Most important outcomes of the project:

Generative-Adversarial networks for advanced artificial data generation

GAN networks allow for advanced generation of new data samples for the purpose of classifier training.

Preliminary findings published in: Proceedings of the EWSHM conference

Recommendation system for data processing and interpretation

It is possible to recommend a good data processing and interpretation approach based on historical data. So far it is proven that the system is operational for Condition Monitoring problems.

This result is currently under submission.

Bayesian approach to data interpretation

Decisions based on measurements taken in distinct points in time can be aggregated using Bayesian approach for increased reliability

This result is currently under submission.

Super-diverse ensembles for robust data interpretation

Ensembles of classifiers of different structures trained on different subsets of data pre-processed using different methods allow for a multi-level decision fusion that increases accuracy of diagnoses

This result is currently under submission.