# Modified PINN

**Repository Path**: wungchb/modified-pinn

## Basic Information

- **Project Name**: Modified PINN
- **Description**: modified PINN（经过注释和少量修改的PINN）
https://github.com/PML-UCF

- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2022-01-12
- **Last Updated**: 2022-01-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# README

## 1、文件结构

`pinn_wind_bearing`为`Ref.64`

`pinn_corrosion_fatigue`为`Ref.65`

`pinn_corrosion_fatigue/2020-jcise`
是论文“基于物理信息 ...”的实现

`pinn_corrosion_fatigue/2020-jcise`包括：
- **bias_model.py**: contains function that generates the proposed hybrid recurrent neural network.
- **run00_MLP.py**: file that initializes multiple architectures for the multilayer perceptron considering several auxiliary planes. 
- **run01_rnn_training.py**: trains the proposed hybrid recurrent neural network considering the multiple pre-trained multilayer perceptrons.
- **run02_rnn_x_validation.py**: file that implements a cross-validation procedure to help to select the most suitable combination of multilayer perceptron acrchitecture and auxiliary plane for damage prediction. 
- **run03_rnn_diagnosis.py**: predicts the crack length in the entire fleet at the 5th year (or 15,000 flights) of operation. 
- **run04_rnn_prognosis.py**: damage forecast of the entire fleet at the 6th year (or 18,000 flights) of operation. 


## 2、复现
### 2.1 代码
`pinn_corrosion_fatigue/2020-jcise`：

`bias_model.py`使用PINN自己搭建的库，其中的某些自定义`Layer`搭建了一个RNN模型

使用到了`inputsSelection`、`CumulativeDamageCell`、`StressIntensityRange`、`WalkerModel`
等PINN自定义Layer，需要进一步与论文对应分析

## 3、分析

代码貌似有部分需要重写，不符合文件书写要求/代码不利于计算

如`run00_MLPs.py`

## 4、改进
- `run00_MLPs.py`读取文件应该放在遍历以外
-