Inteligent IoTSP - Implementation of Embedded ML AI Tensorflow Algorithms on the NVIDIA Jetson Tx Chip
Piotr Lipnickia b, Daniel Lewandowskia b, Michał Syfertc d, Anna Sztyberc e, Paweł Wnukf
- aABB
- bCorporate Research Center, PLCRC ABB, Kraków, Poland
- cPolitechnika Warszawska, IAiR
- dPolitechnika Warszawska, IAiR, Warsaw, Poland
- ePolitechnika Warszawska IAiR, Warszawa, Poland
- fPolitechnika Warszawska IAiR
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
This article presents a description of the project and implementation of the system for the execution of on-line diagnostics of compressors using the methods of artificial intelligence and the Tensorflow library. The main tasks of the system are: on-line acquisition of process data from the compressor set, on-line state monitoring (fault detection) of the compressor set based on the analysis of process data and using the classifiers modelled using the Tensorflow library. The system is intended to be a proof of concept, it should show the possibility of using Tensorflow library models running on the Jetson platform for on-line monitoring of compressor faults. The sample models proposed and prepared during previous research and development projects were used for testing. The algorithms used to identify and detect failures are based on MLP, CNN, SVM and LSTM - keras.