Finite Element Pilot Deep Learning Proxy Model for Bridge State Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current bridge monitoring technologies rely on alarm thresholds and trend analysis, with simulation models only established during the design phase and lacking real-time online simulation capabilities, and require extensive data preparation for deep learning models, leading to inefficiencies and unreliable predictions.
Innovation Solution
A finite element pilot-based deep learning proxy model is developed, incorporating physical information constraints during training to enhance generalization capabilities, allowing for real-time simulation of bridge states using fewer data samples and autonomous model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a data-driven deep learning network is used to predict the next state point in iteration, then calculation efficiency and convergence of the finite element model are optimized, but a large amount of time is consumed for manual preparation of training data and the model has poor reliability without sufficient data input and understanding
Solution Approach 1:
The patent applies the self-service principle by enabling the deep learning model to autonomously complete the training process through physical guidance. The model automatically learns from physical information constraints without requiring manual data preparation, thus resolving the contradiction between calculation efficiency and time consumption for data preparation.
Solution Approach 2:
The patent replaces the mechanical manual data preparation process with an automated physical information constraint system. By substituting manual data collection and processing with automated physical guidance mechanisms, the system achieves both high calculation efficiency and eliminates time-consuming manual operations.
2Reliability
If conventional network learning is used, then model training can be completed, but simulation time is time-consuming and simulation efficiency is low
Solution Approach 1:
The patent applies parameter changes by transforming the training process from conventional network learning to physical information constraint-based learning. By changing the fundamental parameters of the training mechanism to incorporate physical guidance, the system maintains reliable model training completion while dramatically reducing simulation time and improving efficiency.
3Manufacturing precision
If a finite element simulation model is established during the design phase, then the model can simulate a single existing condition, but real-time online simulation of the bridge structure cannot be implemented
Solution Approach 1:
The patent applies the dynamics principle by transitioning from a static finite element model established during the design phase to a dynamic deep learning model that can perform real-time online simulation. The model adapts to varying bridge conditions in real-time, resolving the contradiction between simulation accuracy for single conditions and real-time online simulation capability.
Solution Approach 2:
The patent applies universality by enabling the model to perform multiple functions: it can simulate single existing conditions with high accuracy while also providing real-time online simulation for various bridge states. The deep learning model generalizes the capabilities of conventional finite element models, achieving both precision and versatility.
4Ease of manufacture
If pure data-driven neural network learning is used, then the model can be trained, but physical information constraints are not incorporated and generalization capability is limited
Solution Approach 1:
The patent applies the merging principle by combining data-driven deep learning with physical information constraints. The model integrates both data learning and physical guidance mechanisms, achieving both ease of training and enhanced generalization capability. This resolution merges the advantages of both approaches while eliminating their respective disadvantages.
Data Source
AI summary
This application provides a method for evaluating a running state of a bridge using a finite element pilot-based deep learning proxy model, belonging to the field of online simulation technologies of bridge structures. The method includes: S1:establishing a finite element simulation model; S2: obtaining vehicle load information according to vehicle positions, number plate information, and axle load-number plate information, obtaining environment load information according to temperature, humidity, and wind speed and direction information of the bridge, and obtaining vehicle-environment load information based on the vehicle load information and the environment load information; S3: adaptively training a finite element pilot-based deep learning neural network proxy model; and S4: inputting the vehicle-environment load information into a finite element pilot-based deep learning neural network proxy model, and outputting a real-time structural state of running of the bridge.