Impurity gas detection dataset on 2/3-scaled canister mock-up

This repository contains datasets used for training Variational Autoencoders (VAE) and Wasserstein Autoencoders (WAE) in the detection of impurity gases on a 2/3-scaled sealed canister mock-up. The research focuses on identifying abnormal gas concentrations using acoustic sensing techniques. These datasets were utilized in the study titled "Acoustic Sensing and Auto-encoder Approach for Abnormal Gas Detection on a Spent Nuclear Fuel Canister Mock-up."

Files

1. X_train_helium(argon).npy: This file contains the training data with helium in day-1.
2. Argon_Percentage.npy: This file contains the different impurity levels of argon used in the experiments.
3. impurity_data.npz: This file is a dictionary containing ultrasonic response data for various argon impurity levels.
4. X_train_helium(air).npy: This file contains the training data with helium in day-2.
5. Air_Percentage.npy: This file contains the different impurity levels of argon used in the experiments.
6. impurity_data.npz: This file is a dictionary containing ultrasonic response data for various air impurity levels.

Data Structure

X_train_helium(argon).npy
- Shape: (1,991, 301)
- Description: This array contains the training data, where:
  - Each of the 1,991 rows represents a unique sample.
  - The 301 columns represent the truncated ultrasonic response for each sample.

X_train_helium(air).npy
- Shape: (988, 301)
- Description: This array contains the training data, where:
  - Each of the 988 rows represents a unique sample.
  - The 301 columns represent the truncated ultrasonic response for each sample.

Argon_Percentage.npy or Air_Percentage.npy
- Shape: (41, 1)
- Description: This array contains the impurity levels of argon and air used in the experiments. There are 41 different impurity levels, each represented by a percentage in this file.

impurity_data.npz
- Structure: This file is a dictionary containing 40 items, each representing data for a specific argon impurity concentration level.
- Description: Each item in the dictionary is structured as follows:
  - Shape: (51, 301)
  - Details:
    - The 51 rows represent repeated ultrasonic response measurements at each impurity concentration level.
    - The 301 columns represent the truncated ultrasonic response corresponding to each measurement.

Contact
For any questions or issues related to this dataset, please contact the repository owner via bozhouzh@usc.edu.