Method and device for detecting residual service life of transmission part of construction waste disposal equipment

Abstract

The invention relates to the field of residual service life detection, in particular to a method and device for detecting the residual service life of a transmission part of construction waste disposal equipment. The method comprises the following steps: acquiring data samples of a whole life cycle containing residual service life labels of a plurality of equipment transmission parts; performing data enhancement on the plurality of data samples to obtain a data set; a detection model is constructed, the detection model comprises a multi-scale attention feature extraction module and a predictor, the multi-scale attention feature extraction module comprises a plurality of dense connection blocks with different scales and an attention mechanism, and the predictor is used for predicting the remaining service life with the multi-scale service life state features as input; training the detection model by using the data set to obtain a trained detection model; and detecting a degradation vibration signal of the equipment transmission part by using the trained detection model to obtain the remaining service life. The problems that in the prior art, the residual life detection generalization performance is poor, and the influence of noise interference is large are solved.

Description

Technical Field

[0001] The present invention relates to the field of remaining service life detection, and in particular to a method and device for detecting the remaining service life of transmission components of construction waste disposal equipment. Background Art

[0002] Excavators, loaders, crushers, and other engineering equipment or mechanical equipment are extensively used in construction waste disposal. Load transmission components, such as rolling bearings, are key components of construction waste disposal equipment and are among the most vulnerable to damage because they typically need to withstand large loads and operate continuously. The service life of transmission components affects the use of construction waste disposal equipment and seriously impacts normal production progress. Real-time status monitoring of transmission components and using this information to assess the equipment's operating status and degree of degradation can more accurately determine the equipment's remaining service life. This helps to perform timely repairs before failures occur and arrange for maintenance personnel to perform necessary maintenance, thereby maximizing equipment life, significantly reducing downtime, and lowering the manpower and material costs associated with excessive maintenance.

[0003] Current methods for detecting the remaining useful life of transmission components are mainly divided into model-based methods and data-driven methods. Model-based methods simulate the degradation process of transmission components by establishing physical or mathematical models, but are limited by the complexity of the model and have great limitations under complex working conditions. Data-driven methods extract degradation characteristics by analyzing a large amount of historical data of transmission components, thereby revealing the relevant mapping relationship between monitoring data and remaining useful life. With the rapid development of artificial intelligence big data methods in recent years, especially the outstanding capabilities of neural networks in nonlinear fitting and big data processing, deep learning methods can automatically extract data features and determine the mapping relationship between data without human intervention or prior knowledge. Therefore, the research on remaining useful life prediction using deep learning has become a hot field in academia and industry.

[0004] While deep learning methods have performed well in detecting the remaining life of transmission components, in real industrial scenarios, transmission components operate under a wide range of conditions. Deep learning models trained under one operating condition have poor generalization capabilities and severe performance degradation when applied to different operating conditions. Furthermore, when significantly affected by ambient noise, prediction accuracy is low. Therefore, a method for detecting the remaining service life of transmission components in construction waste disposal equipment with strong generalization capabilities and noise immunity is needed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the existing technology and provide a method for detecting the remaining service life of the transmission components of construction waste disposal equipment. It solves the problems of poor generalization performance of remaining life detection and significant influence of noise interference in the existing technology, and can realize accurate prediction of the remaining service life of the transmission components of construction waste disposal equipment.

[0006] In order to solve the above technical problems, the technical solution of the present invention is: a method for detecting the remaining service life of transmission components of construction waste disposal equipment, comprising: S1, obtaining data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; S2, perform data enhancement on multiple data samples to obtain a data set; S3: Build a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using the multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighted fusion of the life state features output by all densely connected blocks. S4, using the data set to train the detection model to obtain a trained detection model; S5, using the trained detection model to detect the degradation vibration signal of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

[0007] Furthermore, S1 includes: S11, collecting full life cycle vibration signals of transmission components of multiple construction waste disposal equipment under different working conditions; wherein the same speed and load are considered the same working condition; S12, dividing the full life cycle vibration signal into data samples of the same length, marking the remaining service life labels, and normalizing them; S13, dividing the data samples into a training sample set and a test sample set according to different working conditions.

[0008] Furthermore, in S2, data samples are enhanced, including adaptive scaling and transformation, and the formula is: ; Where, represents the i-th data sample, are learnable affine parameters, and Corresponding to the input data samples The mean and variance of .

[0009] Furthermore, in S3, each densely connected block includes several convolutional attention layers and an average pooling layer; The input of the Kth convolutional attention layer is the input feature of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension; Output features of the Kth convolutional attention layer The calculation formula is: ; Where, Represents the input features of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension, is the K-th layer nonlinear transformation function, which includes BN, ReLU, Conv and attention layers; The average pooling layer is used to flatten the output features of the last convolutional attention layer into a one-dimensional vector.

[0010] Furthermore, the calculation method of the attention layer is: First, perform maximum pooling on each channel of the input feature F to obtain the pooling vector; Then, the pooling vector is convolved in one dimension and activated by the Sigmoid function to obtain the channel weight vector , the formula is: ; Where σ is the Sigmoid activation function, is a convolution operation with a kernel size of k, represents maximum pooling; Finally, the channel weight vector Multiply it with the feature F to get the attention-weighted feature, which is used as the output of the attention layer.

[0011] Furthermore, in S3, the formula for adaptive weighted fusion of features output by several densely connected blocks is: ; Where, is the life state feature extracted from the 1st to the mth scale, are learnable weight parameters with the same initial value, normalized by the Softmax function, and their sum of probabilities is 1. It is a multi-scale lifetime state characteristic.

[0012] Furthermore, in S3, the predictor includes several fully connected layers.

[0013] Furthermore, in S4, during the training process, the root mean square damage function (MSE) is minimized to optimize the model: ; Where, Indicates the true value of the remaining useful life, Represents the predicted value of remaining useful life, and N represents the number of input samples.

[0014] The present invention also provides a device for detecting the remaining service life of a transmission component of a construction waste disposal equipment, comprising: A data sample acquisition module is used to obtain data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; The data enhancement module is used to perform data enhancement on multiple data samples to obtain a data set; A detection model construction module is used to construct a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighting the life state features output by all densely connected blocks. The detection model training module is used to train the detection model using the data set to obtain a trained detection model; The detection module is used to use the trained detection model to detect the degraded vibration signals of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

[0015] The present invention also provides a device for detecting the remaining service life of transmission parts of construction waste disposal equipment, including a device for detecting the remaining service life of transmission parts of construction waste disposal equipment; or including a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the computer program, a method for detecting the remaining service life of transmission parts of construction waste disposal equipment is implemented.

[0016] After adopting the above technical solution, the present invention overcomes the singleness of samples in the training sample set through data enhancement, obtains samples with diversified distribution for training the detection model, and obtains the life state characteristics of samples at different scales by establishing a multi-scale attention feature extraction module. The attention mechanism is integrated into the densely connected block to enhance the model's learning ability of state information and reduce noise interference to focus on key state information, which significantly improves the generalization ability and prediction accuracy of the model, and can be effectively applied to the remaining service life detection of transmission components of construction waste disposal equipment under different working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for detecting the remaining service life of a transmission component of construction waste disposal equipment according to the present invention; Figure 2 Schematic diagram of the training process of the detection model of the present invention; Figure 3 Schematic diagram of the structure of the dense connection block of the present invention; Figure 4 Schematic diagram of the detection process of the detection model of the present invention; Figure 5 The remaining useful life prediction result diagram of the detection model on the test sample set; Figure 6 It is a structural schematic diagram of the device for detecting the remaining service life of transmission components of construction waste disposal equipment according to the present invention. DETAILED DESCRIPTION

[0018] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments in conjunction with the accompanying drawings.

[0019] Example 1: Figures 1 to 4 As shown, a method for detecting the remaining service life of a transmission component of construction waste disposal equipment comprises: S1, obtaining data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; S2, perform data enhancement on multiple data samples to obtain a data set; S3: Build a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using the multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighted fusion of the life state features output by all densely connected blocks. S4, using the data set to train the detection model to obtain a trained detection model; S5, using the trained detection model to detect the degradation vibration signal of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

[0020] Specifically, this detection method overcomes the monotony of samples in the training sample set through data enhancement, obtains samples with diverse distributions for training the detection model, and obtains the life state characteristics of samples at different scales by establishing a multi-scale attention feature extraction module. The attention mechanism is integrated into the densely connected block to enhance the model's learning ability of state information and reduce noise interference to focus on key state information. It significantly improves the generalization ability and prediction accuracy of the model, and can be effectively applied to the remaining service life detection of transmission components of construction waste disposal equipment under different working conditions.

[0021] In this embodiment, S1 includes: S11, collecting full life cycle vibration signals of transmission components of multiple construction waste disposal equipment under different working conditions; wherein the same speed and load are considered the same working condition; S12, dividing the full life cycle vibration signal into data samples of the same length, marking the remaining service life labels, and normalizing them; S13, dividing the data samples into a training sample set and a test sample set according to different working conditions; wherein, the remaining useful life labels contained in the training sample set can be used for model training, and the remaining useful life labels contained in the test sample set do not participate in model training, but are only used to verify the accuracy of the remaining useful life predicted by the model for the test sample set.

[0022] As an example, the publicly available XJTU-SY rolling bearing accelerated life test dataset is used. Rolling bearings are one of the key transmission components of equipment. The sampling frequency in this experiment was set to 25.6kHz, the sampling interval was 1 minute, and each sampling duration was 1.28 seconds. Bearing2_1, Bearing2_2, Bearing2_3, Bearing2_4, and Bearing2_5 under the second operating condition (speed of 2250 r / min, radial force of 11 kN) were used as the training sample set to train the constructed detection model. Bearing1_3 under the first operating condition (speed of 2100 r / min, radial force of 12 kN) was used as the test sample set to verify the constructed life detection model. A description of the dataset used is shown in Table 1: .

[0023] It should be noted that the actual life here refers to the actual service life of the bearing in the experiment, rather than the remaining service life value of a single sample.

[0024] For example, Bearing1_3 has a total of 158 samples, and its actual lifespan is 2 hours and 38 minutes, or 158 minutes. Samples are collected every minute, for a total of 158 samples collected over the 158-minute run. This means that the first sample collected has a remaining lifespan of 158 minutes, while the last sample has a remaining lifespan of 1 minute. The corresponding labels are available in the public dataset. In this embodiment, in S2, data enhancement is performed on the data samples, including adaptive scaling and conversion, to obtain diversified samples as input for model training. The formula is: ; Where, represents the i-th data sample, are learnable affine parameters, and Corresponding to the input data samples The mean and variance of .

[0025] It should be noted that for the training sample set, the probability of each sample being data augmented is 0.5 to ensure that there are a sufficient number of original samples and augmented samples to train the model.

[0026] In this embodiment, if Figure 3 As shown in Figure 3, each densely connected block in S3 includes several convolutional attention layers and an average pooling layer; The input of the Kth convolutional attention layer is the input feature of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension; Output features of the Kth convolutional attention layer The calculation formula is: ; Where, Represents the input features of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension, is the K-th layer nonlinear transformation function, which includes BN, ReLU, Conv and attention layers; The average pooling layer is used to flatten the output features of the last convolutional attention layer into a one-dimensional vector.

[0027] Specifically, the feature transfer method of the dense connection block is to directly connect the features of all previous layers in the channel dimension and input them into the next layer, so that the features can be transferred more effectively to enhance the model's ability to learn state information.

[0028] In this embodiment, the calculation method of the attention layer is: First, perform maximum pooling on each channel of the input feature F to obtain the pooling vector; Then, the pooling vector is convolved in one dimension and activated by the Sigmoid function to obtain the channel weight vector , the formula is: ; Where σ is the Sigmoid activation function, is a convolution operation with a kernel size of k, represents maximum pooling; Finally, the channel weight vector Multiply it with the feature F to get the attention-weighted feature, which is used as the output of the attention layer.

[0029] Specifically, the attention layer is used to assign higher weights to key state information, reducing the interference of noise in vibration signal samples and focusing on more important life state features.

[0030] In this embodiment, if Figure 2 As shown in Figure 3, the formula for adaptive weighted fusion of features output by several densely connected blocks in S3 is: ; Where, is the life state feature extracted from the 1st to the mth scale, are learnable weight parameters with the same initial value, normalized by the Softmax function, and their sum of probabilities is 1. It is a multi-scale lifetime state characteristic.

[0031] Specifically, a multi-scale attention feature extraction module is used to extract rich state information at multiple scales, alleviating the problem of difficulty in extracting state information caused by changes in working conditions.

[0032] As an example, Figure 2 As shown in the figure, there are three dense connection blocks. Convolution kernels of different sizes are used in the dense connection blocks at three scales. The convolution kernel sizes at three scales are 1×11, 1×15 and 1×19 respectively. The last layer of the dense connection block is connected to an average pooling layer to flatten the extracted features into a one-dimensional vector. Then, the one-dimensional vector features at different scales are adaptively weighted and fused to dynamically measure the importance of state features at different scales, and finally obtain multi-scale life state features. Among them, the weight parameter Added to the parameter list of the detection model, it can be trained and optimized in the back-propagation optimization, so, Can be a learnable parameter.

[0033] In this embodiment, in S3, the predictor includes several fully connected layers.

[0034] As an example, the predictor includes two fully connected layers, and the number of nodes in the two fully connected layers is 512 and 1 respectively.

[0035] In this embodiment, in S4, during the training process, the MES optimization model is adopted to minimize the root mean square damage function: ; Where, Indicates the true value of the remaining useful life, Represents the predicted value of remaining useful life, and N represents the number of input samples.

[0036] In this embodiment, the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) are used as performance evaluation indicators of the model. RMSE, MAE, and R2 are only used to evaluate the performance of the model and are not involved in the model parameter update: ; ; ; Where, for The average value of .

[0037] In this embodiment, the number of iterative training times of the detection model is 200, and the adaptive moment estimation algorithm is used as the optimization algorithm. Figure 2 The training model shown is back-propagation optimized, and the training stops when the number of iterative training reaches the set value.

[0038] In this embodiment, the model after training is as follows Figure 4 As shown, it includes a multi-scale attention feature extraction module and a predictor.

[0039] The results of the accuracy evaluation of the remaining useful life detection of the test sample set by the method involved in the above embodiment are shown in Table 2: .

[0040] Among them, the root mean square error (RMSE) is more sensitive to large deviations, and the mean absolute error (MAE) provides an intuitive error metric. The smaller its value, the smaller the prediction error of the model. The coefficient of determination (R 2 ) reflects the degree of fit of the model to the data. The higher its value, the better the fit of the model to the data.

[0041] It can be seen that this method uses the data set under the second working condition as the training sample set and the data set under the first working condition as the test sample set. The RMSE of the verification result is 0.0351, and the MAE of the verification result is 0.0289, which reflects that the difference between the remaining service life prediction value and the true value is relatively small. 2 The result is 0.9852, which also shows that the model of this method has a high degree of fit to the data and the remaining service life prediction results of the test sample set are relatively accurate.

[0042] The remaining service life detection results of the test sample set by the method involved in the above embodiment are as follows: Figure 5 As shown, it can be seen that the predicted remaining service life has the same trend as the actual remaining service life, and the degree of fit is also high, which also reflects that the method of the present invention has good generalization ability and can achieve accurate prediction of the remaining service life of the equipment.

[0043] Example 2: Figure 6 As shown, a device for detecting the remaining service life of a transmission component of a construction waste disposal equipment comprises: A data sample acquisition module is used to obtain data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; The data enhancement module is used to perform data enhancement on multiple data samples to obtain a data set; A detection model construction module is used to construct a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighting the life state features output by all densely connected blocks. The detection model training module is used to train the detection model using the data set to obtain a trained detection model; The detection module is used to use the trained detection model to detect the degraded vibration signals of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

[0044] Embodiment 3: A device for detecting the remaining service life of transmission parts of construction waste disposal equipment, comprising the device for detecting the remaining service life of transmission parts of construction waste disposal equipment in embodiment 2; or comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting the remaining service life of transmission parts of construction waste disposal equipment as described in embodiment 1 is implemented.

[0045] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the contents of the specification and must be determined according to the scope of the claims.

Claims

1. A method for detecting the remaining service life of transmission components of construction waste disposal equipment, characterized in that: include: S1, obtaining data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; S2, perform data enhancement on multiple data samples to obtain a data set; S3: Build a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using the multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighted fusion of the life state features output by all densely connected blocks. S4, using the data set to train the detection model to obtain a trained detection model; S5, using the trained detection model to detect the degradation vibration signal of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

2. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: S1 includes: S11, collecting full life cycle vibration signals of transmission components of multiple construction waste disposal equipment under different working conditions; wherein the same speed and load are considered the same working condition; S12, dividing the full life cycle vibration signal into data samples of the same length, marking the remaining service life labels, and normalizing them; S13, dividing the data samples into a training sample set and a test sample set according to different working conditions.

3. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: In S2, data samples are enhanced, including adaptive scaling and transformation. The formula is: ； Where, represents the i-th data sample, are learnable affine parameters, and Corresponding to the input data samples The mean and variance of .

4. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: In S3, each densely connected block includes several convolutional attention layers and an average pooling layer; The input of the Kth convolutional attention layer is the input feature of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension; Output features of the Kth convolutional attention layer The calculation formula is: ； Where, Represents the input features of the densely connected block And the output features of the first 1st to K-1th convolutional attention layers Connections in the channel dimension, is the K-th layer nonlinear transformation function, which includes BN, ReLU, Conv and attention layers; The average pooling layer is used to flatten the output features of the last convolutional attention layer into a one-dimensional vector.

5. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 4, characterized in that: The calculation method of the attention layer is: First, perform maximum pooling on each channel of the input feature F to obtain the pooling vector; Then, the pooling vector is convolved in one dimension and activated by the Sigmoid function to obtain the channel weight vector , the formula is: ； Where σ is the Sigmoid activation function, is a convolution operation with a kernel size of k, represents maximum pooling; Finally, the channel weight vector Multiply it with the feature F to get the attention-weighted feature, which is used as the output of the attention layer.

6. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: In S3, the formula for adaptive weighted fusion of features output by several densely connected blocks is: ； Where, is the life state feature extracted from the 1st to the mth scale, are learnable weight parameters with the same initial value, normalized by the Softmax function, and their sum of probabilities is 1. It is a multi-scale lifetime state characteristic.

7. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: In S3, the predictor consists of several fully connected layers.

8. The method for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 1, characterized in that: In S4, during the training process, the MES optimization model is optimized by minimizing the root mean square damage function: ； Where, Indicates the true value of the remaining useful life, Represents the predicted value of remaining useful life, and N represents the number of input samples.

9. A device for detecting the remaining service life of transmission components of construction waste disposal equipment, characterized in that: include: A data sample acquisition module is used to obtain data samples of the entire life cycle of transmission components of multiple construction waste disposal equipment, including remaining service life labels; The data enhancement module is used to perform data enhancement on multiple data samples to obtain a data set; A detection model construction module is used to construct a detection model. The detection model includes a multi-scale attention feature extraction module and a predictor. The multi-scale attention feature extraction module includes several densely connected blocks at different scales that are integrated with the attention mechanism. The predictor is used to predict the remaining useful life using multi-scale life state features as input. The multi-scale life state features are obtained by adaptively weighting the life state features output by all densely connected blocks. The detection model training module is used to train the detection model using the data set to obtain a trained detection model; The detection module is used to use the trained detection model to detect the degraded vibration signals of the transmission parts of the construction waste disposal equipment to be detected, and obtain the remaining service life.

10. A device for detecting the remaining service life of transmission components of construction waste disposal equipment, characterized in that: The device for detecting the remaining service life of transmission components of construction waste disposal equipment according to claim 9 is included; Or it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the method for detecting the remaining service life of the transmission components of the construction waste disposal equipment as described in any one of claims 1 to 8.

Images