andres2203 / rig_simulation

Testing rig simulation and fault prediction

TEST RIG SIMULATION AND FAULT PREDICTION

Fault Detection, Isolation and Recovery (FDIR)

An IRONHACK/Berlin Machine Learning project by Andres Lucht (github.com/andres2203) and Georgios Papadopoulos (github.com/GeorgiosKP).

1. Background

Data Source: Condition monitoring of hydraulic systems Data Set

Examination and evaluation of sensor related data of a hydraulic testing rig. Prediction 4 different target values for stable situation/fault detection.

2. Content

\data: sensor data

\descriptions: Further descriptions

\experimental: experimental notebooks (see experimental)

merged_df.pkl: data file

01-load_and_examine_data.ipynb

• Loading, processing and examination of data

02-regression.ipynb

• Regression models to analyze system dependencies

03-machine-learning.ipynb

• ml part (see applied methods)

3. Applied Methods

1. Cross validation on train set
2. Choose best performing algorithm, applied methods
   1. LinearRegression
   2. LogisticRegression
   3. RandomForest
   4. DecisionTree
   5. KNearestNeighbors
3. Feature elimination (RFECV)
4. Evaluation on test set

experimental

1. Clustering
   1. PCA
   2. LDA
2. Analysis
   1. FFT