Blasting vibration monitoring data processing method based on neural network and kernel ridge regression
By combining BP neural networks and kernel ridge regression models, the blasting vibration monitoring data is corrected, solving the data distortion problem when the blasting vibration meter is installed in a fixed position. This achieves long-term online monitoring and data accuracy, making it suitable for mining environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEASTERN UNIV CHINA
- Filing Date
- 2024-08-15
- Publication Date
- 2026-07-21
AI Technical Summary
When existing blasting vibration meters are installed in a fixed position, the direction of the vibration velocity sensor cannot be adjusted as the blast source moves, resulting in test data that does not comply with the "Blasting Safety Regulations" and making long-term online monitoring impossible.
By combining BP neural networks and kernel ridge regression models, the blasting vibration waveform is corrected using a training dataset to construct a BP-KRR model, which corrects data distortion caused by changes in sensor orientation and improves the accuracy and flexibility of monitoring data.
It enables long-term online monitoring at the mine site, improves the flexibility and data accuracy of the blasting vibration meter, and is suitable for complex mining environments.
Smart Images
Figure CN119089406B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of engineering blasting technology and relates to a method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression. Background Technology
[0002] In metal mining, rock drilling and blasting are the primary methods of rock breaking. Therefore, frequent blasting is necessary in daily mine operations. However, the disturbance caused by frequent blasting can induce cumulative damage to open-pit slopes, underground pillars, and the roof. This damage can further develop into landslides, roof collapses, and goaf collapses, affecting the safe operation of the mine. To assess this impact, the "Safety Regulations for Blasting" GB6722-2014 specifies the methods for vibration measurement and data processing during blasting.
[0003] Conventional blasting vibration meters have two core components: a blasting vibration recorder and a vibration velocity sensor. The vibration velocity sensor has three channels: X, Y, and Z. In daily use, the X-axis of the vibration velocity sensor is aligned with the blasting vibration source as a reference for monitoring vibration, and the waveform received in the X-axis direction is used for data analysis. Currently, when using this equipment for monitoring in mines, the equipment often moves with the blast hole, but this vibration measurement method requires the testing personnel to remain on-site in the mine, which is not feasible for long-term monitoring. To monitor the impact of frequent blasting vibrations on slopes, pillars, etc., long-term online monitoring equipment is needed. Online monitoring requires the blasting vibration meter to be installed in a fixed position, and the direction of the vibration velocity sensor cannot be adjusted with the movement of the blast source. This results in the test data not conforming to the "Blasting Safety Regulations" and cannot be directly used for blasting vibration analysis. Therefore, a data processing method is needed to correct the test data.
[0004] Current technology for monitoring vibration at blasting sites primarily utilizes blasting vibration meters, which consist of two main components: a blasting vibration recorder and a vibration velocity sensor. To address the limitation of blasting vibration meters, which require the vibration velocity sensor to be mounted in a fixed position and cannot adjust its direction with the movement of the blast source, a technology for remotely controlling the vibration velocity sensor has been proposed. This technology aims to achieve remote control by modifying the sensor's structure, adding a support plate and a movable base to the bottom of the vibration velocity sensor, and mounting a remote drive on the support plate. However, this technology has the following drawbacks:
[0005] 1. Modifying equipment will not only increase costs, but will also make it impossible to analyze the errors generated by the equipment;
[0006] 2. Although the goal of remote control has been achieved, personnel still need to operate the equipment for each blast, which cannot meet the requirement of long-term online monitoring. Summary of the Invention
[0007] To address the shortcomings of existing technologies, this invention tackles the problem of inaccurate test results caused by the movement of the blast source in online monitoring of blasting vibrations. It proposes a data processing method for online monitoring of blasting vibrations based on BP neural networks and BP-KRR (Backpropagation-Based Regression) to correct online monitoring data of blasting vibrations in mines. This method generates a large amount of vibration data through laboratory impact tests, which serves as the training dataset for the model. The model parameters are corrected, and the blasting vibration waveform data is trained and corrected using an artificial neural network BP model and a BP-KRR model. This results in a correction model for data distortion caused by changes in sensor orientation. This model improves the flexibility of blasting vibration meters in field use and significantly enhances their practical application capabilities.
[0008] To achieve the above objectives, this invention proposes a data processing method for blasting vibration monitoring based on neural networks and kernel ridge regression, the steps of which are as follows:
[0009] Step 1: Obtain vibration waveforms under different vibration conditions and construct a dataset;
[0010] Specifically, the process involves: First, identifying several seismic sources and setting different monitoring radii for each source. Then, arranging monitoring points with the same radius but different angles around each source. At each monitoring point, vibration velocity sensors are used to obtain vibration waveforms at different angles for each seismic source, and these waveforms are automatically saved in a blasting vibration recorder, resulting in a dataset containing several vibration waveforms. The monitoring radius is the distance from the seismic source to the monitoring point. Each sample in the dataset represents a vibration waveform and its corresponding distance from the seismic source to the monitoring point, angle, and energy of the seismic source.
[0011] Step 2: Export the vibration waveform from the blasting vibration recorder and preprocess the vibration waveform to obtain the preprocessed dataset;
[0012] The preprocessing specifically involves: performing spectral analysis on the vibration waveform using Fourier Transform (FFT) to extract the dominant frequency and peak vibration velocity of the vibration waveform at each monitoring point; each sample in the preprocessed dataset represents a vibration waveform and its corresponding distance from the source to the monitoring point, angle, energy of the source, dominant frequency, and peak vibration velocity of the vibration waveform;
[0013] Step 3: Divide the preprocessed dataset into training and test sets according to a set ratio;
[0014] Step 4: Construct the BP-KRR model; the BP-KRR model includes a BP neural network model and a kernel ridge regression KRR model connected in sequence;
[0015] Step 5: Train the BP-KRR model using the training set to obtain the trained BP-KRR model;
[0016] Step 5.1: Input the training set into the BP neural network model to train the model and obtain the predicted main frequency and vibration velocity output by the BP neural network model;
[0017] Step 5.2: Import the predicted dominant frequency and vibration velocity into the kernel ridge regression (KRR) model for training, correct the predicted dominant frequency and vibration velocity, and obtain the BP-KRR correction value;
[0018] Step 5.3: Repeat steps 5.1 to 5.2 for a set number of training rounds to obtain the trained BP-KRR model;
[0019] Step 6: Evaluate and adjust the trained BP-KRR model using the test set to obtain the final BP-KRR model;
[0020] Specifically: Input the test set into the trained BP-KRR model, output the BP-KRR correction value, calculate the BP-KRR correction value accuracy, adjust the iteration function of the BP-KRR model according to the accuracy result, until the accuracy result reaches the set threshold, and obtain the final BP-KRR model;
[0021] Step 7: Process the subsequent monitoring data using the final BP-KRR model to obtain accurate long-term online monitoring data; the long-term online monitoring data includes vibration velocity and frequency;
[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0023] To address the data distortion caused by the movement of the blast source in long-term online monitoring, a data processing method for blasting vibration monitoring based on neural networks and kernel ridge regression is proposed. This method has the advantages of BP neural networks and kernel ridge regression prediction models, has low requirements for the quality of the initial acquired data, can be applied to the complex testing environment of mines, and can correct long-term online monitoring data of blasting vibration measurement systems with high accuracy. Attached Figure Description
[0024] Figure 1 Vibration velocity sensor arrangement diagram in an embodiment of the present invention;
[0025] Figure 2 BP neural network structure diagram in this embodiment of the invention;
[0026] Figure 3 In this embodiment of the invention, a BP neural network and a kernel ridge regression prediction model are used.
[0027] Figure 4A comparison chart of BP+KRR predicted values and actual values in an embodiment of the present invention;
[0028] (a) is a comparison diagram of vibration velocity models; (b) is a comparison diagram of main frequency models. Detailed Implementation
[0029] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0030] This invention proposes a data processing method for blasting vibration monitoring based on neural networks and kernel ridge regression to correct the distortion problem of monitoring data. The steps are as follows:
[0031] Step 1: Obtain vibration waveforms under different vibration conditions and construct a dataset;
[0032] Specifically, the process involves: First, identifying several seismic sources and setting different monitoring radii for each source. Then, arranging monitoring points with the same radius but different angles around each source. At each monitoring point, vibration velocity sensors are used to obtain vibration waveforms at different angles for each seismic source, and these waveforms are automatically saved in a blasting vibration recorder, resulting in a dataset containing several vibration waveforms. The monitoring radius is the distance from the seismic source to the monitoring point. Each sample in the dataset represents a vibration waveform and its corresponding distance from the seismic source to the monitoring point, angle, and energy of the seismic source.
[0033] In this embodiment, eight identical blasting vibration meters, a standard measuring tape, an angle measuring device, and a 4.14 kg weight are prepared; according to... Figure 1 The vibration velocity sensors are arranged as shown. A weight is placed 1m above the ground and dropped freely, generating vibration waves upon impact. A circle with a monitoring radius of 1m is drawn around the point of impact, and eight blasting vibration meters are positioned at angles of 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°, respectively, and numbered 1-8. The weight is dropped from a height of 1m to 0.1m, with intervals of 0.1m. The weight is dropped three times at each height interval, and vibration data are recorded. In the experiment, the x-axis of the vibration velocity sensor is required to point towards the center of vibration, and the vibration data is used to verify the uniformity of the site.
[0034] After verifying the uniformity of the site, the angle between the x-axis and the center of vibration of the eight vibration velocity sensors was changed. The changed angles were 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°.
[0035] With the angles of the eight vibration velocity sensors remaining constant, the distance between the vibration velocity sensors and the point where the hammer lands is changed. Figure 1The monitoring radius was set to 11 concentric circles with monitoring radii of 3m, 2.8m, 2.6m, 2.4m, 2.2m, 2.0m, 1.8m, 1.6m, 1.4m, 1.2m, and 1.0m respectively. The data collected from each concentric circle with a different monitoring radius was recorded as a group, for a total of 11 groups of data.
[0036] According to the monitoring radius set above, the falling height of the hammer is changed at each monitoring radius, from 2m to 1m, with an interval of 0.1m. The hammer is dropped 3 times at each height interval, and the vibration data is recorded.
[0037] Step 2: Perform waveform analysis preprocessing on the collected data. After filtering the valid vibration waveform data, perform spectrum analysis on the waveforms using Fourier transform (FFT) to extract the dominant frequency and peak vibration velocity of the vibration waveform at each monitoring point.
[0038] Step 3: Divide the preprocessed dataset into training and test sets according to a ratio of 80% and 20% respectively. The divided datasets are used to combine the algorithm to prepare for subsequent training of BP neural network models, BP-KRR models and model accuracy evaluation.
[0039] Step 4: Construct the BP-KRR model; the BP-KRR model includes a BP neural network model and a kernel ridge regression KRR model connected in sequence;
[0040] Step 5: Train the BP-KRR model using the training set to obtain the trained BP-KRR model;
[0041] Step 5.1: Input the training set into the BP neural network model to train the model and obtain the predicted main frequency and vibration velocity output by the BP neural network model;
[0042] like Figure 2 The impact energy generated by the free fall of the hammer at different heights is used as input x1 of the input layer. The angle between the x-axis of the vibration velocity sensor and the center of vibration is set as input x2 of the input layer, the distance between the landing center and the sensor is set as input x3 of the input layer, and the main frequency of the waveform and the peak vibration velocity are set as input x4 and x5 of the input layer, respectively. The BP neural network has two hidden layers, and the output layer outputs the main frequency y1 and the vibration velocity y2.
[0043] Step 5.2: As Figure 3 As shown, the predicted dominant frequency and vibration velocity are imported into the Kernel Ridge Regression (KRR) model for training, and the predicted dominant frequency and vibration velocity are corrected to obtain the BP-KRR correction value.
[0044] Step 5.3: Repeat steps 5.1 to 5.2 for a set number of training rounds to obtain the trained BP-KRR model;
[0045] Step 6: Evaluate and adjust the trained BP-KRR model using the test set to obtain the final BP-KRR model;
[0046] In this embodiment, 80% of the preprocessed data is used as the training set and imported into the BP neural network model for model training. The model parameters are adjusted to obtain the predicted value with the smallest error from the actual value. The obtained predicted value is then imported into the kernel ridge regression model for model training again. The prediction accuracy is calculated using the test set to evaluate the performance of the prediction model. Based on the evaluation results, the model structure of the algorithm is adjusted until the accuracy reaches more than 90%.
[0047] Step 7: Use the final BP-KRR model to process the subsequent monitoring data to obtain accurate long-term online monitoring data;
[0048] like Figure 4 As shown, by comparing the predicted values of the BP neural network, the predicted values of BP-KRR, and the actual values, the predicted values of BP-KRR in this invention are closer to the actual values.
Claims
1. A method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression, characterized in that, Includes the following steps: Step 1: Obtain vibration waveforms under different vibration conditions and construct a dataset; Step 2: Export the vibration waveform from the blasting vibration recorder and preprocess the vibration waveform to obtain the preprocessed dataset; Step 3: Divide the preprocessed dataset into training and test sets according to a set ratio; Step 4: Construct the BP-KRR model; the BP-KRR model includes a BP neural network model and a kernel ridge regression KRR model connected in sequence; Step 5: Train the BP-KRR model using the training set to obtain the trained BP-KRR model; Step 5 specifically includes: Step 5.1: Input the training set into the BP neural network model to train the model and obtain the output of the BP neural network model. Predicted dominant frequency and vibration velocity; Step 5.2: Import the predicted dominant frequency and vibration velocity into the kernel ridge regression (KRR) model for training, correct the predicted dominant frequency and vibration velocity, and obtain the BP-KRR correction value; Step 5.3: Repeat steps 5.1 to 5.2 for a set number of training rounds to obtain the trained BP-KRR model; Step 6: Evaluate and adjust the trained BP-KRR model using the test set to obtain the final BP-KRR model; Step 7: Use the final BP-KRR model to process the subsequent monitoring data to obtain accurate long-term online monitoring data.
2. The method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression according to claim 1, characterized in that, Step 1 is as follows: First, determine several seismic sources and set different monitoring radii for each seismic source. Then, arrange monitoring points with the same monitoring radius but different angles around each seismic source. At each monitoring point, use a vibration velocity sensor to obtain the vibration waveforms of the monitoring points at different angles under each seismic source condition, and automatically save them in the blasting vibration recorder to obtain a dataset including several vibration waveforms. The monitoring radius is the distance from the seismic source to the monitoring point.
3. The method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression according to claim 1, characterized in that, Each sample in the dataset described in step 1 is a vibration waveform and its corresponding distance from the source to the monitoring point, angle, and energy of the source.
4. The method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression according to claim 1, characterized in that, The preprocessing in step 2 specifically involves: performing spectral analysis on the vibration waveform using Fourier transform (FFT) to extract the dominant frequency and peak vibration velocity of the vibration waveform at each monitoring point; each sample in the preprocessed dataset is a vibration waveform and its corresponding distance from the source to the monitoring point, angle, energy of the source, dominant frequency, and peak vibration velocity of the vibration waveform.
5. The method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression according to claim 1, characterized in that, Step 6 specifically involves: inputting the test set into the trained BP-KRR model, outputting the BP-KRR correction value, calculating the BP-KRR correction value accuracy, adjusting the iteration function of the BP-KRR model based on the accuracy result, until the accuracy result reaches the set threshold, and obtaining the final BP-KRR model.
6. The method for processing blasting vibration monitoring data based on neural networks and kernel ridge regression according to claim 1, characterized in that, The long-term online monitoring data mentioned in step 7 includes vibration velocity and frequency.