CNC Milling Tool (Tool Wear Detection)

Project Overview

Project Details

In this project machining data was collected from a CNC machine for variations of tool condition, feed rate, and clamping pressure.

In 18 machining experiments, time series data was collected with a sampling rate of 100 ms from the 4 motors in the CNC (X, Y, Z axes and spindle).

Goal

The main goal of the project is to find the best model to predict the machine failure.

Tools:

Models:

• Machine Learning: LogisticRegression, SVC, RandomForest, GradientBossting, KNN (scikit-learn)
• Deep Learning: ANN(TensorFlow)

Data Visualization:

• Matplotlib
• Seaborn

Data Preprocessing:

• Missing Data Imputation: ArbitraryNumberImputer, EndTailImputerm, RandomSampleImputer
• Feature Scaling: StandardScaler, RobustScaler, MinMaxScaler
• Feature Selection: DropConstantFeatures, DropDuplicateFeatures, DropCorrelatedFeatures

Metrics:

• Accuracy
• Roc-auc
• Recall
• Precision
• F1
• Confusion matrix

Other:

• Grid Search (hyperparameters tuning)
• Pipeline
• Pandas
• Numpy