Code associated with the publication "Tinyml anomaly detection for industrial machines with periodic duty cycles" at the Sensor Application Symposium 2024. It contains a jupyter notebook to train/test the ML models and the c-code to run the models on microcontrollers.
The csv input files ('Confidential_Drive_data_Jun2021.csv', 'Confidential_Drive_data_Okt2021.csv','Confidential_Drive_data_Jan2022.csv' and 'Confidential_Drive_data_April2022.csv') are not available due to conflicts of interest of the parties involved in this project. However, an "input_data.csv" example file is provide to evaluate the models.
See the 'requirement.txt' file
If you use this code or the paper, please cite as:
L. S. Martinez-Rau, Y. Zhang, B. Oelmann and S. Bader, "TinyML Anomaly Detection for Industrial Machines with Periodic Duty Cycles," 2024 IEEE Sensors Applications Symposium (SAS), Naples, Italy, 2024, pp. 1-6, doi: 10.1109/SAS60918.2024.10636584.
or use the BibTeX:
@INPROCEEDINGS{10636584, author={Martinez-Rau, Luciano Sebastian and Zhang, Yuxuan and Oelmann, Bengt and Bader, Sebastian}, booktitle={2024 IEEE Sensors Applications Symposium (SAS)}, title={TinyML Anomaly Detection for Industrial Machines with Periodic Duty Cycles}, year={2024}, volume={}, number={}, pages={1-6}, keywords={Productivity;Machine learning algorithms;Recurrent neural networks;Microcontrollers;Tiny machine learning;Belts;Real-time systems;anomaly detection;conveyor belt;industry 4.0;low-power microcontroller;machine learning;maintenance;tinyML}, doi={10.1109/SAS60918.2024.10636584}}