normalized amplitude frequency histogram into a number of equal-width intervals, and counting the number of data points in each interval; Calculate the probability density based on the number of data points in each interval; Normalizing the probability density to obtain a normalized probability density; Converting the normalized probability density into polar angles; constructing a Gram matrix based on the sum of the cosines of the polar angles; Based on the size of the element values ​​in the Gram matrix, red, green and blue colors are assigned to obtain a historical Gram matrix two-dimensional image.

5. An application method of a wear particle shape recognition model, characterized in that: The application method of the wear particle shape recognition model includes: Acquire the wear particle signal to be measured; performing noise reduction processing on the wear particle signal to be measured to obtain the wear particle signal to be measured after noise reduction processing; Performing data probability density extraction and Gram angle field encoding on the wear particle signal to be measured after the noise reduction process to obtain a Gram two-dimensional image to be measured; The Gram two-dimensional image to be measured is input into the wear particle shape recognition model to obtain a state label corresponding to the Gram two-dimensional image to be measured; the wear particle shape recognition model is a model obtained based on the wear particle shape recognition model determination method according to any one of claims 1-4.

6. The application method of the wear particle shape recognition model according to claim 5, characterized in that: Performing noise reduction processing on the wear particle signal to be measured to obtain the wear particle signal to be measured after noise reduction processing specifically includes: Using Daubechies 4th order wavelet as basis function, the wear particle signal to be measured is decomposed into 4 layers to obtain a coefficient set; the coefficient set includes: 1 layer of low-frequency approximation coefficients and 4 layers of high-frequency detail coefficients; Using unbiased risk estimation, the threshold of the high-frequency detail coefficient of each layer is calculated; Based on the threshold of the high-frequency detail coefficient of each layer, the high-frequency detail coefficient of the corresponding layer is shrunk to obtain the high-frequency detail coefficient after shrunk; Constructing a new coefficient set based on the high-frequency detail coefficients and low-frequency approximation coefficients after the shrinkage processing; The new coefficient set is reconstructed by using inverse wavelet transform to obtain the wear particle signal to be measured after noise reduction processing.

7. The application method of the wear particle shape recognition model according to claim 5, characterized in that: The wear particle signal to be measured after the noise reduction process is subjected to data probability density extraction and Gram angle field encoding to obtain a Gram two-dimensional image to be measured, specifically comprising: constructing an amplitude frequency histogram based on the wear particle signal to be measured after the noise reduction processing; Normalizing the amplitude-frequency histogram to obtain a normalized amplitude-frequency histogram; Dividing the normalized amplitude frequency histogram into a number of equal-width intervals, and counting the number of data points in each interval; Calculate the probability density based on the number of data points in each interval; Normalizing the probability density to obtain a normalized probability density; Converting the normalized probability density into polar angles; constructing a Gram matrix based on the sum of the cosines of the polar angles; Based on the size of the element values ​​in the Gram matrix, red, green and blue colors are assigned to obtain a Gram two-dimensional image to be measured.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for determining the abrasive shape recognition model according to any one of claims 1 to 4 or the method for applying the abrasive shape recognition model according to any one of claims 5 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for determining the wear particle shape recognition model according to any one of claims 1 to 4 or the method for applying the wear particle shape recognition model according to any one of claims 5 to 7 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for determining the wear particle shape recognition model according to any one of claims 1 to 4 or the method for applying the wear particle shape recognition model according to any one of claims 5 to 7 is implemented.

Images