RSW gun fault prediction benchmark data set (demo)

The resistance spot welding (RSW) welding gun fault prediction benchmark data set has 72 multivariate time series in the training set and 8 in the testing set. Each time series length 604800 sampled at 1 Hz with missing values and has 20 dimensions (c1-c19 and the error code). We retain the missing value and the outliers of the welding gun time series for the potential of imputation research in the future.
This data set supports an academic paper named 'benchmark study for welding gun fault prediction'.

Feature name and explanation:

c1 :  Electrode cap offset;

c2 :  Electrode force;

c3 :  Electrode position;

c4 :   Force build-up;

c5 :  Balance pressure;

c6 :  Friction;

c7 :  Maximum aperture;

c8 :  Maximum electrode force;

c9 :  Mtart friction;

c10 :  US2;

c11 :  Welding point count;

c12 :  Position count;

c13:  Setpoints of counterbalance pressure;

c14:  Setpoints of electrode force;

c15 :  Setpoints of electrode position;

c16:  Setpoints of sheet thickness;

c17 :  Setpoints of velocity;
c18:  Setpoints of force build-up;
c19 :  Offset value in robot.

Machine Learning Task:
This dataset is suitable for a time series forecasting task, where machine learning models can be trained to predict future welding parameters based on the provided welding parameters time series in history. 

Code for quick start:

https://zenodo.org/record/7655025

If you want to have an overview of the data before downloading all of it, you can download only the files with the word "Damo" in the file name.

For any question, please contact 1910633@stu.neu.edu.cn