Intelligent detection method and system for material crushing

Abstract

The invention provides an intelligent detection method and system for material crushing. The method comprises the following steps: acquiring sound data, vibration data and fault types of crushing equipment in a crushing operation process, extracting time domain features and frequency domain features of the sound data and the vibration data, splicing the time domain features and the frequency domain features into a vector, and taking a splicing vector of the sound data and a splicing vector of the vibration data as model input to obtain a model; and performing model training by taking the fault type of the crushing equipment as model output, inputting new sound data and vibration data into the trained model to obtain a prediction result of the fault type, and adjusting the crushing equipment according to the prediction result. According to the method provided by the invention, the fault type of the crushing equipment is identified based on the feature vectors of the sound data and the vibration data, the fault type detection efficiency is improved, the maintenance cost of the crushing equipment is reduced, the features of the sound data and the vibration data are deeply learned through the model, and the fault type detection accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of crushing detection technology, and in particular to a material crushing intelligent detection method and system. Background Art

[0002] With the rapid development of society, the treatment of scrap metals, scrapped vehicles, industrial scrap equipment, etc. has become increasingly prominent. How to efficiently convert them into reusable metal fragments or particles to promote resource recycling and environmental protection has become a difficult problem that needs to be solved in this field.

[0003] Since the internal components of some scrapped equipment are made of stainless steel, brass, titanium and other materials with high rigidity and elasticity, if the crushing rate is too fast during the crushing process, it will be easy for the rigid material to not be able to enter the cutter head of the crusher smoothly or get stuck in the cutter head. If the cutter head is stuck for a long time, it may also cause problems such as bearing failure. Therefore, it is necessary to adjust the crushing rate during the crushing process to prevent raw materials with large deflection from repeatedly jumping near the cutter head and not entering the working area of ​​the cutter head. However, blindly reducing the working rate will also affect the working efficiency of the entire crushing process. Therefore, a crushing detection method for a crushing device is needed, which can detect the above problems in real time during the crushing process, and adjust the speed of the crushing device or repair the crushing device in real time according to the detected problems, so as to improve the working efficiency and reduce the maintenance cost of the crushing device. Summary of the invention

[0004] The present invention describes a material crushing intelligent detection method and system, which can solve the above technical problems.

[0005] According to a first aspect, a material crushing intelligent detection method is provided. The method comprises:

[0006] Collect the sound data, vibration data and fault types of the crushing equipment during the crushing operation, and record the time of data collection;

[0007] Extracting time domain features of the sound data and the vibration data, and converting the sound data and the vibration data from the time domain to the frequency domain by Fourier transform to extract the frequency domain features of the data;

[0008] The time domain features and the frequency domain features are concatenated into a vector to obtain a concatenated vector of the sound data and a concatenated vector of the vibration data respectively;

[0009] Using the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input and the fault type of the crushing equipment as the model output, the model is trained to obtain a trained model;

[0010] Collect new sound and vibration data and input them into the trained model to obtain the prediction results of the fault type and adjust the crushing equipment according to the prediction results.

[0011] In some embodiments, the fault types include: the cutter head is stuck, the material does not enter the cutter head, and the bearing is faulty.

[0012] In some embodiments, the time domain features include: mean, variance, peak, kurtosis, kurtosis, skewness, and energy.

[0013] In some embodiments, the frequency domain features include: spectrum, power spectrum density, harmonic frequency, fundamental frequency, bandwidth, spectrum spread, spectrum attenuation, and spectrum peak.

[0014] In some embodiments, after the sound data and the vibration data are collected, the sound data and the vibration data are first filtered to remove background noise and interference.

[0015] In some embodiments, after the time domain features and frequency domain features of the sound data and vibration data are extracted, the time domain features and frequency domain features data are first normalized.

[0016] In some embodiments, the model includes: an input layer, a CNN module, an LSTM module, a feature fusion layer, a fully connected layer and an output layer; wherein the CNN module includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer, the convolution kernel size of the convolution layer is 3, and the activation function is ReLU, which is used to extract local features of sound data, the pooling layer is used to reduce the spatial dimension of the feature map, and the flattening layer is used to flatten the output of the convolution layer so as to connect with the fully connected layer, and the fully connected layer is used to further extract sound data features; the LSTM module includes two LSTM layers and one fully connected layer, the two LSTM layers are used to capture features of different levels of vibration data, and the fully connected layer is used to further extract vibration data features and perform nonlinear transformation; the feature fusion layer is used to splice the outputs of the CNN module and the LSTM module; the fully connected layer is used to fuse the sound data features extracted from the CNN module with the vibration data features extracted from the LSTM module.

[0017] In some embodiments, the feature fusion layer in the model uses a self-attention mechanism to fuse the outputs of the CNN module and the LSTM module.

[0018] According to a second aspect, a material crushing intelligent detection system is provided, the system comprising:

[0019] The data collection module is used to collect the sound data, vibration data and fault types of the crushing equipment during the crushing operation, and record the time of data collection;

[0020] A feature extraction module is used to extract time domain features of sound data and vibration data, and convert the sound data and vibration data from the time domain to the frequency domain through Fourier transform to extract the frequency domain features of the data;

[0021] A vector splicing module is used to splice the time domain features and the frequency domain features into a vector, and obtain a splicing vector of the sound data and a splicing vector of the vibration data respectively;

[0022] A model training module, used for taking the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input and taking the fault type of the crushing equipment as the model output to perform model training;

[0023] The model prediction module is used to input the collected new sound data and vibration data into the trained model to obtain the prediction results of the fault type and adjust the crushing equipment according to the prediction results.

[0024] The present invention provides a material crushing intelligent detection method and system. The sound data, vibration data and fault types of the crushing equipment during the crushing operation are collected, the time domain features and frequency domain features of the sound data and vibration data are extracted, the time domain features and frequency domain features are spliced ​​into a vector, the spliced ​​vectors of the sound data and the spliced ​​vectors of the vibration data are used as model inputs, and the fault types of the crushing equipment are used as model outputs. Model training is performed, and new sound data and vibration data are input into the trained model to obtain the prediction results of the fault types, and the crushing equipment is adjusted according to the prediction results. The method provided by the present invention can identify the fault type of the crushing equipment based on the feature vectors of the sound data and vibration data, improve the efficiency of fault type detection, reduce the maintenance cost of the crushing device, and improve the accuracy of fault type detection by deeply learning the characteristics of the sound data and vibration data through the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A schematic flow chart of a material crushing intelligent detection method provided by an embodiment of the present invention is shown;

[0027] Figure 2 A schematic diagram of the framework of a material crushing intelligent detection system provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0029] In the description of the embodiments of the present application, words such as "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a concrete way.

[0030] In the description of the embodiments of the present application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the term "plurality" means two or more.

[0031] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprises", "has" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0032] Figure 1 A schematic diagram of a material crushing intelligent detection method provided in an embodiment of this specification is shown, such as Figure 1 As shown, the method comprises the following steps:

[0033] Step S101, collecting sound data, vibration data and fault types of the crushing equipment during the crushing operation, and recording the time of data collection.

[0034] During the pulverizing operation of the pulverizing equipment, sound data can be collected through sound sensors, such as microphones, ensuring that they have sufficient sensitivity and frequency response range. If the microphone output signal is weak, a preamplifier can be used to amplify the signal, or a high-quality sound card can be used to convert the analog signal into a digital signal, and provide better sampling rate and resolution. In addition, vibration data can be collected through vibration sensors, such as by selecting a suitable three-axis accelerometer to measure the vibration of the pulverizing equipment in the three directions of X, Y, and Z. A data acquisition card (DAQ) or a dedicated data acquisition module can be used to convert the analog signal of the accelerometer into a digital signal. Use an industrial PC, laptop or embedded system to enter and store the collected data in real time, and record the corresponding data collection time.

[0035] In one embodiment, the fault types include: the cutter head is stuck, the material does not enter the cutter head, the bearing is faulty, etc.

[0036] During the crushing operation, when the cutter head is stuck, the motor and cutter head will try to continue working, which may cause the sound to suddenly become louder, and abnormal high-frequency sounds may appear. The sound may also change periodically, reflecting that the cutter head tried to rotate but was stuck. The amplitude of the vibration signal will increase significantly because the stuck cutter head will cause greater mechanical stress on the equipment. The frequency of the vibration signal may change and no longer have a regular rotation frequency. There may also be impact signals, reflecting the collision between the cutter head and the stuck material. More harmonic components may appear in the vibration signal, reflecting nonlinear mechanical behavior. When the material does not enter the cutter head, the sound of the equipment during operation may become softer and the frequency of the sound will be relatively stable because there is no material to interfere with the normal rotation of the cutter head. The noise level will be reduced due to the lack of material cutting and crushing. The sound of the cutter head idling may be heard, showing periodic characteristics. The amplitude of the vibration signal will decrease because there is no material entering and the equipment load is reduced. The harmonic components in the vibration signal will decrease, reflecting a smoother mechanical movement. There may be some background noise, but the vibration signal is smoother overall. When a bearing failure occurs, you may hear a high-frequency "squeak" or "rattle" sound, which is caused by bearing wear or poor lubrication. You may hear periodic noise, reflecting the contact between the bearing rolling elements and the inner and outer rings. You may hear irregular noises, reflecting uneven wear inside the bearing. Frequency components unique to bearing failures will appear in the vibration signal, such as the rolling element pass frequency (BPFO), rolling element rotation frequency (BSF), etc. More harmonic components may also appear in the vibration signal, reflecting the nonlinear behavior of the bearing. Impact signals may also appear, reflecting collisions or friction inside the bearing. It is worth noting that this solution is not limited to the above-mentioned faults. The faults of the crushing equipment can be judged according to the actual situation during the actual crushing operation, and they will not be listed one by one here.

[0037] In one implementation, after the sound data and the vibration data are collected, the sound data and the vibration data are first filtered to remove background noise and interference.

[0038] First, you need to clearly identify the type of noise to be removed and the frequency range, as well as the signal components to be retained, and select the appropriate filter type, such as Butterworth, Chebyshev, Elliptic, etc., based on your needs. Set the order, cutoff frequency, or passband frequency range of the filter. Design the filter based on the selected type and parameters. Apply the designed filter to the original sound data and vibration data to generate filtered data. Verify whether the filtered data meets expectations through visualization or statistical analysis, and adjust the filter parameters if necessary.

[0039] Filtering can remove or attenuate background noise, thereby increasing the ratio of useful information in the signal to noise. In pulverizing equipment fault diagnosis, a higher signal-to-noise ratio helps to more accurately identify fault modes and abnormal conditions.

[0040] Step S102 , extracting time domain features of the sound data and the vibration data, and converting the sound data and the vibration data from the time domain to the frequency domain through Fourier transform to extract the frequency domain features of the data.

[0041] The time domain characteristics of sound data and vibration data can be calculated using mathematical methods, such as by adding up the values ​​of all sampling points of the signal and then dividing by the number of sampling points to obtain the average value of the signal. Calculating time domain characteristics by mathematical methods is a general technology and will not be described in detail here. Sound data and vibration data can be converted from the time domain to the frequency domain by the Fourier transform method. After the conversion, frequency domain characteristics can be extracted from the frequency domain data by data calculation. Calculating frequency domain characteristics by mathematical methods is a general technology and will not be described in detail here. The Fourier transform formula is as follows:

[0042]

[0043] Where x(t) is the time domain signal, X(f) is the frequency domain signal, f is the frequency, j is the imaginary unit, and e -j2πft is a complex exponential function.

[0044] In one implementation, the time domain features include: mean, variance, peak, kurtosis, kurtosis, skewness, energy, etc. Other time domain features may also be included, but this embodiment is not limited thereto, and features may be expanded or deleted according to actual conditions.

[0045] In one implementation, the frequency domain features include: spectrum, power spectrum density, harmonic frequency, fundamental frequency, bandwidth, spectrum spread, spectrum attenuation, spectrum peak, etc. Other frequency domain features may also be included, but this embodiment is not limited thereto, and features may be expanded or deleted according to actual conditions.

[0046] In one implementation, after the time domain features and frequency domain features of the sound data and vibration data are extracted, the time domain features and frequency domain features data are first normalized.

[0047] The sound data and vibration data can be normalized by the Min-Max Normalization method, and the formula is as follows:

[0048]

[0049] Among them, x is the original feature data, min(x) is the minimum value in the feature data, and max(x) is the maximum value in the feature data.

[0050] Step S103, concatenating the time domain features and the frequency domain features into a vector, and obtaining a concatenated vector of the sound data and a concatenated vector of the vibration data respectively.

[0051] The normalized time domain features and frequency domain features are concatenated into a vector. Assuming that the time domain feature vector is Ts and the frequency domain feature vector is Fs, the concatenated vector is Vs = [Ts, Fs]. It should be noted that when concatenating vectors, the time domain features and frequency domain features must be aligned in time. Optionally, to ensure that the concatenated vector is not too large to avoid the dimensionality disaster, the number of features can be reduced through feature selection or dimensionality reduction techniques (such as PCA).

[0052] Time domain features and frequency domain features each provide different types of information. Time domain features reflect the characteristics of the signal changing over time, while frequency domain features reveal the components of the signal at different frequencies. By concatenating these two feature sets, a more comprehensive data representation can be obtained. Concatenating the time domain and frequency domain features into a vector can simplify the subsequent data processing and feature processing process. There is no need to process the two sets of features separately, but rather as a whole. By combining time domain and frequency domain features, the model can better cope with different types of data changes and noise interference. Time domain features and frequency domain features can complement each other and improve the robustness of the model under different conditions. The concatenated feature vector can support more complex model structures, such as deep learning models. These models can better utilize the rich feature information and improve the performance of the model.

[0053] Step S104, using the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input, and using the fault type of the crushing equipment as the model output, to perform model training and obtain a trained model.

[0054] The concatenated vector of the sound data and the concatenated vector of the vibration data can be used as the model input, and the type of fault that occurs in the crushing equipment can be used as the model output for model training. Specifically, the data can be divided into a training set, a validation set, and a test set. For example, 70% is used for training, 15% is used for validation, and 15% is used for testing. According to the complexity of the problem and the characteristics of the data, a suitable machine learning model can be selected, such as random forest, convolutional neural network, long short-term memory network, gradient boosting tree, etc. During the training process, a suitable loss function, such as cross entropy loss, is selected, and a suitable optimizer, such as Adam, SGD, etc., is selected, and the number of training iterations is set. After each training cycle, the performance of the model is evaluated using the validation set. Hyperparameters are tuned using methods such as grid search or random search. After training and tuning are completed, the model is finally evaluated using the test set to calculate performance indicators such as accuracy, precision, recall, and F1 score. The trained model is saved for subsequent use.

[0055] In one embodiment, a material crushing intelligent detection model is provided, which includes: an input layer, a CNN module, an LSTM module, a feature fusion layer, a fully connected layer, and an output layer.

[0056] The input layer has two data input interfaces, one of which is used to input the concatenated vector of sound data, and the input interface is connected in series with the CNN module, and the CNN module is used to extract the features of sound data. The other data input interface is used to input the concatenated vector of vibration data, and the input interface is connected in series with the LSTM module, and the LSTM module is used to extract the features of vibration data. The output ends of the CNN module and the LSTM module are connected to the feature fusion layer, and the feature fusion layer, the fully connected layer and the output layer are connected in series.

[0057] Specifically, the CNN module includes a convolutional layer, a pooling layer, a flattening layer, and a fully connected layer. The convolutional layer has a convolution kernel size of 3 and an activation function of ReLU, which is used to extract local features of the sound data. The pooling layer is used to reduce the spatial dimension of the feature map. The flattening layer is used to flatten the output of the convolutional layer so as to connect it with the fully connected layer. The fully connected layer is used to further extract the features of the sound data. The activation function of the fully connected layer is ReLU, and the number of neurons can be selected according to the complexity of the data, such as 128 or 256.

[0058] Furthermore, the LSTM module includes two LSTM layers and one fully connected layer. The two LSTM layers are used to capture the features of vibration data at different levels, and can also capture time dependency and dynamic features. The fully connected layer is used to further extract vibration data features and perform nonlinear transformation.

[0059] Next, the feature fusion layer is used to concatenate the outputs of the CNN module and the LSTM module. The fully connected layer is used to fuse the sound data features extracted from the CNN module with the vibration data features extracted from the LSTM module.

[0060] The model provided in this embodiment can process sound data and vibration data at the same time, and these two types of data provide complementary information. The CNN module is used to process the spectrogram or other two-dimensional representation of sound data and extract local spatial features. The LSTM module is used to process the time series of vibration data to capture time dependency and dynamic features. Nonlinear transformation is performed through feature fusion and fully connected layers, thereby improving the accuracy and robustness of fault detection.

[0061] In one embodiment, the feature fusion layer in the model uses a self-attention mechanism to fuse the outputs of the CNN module and the LSTM module.

[0062] First, we calculate the query, key, and value of the output of the CNN module and the LSTM module respectively. Then we calculate the dot product between the query and the key to get the attention score. We apply the Softmax function to the attention score to get the attention weight. Finally, we use the attention weight to perform weighted summation on the value to get the fused feature.

[0063] For example, suppose the output of CNN is h cnn , the output of LSTM is h lstm , respectively, using the linear transformation W Q , W K , W V Compute queries, keys, and values ​​output by CNN and LSTM modules:

[0064] Q cnn =h cnn W Q K cnn =h cnn W K V cnn =h cnn W V

[0065] Q ltsm =h ltsm W Q K ltsm =h ltsm W K V ltsm =h ltsm W V

[0066] Compute the dot product between the query and the key to get the attention score:

[0067]

[0068]

[0069] where dk is the dimension of the key.

[0070] Apply the Softmax function to the attention score to get the attention weight:

[0071] attention cnn =softmax(score cnn )

[0072] attention ltsm =softmax(score ltsm )

[0073] Use the attention weights to perform weighted summation on the values ​​to obtain the fused features:

[0074] h fusion =attention cnn +attention ltsm

[0075] This embodiment uses a complex feature fusion method to combine the sound data features and the vibration data features. Its attention mechanism can adaptively focus on important features in the sound and vibration data, improve the model's sensitivity to key information, and improve the model's performance and robustness.

[0076] Step S105, collecting new sound data and vibration data, and inputting them into the trained model to obtain the prediction result of the fault type, and adjusting the crushing equipment according to the prediction result.

[0077] The method for collecting new sound data and vibration data is the same as the method for collecting sound data and vibration data in step S101, which will not be repeated here. Before inputting the data into the model, the data needs to be denoised, normalized and vector spliced. The specific method is the same as the method for denoising, normalizing and vector splicing in steps S102 and S103, which will not be repeated here. After the sound data and vibration data are processed as above, they are input into the trained model to obtain the prediction result of the fault type. The model output can be a probability distribution, indicating the probability of different fault types, and the fault type with the highest probability is selected as the prediction result. It can also be a direct output of the fault type, which is not limited here. According to the predicted fault type, take corresponding adjustment measures to adjust the crushing equipment. For example, stop the operation of the crushing equipment immediately, check whether the cutter head is damaged or worn, replace the cutter head if necessary, reduce the feed rate, and avoid jamming caused by too much material. Check whether the feed port is blocked or deformed, and clean or repair the feed port. Appropriately increase the feed rate to ensure that the material can enter the cutter head evenly. Check the properties of the material (such as humidity, viscosity, etc.) to ensure that the material is suitable for the processing capacity of the current equipment. Open the equipment and check the wear of the bearings. Properly lubricate the bearings to reduce friction and wear, etc.

[0078] Corresponding to the above method provided by the present invention, the present invention also provides an intelligent detection system for material crushing. Figure 2 FIG. 2 is a schematic diagram showing a structure of a material crushing intelligent detection system provided in an embodiment of this specification. Figure 2 As shown, the system includes:

[0079] The data collection module S201 is used to collect the sound data, vibration data and fault types of the crushing equipment during the crushing operation, and record the time of data collection.

[0080] The feature extraction module S202 is used to extract the time domain features of the sound data and the vibration data, and convert the sound data and the vibration data from the time domain to the frequency domain through Fourier transform to extract the frequency domain features of the data.

[0081] The vector splicing module S203 is used to splice the time domain features and the frequency domain features into a vector, and obtain a splicing vector of the sound data and a splicing vector of the vibration data respectively.

[0082] The model training module S204 is used to perform model training by taking the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input and taking the fault type of the crushing equipment as the model output.

[0083] The model prediction module S205 is used to input the collected new sound data and vibration data into the trained model to obtain the prediction result of the fault type and adjust the crushing equipment according to the prediction result.

[0084] It is worth mentioning that Figure 2 The system shown is Figure 2 The method embodiments shown correspond to and can be applied to Figure 2 The acquisition of sound data and vibration data, feature extraction, vector splicing, model training and model prediction in the illustrated method embodiment will not be described in detail here. Those skilled in the art should easily appreciate that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0085] According to another embodiment of the present invention, a computing device is provided, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the Figure 2 The method described.

[0086] According to another embodiment, there is also provided a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to execute a combination of Figure 2 The method described.

[0087] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, firmware or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0088] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A material crushing intelligent detection method, characterized in that: The method comprises: Collect the sound data, vibration data and fault types of the crushing equipment during the crushing operation, and record the time of data collection; Extracting time domain features of the sound data and the vibration data, and converting the sound data and the vibration data from the time domain to the frequency domain through Fourier transform to extract the frequency domain features of the sound data and the vibration data; The time domain features and the frequency domain features are concatenated into a vector to obtain a concatenated vector of the sound data and a concatenated vector of the vibration data respectively; Using the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input and the fault type of the crushing equipment as the model output, the model is trained to obtain a trained model; Collect new sound and vibration data and input them into the trained model to obtain the prediction results of the fault type and adjust the crushing equipment according to the prediction results.

2. The method according to claim 1, characterized in that The fault types include: the cutter head is stuck, the material does not enter the cutter head, and the bearing is faulty.

3. The method according to claim 1, characterized in that The time domain features include: mean, variance, peak, kurtosis, kurtosis, skewness, and energy.

4. The method according to claim 1, characterized in that: The frequency domain features include: spectrum, power spectrum density, harmonic frequency, fundamental frequency, bandwidth, spectrum spread, spectrum attenuation, and spectrum peak.

5. The method according to claim 1, characterized in that After the sound data and vibration data are collected, they are first filtered to remove background noise and interference.

6. The method according to claim 1, characterized in that After the time domain features and frequency domain features of the sound data and vibration data are extracted, the time domain features and frequency domain features data are first normalized.

7. The method according to claim 1, characterized in that The model includes: an input layer, a CNN module, an LSTM module, a feature fusion layer, a fully connected layer and an output layer; wherein the CNN module includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer, the convolution kernel size of the convolution layer is 3, the activation function is ReLU, which is used to extract local features of sound data, the pooling layer is used to reduce the spatial dimension of the feature map, the flattening layer is used to flatten the output of the convolution layer so as to connect with the fully connected layer, and the fully connected layer is used to further extract the features of the sound data; the LSTM module includes two LSTM layers and one fully connected layer, the two LSTM layers are used to capture features of different levels of vibration data, and the fully connected layer is used to further extract the features of the vibration data and perform nonlinear transformation; the feature fusion layer is used to splice the outputs of the CNN module and the LSTM module; the fully connected layer is used to fuse the sound data features extracted from the CNN module with the vibration data features extracted from the LSTM module.

8. The method according to claim 7, characterized in that The feature fusion layer in the model uses a self-attention mechanism to fuse the outputs of the CNN module and the LSTM module.

9. A material crushing intelligent detection system, characterized in that: Material crushing detection is performed by the method according to any one of claims 1 to 8, the system comprising: The data collection module is used to collect the sound data, vibration data and fault types of the crushing equipment during the crushing operation, and record the time of data collection; A feature extraction module is used to extract the time domain features of the sound data and the vibration data, and convert the sound data and the vibration data from the time domain to the frequency domain through Fourier transform to extract the frequency domain features of the data; A vector splicing module is used to splice the time domain features and the frequency domain features into a vector, and obtain a splicing vector of the sound data and a splicing vector of the vibration data respectively; A model training module, used for taking the concatenated vector of the sound data and the concatenated vector of the vibration data as the model input and taking the fault type of the crushing equipment as the model output to perform model training; The model prediction module is used to input the collected new sound data and vibration data into the trained model to obtain the prediction results of the fault type and adjust the crushing equipment according to the prediction results.

Images