Skip to content

Repository files navigation

🧪 UCM-Parameter-by-ML

🇨🇳 基于机器学习的 UCM 本构模型参数识别 | 🇬🇧 ML-based UCM Parameter Identification


Python PyTorch CUDA License


🇬🇧 English Documentation   ·   🇨🇳 中文文档


📖 Overview / 概述

This project uses a fully-connected deep neural network to predict four key parameters of the Unified Constitutive Model (UCM) for structured soils, directly from undrained triaxial compression test curves.

本项目使用全连接深度神经网络,从不排水三轴压缩试验曲线中直接预测结构性土体**统一本构模型(UCM)**的四个关键参数。


🎯 Target Parameters / 目标参数

Symbol Parameter / 参数 Range / 范围
m OCR control / 超固结控制参数 1.0 – 10.0
a Structure control / 结构性控制参数 0.1 – 1.0
OCR Over-consolidation ratio / 初始超固结比 0.75 – 1.5
R* Initial structure / 初始结构性参数 0.0 – 1.0

🏗️ Architecture / 网络架构

╭─────────────────────────────╮
│   Input Layer  (15 features) │
╰──────────────┬──────────────╯
               │
         FC(128) + ReLU
               │
         FC(64)  + ReLU
               │
         FC(32)  + ReLU
               │
         FC(16)  + ReLU
               │
         FC(8)   + ReLU
               │
     ┌────────────────────┐
     │  Output Layer (4)    │
     └────────────────────┘
               │
       [m, a, OCR, R*]

🚀 Quick Start / 快速开始

# 1️⃣  Install dependencies / 安装依赖
pip install -r requirements.txt

# 2️⃣  Generate training data / 生成训练数据
cd dataset && python 001-Generate_Dataset.py

# 3️⃣  Train model / 训练模型
python main_model.py

# 4️⃣  Predict on real data / 真实数据预测
python model_test.py

Made with ❤️ for geotechnical research

⭐ Star this repo if you find it helpful!

About

Paper code - Inversion of constitutive parameters for the Shanghai Model from synthetic triaxial data (Computers and Geotechnics, COGE-106268)

Topics

Resources

Stars

5 stars

Watchers

1 watching

Forks

Releases

Packages

Used by

Contributors

Languages