Mine Microseismic Source Location Using Transfer-Learned P-Wave Picking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Microseismic event location in underground engineering applications suffers from large errors and low accuracy due to complex wave propagation, noise interference, and the need for manual calibration, which is subjective and time-consuming.
Innovation Solution
A massive data-driven method using a deep neural network model with transfer learning to automatically calibrate P-wave arrival times, incorporating a U-net structure, data augmentation, and an optimization algorithm to improve location accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual calibration is used to pick P-wave arrival times, then location accuracy can be improved, but time consumption and labor intensity increase significantly
Solution Approach 1:
The patent replaces manual mechanical calibration with an automated neural network system. The neural network model automatically picks P-wave arrival times by learning from training data, eliminating the need for manual intervention while maintaining high accuracy. This substitution of human expert judgment with an automated intelligent system resolves the contradiction between accuracy and time consumption.
Solution Approach 2:
The patent performs preliminary training of the neural network model using labeled seismic data before actual microseismic event location. The model learns to identify P-wave arrivals from training examples, so that during actual operation, arrival time picking can be performed automatically without manual calibration. This preliminary learning phase enables the system to achieve both high accuracy and efficiency in production use.
2Productivity
If traditional automatic calibration algorithms are used, then processing speed is improved, but accuracy decreases for low SNR and complex waveforms
Solution Approach 1:
The patent transforms the calibration problem from using simple scalar parameters (like amplitude thresholds in traditional methods) to using high-dimensional waveform features processed by a neural network. The model considers multiple characteristics of the waveform simultaneously (amplitude, frequency content, arrival pattern) rather than relying on single-parameter thresholds, enabling accurate picking even for complex, low-SNR signals while maintaining automated processing speed.
Solution Approach 2:
The patent combines multiple waveform characteristics and features into a composite neural network model that processes various signal properties together. Rather than using a single simple algorithm, the system integrates multiple types of information (amplitude envelopes, frequency spectra, arrival patterns) through the neural network architecture, creating a composite intelligent system that outperforms traditional single-method approaches for complex waveforms.
3Measurement precision
If manual calibration is performed repeatedly to account for complex wave velocity structures, then location accuracy is improved, but the complexity and time required for the process increases
Solution Approach 1:
The patent enables the neural network model to automatically adapt to different wave velocity structures and propagation conditions without requiring manual recalibration for each case. The model learns from diverse training data that includes various waveforms and propagation conditions, allowing it to self-adjust and provide accurate arrival time picks across different geological settings. This self-service capability eliminates the need for repeated manual fine-tuning while maintaining accuracy.
Data Source
AI summary
Disclosed is a massive data-driven method for automatically locating a mine microseismic source, including: constructing a microseismic wave calibration data set by using a large-scale seismic data set containing seismic signals and non-seismic signals; constructing a pre-training calibration model based on a full convolution neural network through deep learning of a seismic wave calibration data set; using microseismic data of mine sites for transfer learning of an initial arrival time calibration model to construct an arrival time automatic calibration model suitable for mine microseismic signals; and automatically as well as accurately locating mine microseismic events based on an isokinetic homogeneous isotropic velocity model by using an optimization algorithm to deduce arrival time errors and through repeated iteration and fine-tuning.