EEWnet - Earthquake Early Warning Network

Project Overview

EEWnet is a real-time seismic response prediction method of high-rise buildings based on deep learning for earthquake early warning

Details of the method can be referred to: Real-time seismic response prediction method of high-rise buildings based on deep learning for earthquake early warning

Environment Requirements

• Python 3.9.13
• CUDA 12.4

Dependencies

torch==2.4.1+cu124 numpy==1.22.4 pandas==2.2.3 scikit-learn==1.5.2 matplotlib==3.7.2

Dataset Description

• Input file: input.csv
• Data preprocessing:
  • Logarithmic transformation of data
  • Removal of outliers and NaN values
  • Features include:
    • nStory
    • storyheight
    • year
    • strutype
    • Earthquake Magnitude
    • EpiD
    • Vs30
    • Peak_Ground_Velocity
    • Significant_Duration
    • Arias_Intensity
    • Cumulative_Absolute_Velocity
    • Peak_Ground_Velocity
• Building monitoring data.xlsx 74 measured data used for validating the proposed method.

Model Architecture

• Input layer: 12 feature nodes
• Hidden layers:
  • 512 nodes × 4 layers
  • 256 nodes × 2 layers
  • 128 nodes × 1 layer
  • 64 nodes × 1 layer
  • 32 nodes × 1 layer
• Output layer: 1 node
• Activation function: LeakyReLU
• Regularization: BatchNormalization and Dropout

Training Parameters

• Learning rate: 0.0005
• Batch size: 512
• Epochs: 300
• Optimizer: Adam
• Loss function: MSE

Model Evaluation

The model is evaluated using the following metrics:

• R² score: Measures the goodness of fit between predicted and actual values
• Mean Absolute Error (MAE): Average absolute difference between predicted and actual values
• Mean Squared Error (MSE): Average of squared differences between predicted and actual values
• Root Mean Square Error (RMSE): Square root of MSE, representing the standard deviation of predictions

Output Files

The model training process generates the following files:

• loss_{timestamp}.png: Loss curve during training
• Predicted_vs_actual_{timestamp}.png: Scatter plot comparing predicted vs actual values
• my_best_model.pth: Saved best model weights
• true_pre.txt: Model prediction results
• metrics.txt: Contains all evaluation metric results
• loss.txt: Records loss values during training and testing

Usage Instructions

1. Data Preparation:

   • Place training data in the specified directory
   • Ensure data format meets requirements

2. Model Training:

   python EEWnet.py

3. View Results:

   • Check generated charts and metric files in the output directory
   • Use generated my_best_model.pth for predictions