Intelligent Calibration of Static FEA Computations Based on Terrestrial Laser Scanning Reference

Area of Science:

• Engineering
• Computational Mechanics
• Artificial Intelligence

Background:

• Advanced calibration and sensor monitoring increase demand for accurate finite element analysis (FEA).
• Standard FEA often uses simplified geometry, leading to deviations from real-world structural behavior under static loading.
• Accurate calibration is crucial for reliable structural behavior prediction and monitoring.

Purpose of the Study:

• To develop and validate a deep learning methodology for calibrating FEA results.
• To leverage sensor-based measurements for refining FEA computations.
• To enhance the accuracy and predictive capabilities of FEA for structural monitoring.

Main Methods:

• Utilized long short-term memory (LSTM) networks to learn deviation sequences between standard FEA and reference measurements (e.g., terrestrial laser scanning).
• Trained deep learning models to capture complex deviation patterns during static loading processes.
• Implemented threshold control for final FEA computation calibration.

Main Results:

• The deep learning approach effectively learned and captured complex deviation principles.
• Generated FEA sequence results were calibrated, significantly reducing mean square errors.
• The calibration method demonstrated a substantial improvement in the accuracy of future FEA predictions.

Conclusions:

• Deep learning-based calibration of FEA offers a robust solution for enhancing accuracy.
• This methodology strengthens calibration depth and improves the reliability of structural behavior predictions.
• The approach is highly beneficial for structural monitoring and predicting future behaviors.