Industries are currently going through “The Fourth Industrial Revolution,” as
professionals have called it “Industry 4.0” (I4.0). Integration of physical and
digital systems for the product life cycle mainly concerns Industry 4.0. With
the appearance of I4.0, the concept of prediction management has become an
unavoidable tendency in the framework of big data and smart manufacturing. At
the same time, it offers a reliable solution for handling test fatigue failures.
AI and its key technologies play an essential role -
- 1
to make industrial systems autonomous
like predicting test
failures
- 2
to make
possible the automatized data collection from industrial
machines/components.
Based on these collected data types, machine learning algorithms can be applied
for automated failure detection and diagnosis. However, it is a bit difficult to
select appropriate machine learning (ML) techniques, type of data, data size,
and equipment to apply ML in industrial systems. Selection of inappropriate
technique, dataset, and data size may cause time loss and infeasible result
prediction. Therefore, this study aims to present a comprehensive case study of
predicting the testing failure using ML techniques.
This work presents a novel approach for different parameter- based fatigue
failure (rig testing failure) characterization using artificial intelligence
(AI). The deep learning algorithm is trained on carefully collected physical
testing data (historical data), which helps in predicting the new product
development testing failure cycles based on basic design parameters available at
the start of the program such as loading, component dimensions, distances, and
inclination angle, etc. Rig testing reveals the testing cycles which indicate
either failure or non-failure of the component (depending upon the passing
criteria). Thus, every driveline component subjected to this research work
generates at least one data set (testing values from AI). Based on this study, a
conservative failure prediction accuracy of 88% is achieved. So, this
methodology is pioneering to predict fatigue failure without -
- 1
comprehensive expensive
physical testing.
- 2
the
need for extensive, error-prone, use of complex assessment
methodologies
With expert knowledge of evaluation procedures, the developed AI approach enables
quick and reliable prediction of fatigue failure of components based on
elementary key design parameters which can reduce the overall design cycle
time.