TBM Jamming Prediction via CNN-Transformer and Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for predicting tunnel boring machine (TBM) jamming face challenges such as limited precision and scope in geological exploration, complexity and high computational demands in numerical simulations, and instability in machine learning-based predictions, leading to inaccurate and delayed jamming predictions.
Innovation Solution
A method using a deep neural network combined with numerical simulation, specifically constructing a jamming numerical sample library through numerical simulation and establishing a prediction model using Convolutional Neural Networks (CNN) and Transformers for real-time monitoring and early warning of TBM jamming.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If numerical simulation is used to simulate the interaction between TBM and surrounding rock, then the prediction accuracy of TBM jamming is improved, but the computational complexity and calculation time increase significantly
Solution Approach 1:
The patent pre-generates a comprehensive sample library through numerical simulation before actual TBM operation. The sample library contains pre-calculated stress changes, deformation features, and stability data for various geological conditions. During real-time operation, the system only needs to query and compare this pre-computed library rather than performing complex simulations, thus achieving high prediction accuracy without real-time computational burden.
Solution Approach 2:
The patent creates a simplified copy of the complex numerical simulation results in the form of a lookup table/sample library. Instead of replicating the full simulation process during operation, the system uses pre-computed representative data that captures the essential relationships between geological parameters and TBM jamming risk, significantly reducing computational requirements while maintaining prediction accuracy.
2Speed
If machine learning algorithms are used for real-time prediction, then the real-time performance is improved, but the prediction stability deteriorates due to limited monitoring samples and algorithm limitations
Solution Approach 1:
The patent merges the advantages of numerical simulation (high accuracy) with machine learning (real-time performance) by using simulation data to train robust prediction models. The sample library generated from numerical simulation provides comprehensive training data that overcomes the limitation of few real monitoring samples, enabling the machine learning model to achieve both real-time performance and stable, reliable predictions.
Solution Approach 2:
The system performs preliminary training of machine learning algorithms using extensively generated numerical simulation data before real-time operation. This pre-training phase allows the algorithm to learn from a comprehensive dataset covering various geological conditions, improving its stability and reliability when deployed for real-time prediction with limited actual monitoring samples.
3Loss of information
If geological exploration and prediction are used to analyze bad geological sections, then the advance knowledge of geological conditions is improved, but the coverage and precision are limited
Solution Approach 1:
The patent creates a universal sample library that covers multiple geological conditions and TBM operating parameters through systematic numerical simulation. The library is designed to be applicable to various geological scenarios (different rock types, stress conditions, water conditions) and can predict both TBM jamming and surrounding rock stability, providing comprehensive coverage and high precision that exceeds the limitations of traditional geological exploration methods.
Data Source
AI summary
A method of predicting tunnel boring machine (TBM) jamming based on deep neural network and numerical simulation and a system thereof are provided, belonging to the technical field of tunnel boring. The method including the following steps: constructing a jamming numerical sample library by using s numerical simulation technology; establishing a jamming prediction model based on the jamming numerical sample library by using a Convolutional Neural Network (CNN) and a Transformer; and implementing real-time monitoring and early warning of TBM jamming by using the jamming prediction model. The present application realizes the real-time monitoring and early warning of TBM jamming, reduces or avoids the jamming phenomenon, and improves the safety and efficiency of TBM construction.