Hydraulic machine fault monitoring device

By collecting and preprocessing pressure curve data in the hydraulic press, combining the variational autoencoder and the prediction model of the generation of adversarial network, the advance prediction of hydraulic cylinder failure is achieved, and the problem of failure cannot be predicted in advance in the existing technology is solved, and the problem of failure cannot be predicted in advance in the existing technology is improved. Accuracy and stability of fault detection.

CN119062630AInactive Publication Date: 2024-12-03TIANSHUI METALFORMING MACHINE TOOL GRP ANHUI PRESS
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411112507.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, hydraulic cylinder fault monitoring mainly relies on pressure sensors to identify the pressure curve, and the failure cannot be predicted in advance, resulting in sudden damage to the hydraulic press and product scrapping.

Method used

The data acquisition module is used to collect the pressure curve data of the hydraulic press through the pressure sensor, combine the data preprocessing and prediction model building module, and use the variable autoencoder and the generation adversarial network to build the prediction model, perform filtering processing, feature extraction and fault detection, and output in real time. Failure prediction results.

Benefits of technology

It realizes prediction before a fault occurs, avoiding sudden damage to the hydraulic press and product scrapping, and improves the accuracy and robustness of fault prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119062630A_ABST
    Figure CN119062630A_ABST
Patent Text Reader

Abstract

The invention discloses a hydraulic machine fault monitoring device and belongs to the field of hydraulic machine fault monitoring. The monitoring device comprises a data acquisition module which is used for acquiring pressure curve data of the hydraulic machine during hydraulic work by using a pressure sensor mounted on the hydraulic machine; the data preprocessing module is used for filtering the collected pressure curve data to remove high-frequency noise; feature extraction is carried out on the pressure curve data; the prediction model construction module is used for constructing a prediction model in combination with a variational auto-encoder and a generative adversarial network, learning potential distribution of data through VAE, generating more real data distribution through GAN, and carrying out fault detection; the prediction model training module is used for inputting the preprocessed pressure curve data into the prediction model for training; and the prediction module is used for performing denoising and feature extraction on the pressure curve data acquired by the pressure sensor in real time, inputting the pressure curve data into the trained prediction model and outputting a fault prediction result, so that the fault is predicted before the fault occurs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of hydraulic press fault monitoring, and in particular relates to a hydraulic press fault monitoring device. Background Art

[0002] With the continuous development of science and technology, hydraulic presses occupy an important position in industrial production, providing power source for product press and other assembly processes; hydraulic cylinders are the core power components of hydraulic presses, and the life of hydraulic cylinders directly affects the life of hydraulic presses; during the operation of hydraulic presses, hydraulic cylinders need to be monitored and replaced in time after faults are detected to ensure the normal operation of hydraulic presses.

[0003] In the prior art, the monitoring of the hydraulic cylinder of a hydraulic press is carried out by using a pressure sensor to identify the pressure curve of the hydraulic cylinder, and the fault is identified only after the fault occurs. At this time, for the products being produced, there is a probability that the products will be scrapped. Therefore, it is necessary to predict the fault before the fault occurs to avoid the sudden damage of the hydraulic press causing the product to be scrapped. For this purpose, a hydraulic press fault monitoring device is proposed. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a hydraulic press fault monitoring device to solve the problems in the prior art.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] Hydraulic press fault monitoring device, comprising:

[0007] Data acquisition module: uses the pressure sensor installed on the hydraulic press to collect the pressure curve data of the hydraulic press during hydraulic operation;

[0008] Data preprocessing module: first filter the collected pressure curve data to remove high-frequency noise; then extract features from the pressure curve data;

[0009] Prediction model building module: Combining variational autoencoders and generative adversarial networks to build prediction models, VAE learns the potential distribution of data, and GAN generates more realistic data distribution for fault detection;

[0010] Prediction model training module: inputting the preprocessed pressure curve data into the prediction model for training;

[0011] And, prediction module: after denoising and feature extraction of the pressure curve data collected by the pressure sensor in real time, input it into the trained prediction model and output the fault prediction result.

[0012] Furthermore, the filtering process uses a moving average filter to remove noise from the pressure curve data, specifically:

[0013] Given a set of data y1,y2,...,y n , for the i-th data point y i Using the moving average filter, we can get the filtered data points

[0014]

[0015] Here, m is the size of the window.

[0016] Furthermore, the extracted features include: mean, variance, peak value and power spectral density, and the extracted features are combined into a data set as an input of the prediction model.

[0017] Furthermore, the calculation formula of the mean value M is:

[0018]

[0019] Among them, y i is the i-th data point, n is the total number of data points;

[0020] The calculation formula of the variance V is:

[0021]

[0022] The peak is the maximum or minimum value in the data.

[0023] Furthermore, when calculating the power spectral density, the time domain data is converted into frequency domain data by discrete Fourier transform; let the original data be y1,y2,...,y n , the sampling frequency is f s , after discrete Fourier transform, we get the spectrum data Y(f), and the calculation formula of power spectrum density S(f) is:

[0024]

[0025] Where |Y(f)| is the amplitude of the spectrum, f is the frequency, and n is the total number of data points.

[0026] Furthermore, the prediction model includes:

[0027] 1) Encoder, including:

[0028] Input layer: pressure curve data input;

[0029] Multiple hidden layers: including convolutional layers or fully connected layers for feature extraction and encoding;

[0030] Output layer: outputs the mean μ and variance σ of the latent space 2 ;

[0031] 2) Reparameterization layer: Output mean μ and variance σ from the encoder 2 Reparameterize the latent vector z in ;

[0032] 3) Decoder, including:

[0033] Input layer: receives the potential vector z from the reparameterized layer;

[0034] Multiple hidden layers: including fully connected layers or deconvolution layers, used to decode and reconstruct pressure curve data;

[0035] Output layer: output reconstructed pressure curve data;

[0036] 4) Generator, including:

[0037] Input layer: receives the potential vector of the reconstructed pressure curve data;

[0038] Multiple hidden layers: including fully connected layers or deconvolution layers, used to generate fake pressure curve data from the latent space;

[0039] Output layer: output the generated pressure curve data;

[0040] 5) Discriminator, including:

[0041] Input layer: receives the pressure curve data generated by the generator;

[0042] Multiple hidden layers: including fully connected layers or convolutional layers, used to determine the authenticity of input data;

[0043] Output layer: Outputs the probability that the input data is true data.

[0044] A hydraulic press fault monitoring method, using the above hydraulic press fault monitoring device, comprises the following steps:

[0045] S1, using a pressure sensor installed on a hydraulic press to collect pressure curve data during hydraulic operation on the hydraulic press;

[0046] S2, first filter the collected pressure curve data to remove high-frequency noise; then extract features from the pressure curve data, and combine the extracted features into a data set to input into the prediction model;

[0047] S3, uses a combination of variational autoencoders and generative adversarial networks to build a prediction model, uses VAE to learn the potential representation of data, and uses GAN to generate more realistic data samples;

[0048] S4, inputting the preprocessed pressure curve data into the prediction model for training;

[0049] S5, after denoising and feature extraction of the pressure curve data collected in real time by the pressure sensor, input the data into the trained prediction model and output the fault prediction result.

[0050] A computer storage medium stores a readable program, and when the program is running, the above-mentioned hydraulic press fault monitoring method can be executed.

[0051] An electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0052] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned hydraulic press fault monitoring method.

[0053] A computer program product includes computer instructions, wherein the computer instructions instruct a computing device to execute operations corresponding to the above-mentioned hydraulic press fault monitoring method.

[0054] Beneficial effects of the present invention:

[0055] 1. By setting up a pressure sensor to collect the pressure curve data of the hydraulic press, the data is used as the input of the prediction model to output the fault prediction result, so that the fault can be predicted before it occurs to avoid sudden damage to the hydraulic press and cause product scrapping.

[0056] 2. By combining variational autoencoders and generative adversarial networks in the prediction model, VAE is used to learn the potential representation of data, and GAN is used to generate more realistic data samples, thereby improving the accuracy and robustness of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Figure 1 It is a schematic diagram of the overall structure of the hydraulic press fault monitoring device of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Example 1

[0061] like Figure 1 As shown, the hydraulic press fault monitoring device includes: a data acquisition module, a data preprocessing module, a prediction model building module, a prediction model training module and a prediction module.

[0062] 1) Data acquisition module

[0063] The data acquisition module includes a pressure sensor installed on the hydraulic press to collect the pressure curve data of the hydraulic press during hydraulic operation. These data will be the input of the prediction model and finally output the fault prediction results.

[0064] 2) Data preprocessing module

[0065] In the data preprocessing module, the collected pressure curve data is first filtered to remove high-frequency noise; then the pressure curve data is feature extracted to extract statistical features (such as mean, variance, peak value) and frequency domain features (such as power spectrum density), and the extracted features are combined into a data set to input into the prediction model;

[0066] During filtering, a moving average filter is used to denoise the pressure curve data, specifically:

[0067] Given a set of data y1,y2,...,y n , for the i-th data point y i Using a simple moving average filter with a window size of m, we can get the filtered data points

[0068]

[0069] Where m is the size of the window, usually an odd number, so that the window can symmetrically cover the data around the current data point. The advantages of the simple moving average filter are simple implementation, fast calculation speed, and suitable for stationary data and medium-frequency noise.

[0070] During the feature extraction process, the extracted statistical features include mean, variance, and peak value; the mean M of the pressure curve data can be simply expressed as the arithmetic mean of all data points:

[0071]

[0072] Among them, y i is the ith data point and n is the total number of data points.

[0073] The variance V measures how dispersed the data points are from the mean:

[0074]

[0075] The peak is the maximum or minimum value in the data, which can be divided into positive peak (maximum value) and negative peak (minimum value).

[0076] Frequency domain features analyze the distribution of pressure curve data at different frequencies. Common frequency domain features include power spectral density (PSD). Power spectral density describes the energy distribution of the signal in the frequency domain and is a function of frequency.

[0077] When calculating the power spectral density, the time domain data can be converted into frequency domain data by discrete Fourier transform (DFT); let the original data be y1, y2, ..., y n The sampling frequency is f s , after DFT, we get the spectrum data Y(f), and the calculation formula of power spectrum density S(f) is:

[0078]

[0079] Where |Y(f)| is the amplitude of the spectrum, f is the frequency, and f s is the sampling frequency and n is the total number of data points.

[0080] 3) Prediction model building module:

[0081] The prediction model uses a combination of variational autoencoder (VAE) and generative adversarial network (GAN): VAE can learn the potential distribution of data, and GAN can generate more realistic data distribution for fault detection. Specifically, the prediction model includes: encoder, reparameterization layer, decoder, generator and discriminator;

[0082] The encoder includes:

[0083] Input layer: pressure curve data input;

[0084] Multiple hidden layers: including convolutional layers or fully connected layers for feature extraction and encoding;

[0085] Output layer: outputs the mean μ and variance σ of the latent space 2 ;

[0086] b Reparameterization layer: Output mean μ and variance σ from the encoder 2 Reparameterize the latent vector z in , making it differentiable and easy to sample;

[0087] The c decoder includes:

[0088] Input layer: receives the potential vector z from the reparameterized layer;

[0089] Multiple hidden layers: including fully connected layers or deconvolution layers, used to decode and reconstruct pressure curve data;

[0090] Output layer: output reconstructed pressure curve data;

[0091] d generator includes:

[0092] Input layer: receives the potential vector of the reconstructed pressure curve data;

[0093] Multiple hidden layers: including fully connected layers or deconvolution layers, used to generate fake pressure curve data from the latent space;

[0094] Output layer: output the generated pressure curve data;

[0095] The e discriminator includes:

[0096] Input layer: receives the pressure curve data generated by the generator;

[0097] Multiple hidden layers: including fully connected layers or convolutional layers, used to determine the authenticity of input data;

[0098] Output layer: Outputs the probability that the input data is true data.

[0099] The prediction model combines variational autoencoder (VAE) and generative adversarial network (GAN). This combination uses VAE to learn the potential representation of data and generates more realistic data samples through GAN. It can effectively learn and utilize the potential features of pressure curve data and generate more realistic data samples, thereby improving the accuracy and robustness of fault prediction.

[0100] 4) Prediction model training module:

[0101] The steps to train a prediction model are:

[0102] Step 1: Divide the preprocessed pressure curve data into a training set and a validation set in a ratio of 8:2;

[0103] Step 2: Input the training set data into the prediction model for training and tuning;

[0104] Step 3: Input the validation set into the trained prediction model to verify the prediction ability of the prediction model.

[0105] 5) Fault prediction module:

[0106] The pressure curve data collected by the pressure sensor in real time is preprocessed by the data preprocessing module, input into the trained prediction model, and outputs the fault prediction result.

[0107] Example 2

[0108] Based on the fault monitoring device mentioned in Example 1, a hydraulic press fault monitoring method is proposed in this embodiment, comprising the following steps:

[0109] S1, using the pressure sensor installed on the hydraulic press to collect the pressure curve data of the hydraulic press during hydraulic operation.

[0110] S2, first filter the collected pressure curve data to remove high-frequency noise; then perform feature extraction on the pressure curve data, extract statistical features (such as mean, variance, peak) and frequency domain features (such as power spectral density), and combine the extracted features into a data set to input into the prediction model.

[0111] S3, uses a combination of variational autoencoder (VAE) and generative adversarial network (GAN) to build a prediction model: VAE is used to learn the potential representation of the data, and GAN is used to generate more realistic data samples, so that the potential features of the pressure curve data can be effectively learned and utilized, and more realistic data samples can be generated, thereby improving the accuracy and robustness of fault prediction.

[0112] S4, inputting the preprocessed pressure curve data into the prediction model for training;

[0113] S5, after denoising and feature extraction of the pressure curve data collected in real time by the pressure sensor, input the data into the trained prediction model and output the fault prediction result.

[0114] The method of the present invention may be implemented in hardware, firmware, or as software or computer code that may be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded over a network and will be stored in a local recording medium, so that the method described herein may be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, processor, or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0115] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A hydraulic press fault monitoring device, characterized in that: include: Data acquisition module: uses the pressure sensor installed on the hydraulic press to collect the pressure curve data of the hydraulic press during hydraulic operation; Data preprocessing module: first filter the collected pressure curve data to remove high-frequency noise; then extract features from the pressure curve data; Prediction model building module: Combining variational autoencoders and generative adversarial networks to build prediction models, VAE learns the potential distribution of data, and GAN generates more realistic data distribution for fault detection; Prediction model training module: inputting the preprocessed pressure curve data into the prediction model for training; And, the fault prediction module: after denoising and feature extraction of the pressure curve data collected by the pressure sensor in real time, it is input into the trained prediction model and the fault prediction result is output.

2. The hydraulic press fault monitoring device according to claim 1, characterized in that: The filtering process uses a moving average filter to remove noise from the pressure curve data, specifically: Given a set of data y1,y2,...,y n , for the i-th data point y i Using the moving average filter, we can get the filtered data points Here, m is the size of the window.

3. The hydraulic press fault monitoring device according to claim 1, characterized in that: The extracted features include: mean, variance, peak and power spectral density, and the extracted features are combined into a data set as the input of the prediction model.

4. The hydraulic press fault monitoring device according to claim 3, characterized in that: The calculation formula of the mean value M is: Among them, y i is the i-th data point, n is the total number of data points; The calculation formula of the variance V is: The peak is the maximum or minimum value in the data.

5. The hydraulic press fault monitoring device according to claim 3, characterized in that: When calculating the power spectral density, the time domain data is converted into frequency domain data through discrete Fourier transform; let the original data be y1,y2,...,y n , the sampling frequency is f s , after discrete Fourier transform, we get the spectrum data Y(f), and the calculation formula of power spectrum density S(f) is: Where |Y(f)| is the amplitude of the spectrum, f is the frequency, and n is the total number of data points.

6. The hydraulic press fault monitoring device according to claim 1, characterized in that: The prediction model includes: 1) Encoder, including: Input layer: pressure curve data input; Multiple hidden layers: including convolutional layers or fully connected layers for feature extraction and encoding; Output layer: outputs the mean μ and variance σ of the latent space 2 ; 2) Reparameterization layer: Output mean μ and variance σ from the encoder 2 Reparameterize the latent vector z in ; 3) Decoder, including: Input layer: receives the potential vector z from the reparameterized layer; Multiple hidden layers: including fully connected layers or deconvolution layers, used to decode and reconstruct pressure curve data; Output layer: output reconstructed pressure curve data; 4) Generator, including: Input layer: receives the potential vector of the reconstructed pressure curve data; Multiple hidden layers: including fully connected layers or deconvolution layers, used to generate fake pressure curve data from the latent space; Output layer: output the generated pressure curve data; 5) Discriminator, including: Input layer: receives the pressure curve data generated by the generator; Multiple hidden layers: including fully connected layers or convolutional layers, used to determine the authenticity of input data; Output layer: Outputs the probability that the input data is true data.

7. A method for monitoring a hydraulic press fault, using the hydraulic press fault monitoring device according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1, using a pressure sensor installed on a hydraulic press to collect pressure curve data during hydraulic operation on the hydraulic press; S2, first filter the collected pressure curve data to remove high-frequency noise; then extract features from the pressure curve data, and combine the extracted features into a data set to input into the prediction model; S3, uses a combination of variational autoencoders and generative adversarial networks to build a prediction model, uses VAE to learn the potential representation of data, and uses GAN to generate more realistic data samples; S4, inputting the preprocessed pressure curve data into the prediction model for training; S5, after denoising and feature extraction of the pressure curve data collected in real time by the pressure sensor, input the data into the trained prediction model and output the fault prediction result.

8. A computer storage medium storing a readable program, characterized in that: When the program is running, the hydraulic press fault monitoring method described in claim 7 can be executed.

9. An electronic device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the hydraulic press fault monitoring method as described in claim 7.

10. A computer program product comprising computer instructions, characterized in that: The computer instructions instruct the computing device to execute operations corresponding to the hydraulic press fault monitoring method as described in claim 7.

Citation Information

Cited By

  • Press machine fault early warning method

    CN120429620A