Short name: DataTwin
Long name: DataTwin: Testing and experimentation with a novel data analytics service for predictive maintenance
Company: Nissatech | nissatech.com
Call: F4Fp-03-M (see call details)
Proposal number: F4Fp-03-M20
SUMMARY REMARKS & TESTBEDS
This proposal is about testing and experimentation with an innovative service for the predictive maintenance based on the creation of Digital Twins of industrial systems from the past data using a novel data analytics service. The system relies on D2Lab Framework for developing data analytics solutions (d2lab.nissatech.com). The service is based on learning usual behaviour of the system (aka Digital Twin) through applying different clustering methods on the past data, which is a very data intensive approach and requires scalable implementation and a cloud-based deployment. This leads to a better understanding of the behaviour of the system and will be used for predicting situations which might lead to some problems (anomalies). This kind of systems requires an efficient experimentation with real-time big data, as well as stored data (batch processing) which can be very efficiently implemented using TENGU testbed. In particular, we would like to test the performances of the new service under different configuration settings that will be provided by the Fed4FIRE+ experimentation. Especially important is to understand how the system will behave under different relations between the real-time and batch processing which are parts of the Lambda architecture provided by TENGU testbed, so that an efficient implementation of testing would be possible. The goal is to understand which QoS (Quality of Service) can be offered by D2Lab service for which costs (generated through the usage of the underlying computing architecture). The business impact of the proposal is huge since the predictive maintenance is an emerging area and our approach is enabling a novel approach for providing the maintenance models out of data (using data analytics). We will test the approach in two domains (manufacturing and oil&gas) where we have business partners ready for experimentation (and sharing data).
We argue that this proposal will bring two important novelties to the Fed4FIRE+ infrastructure:
- How this infrastructure can be used for experimenting with big data applications driven by huge real-time data (TENGU testbed)
- What are main advantages comparing to the “traditional” cloud computing infrastructure (like in Grid5000 testbed)
This should lead to important conclusions when and how to use infrastructure for “complex” (real-time and batch) processing, comparing to “ordinary” cloud computing infrastructure. The federate nature of the Fed4FIRE+ testbeds will be the best enabler for this process.