The aim of this project is building an IIoT solution with embedded Artificial Intelligence enabling rapid and custom-made prototyping, design for scalability and industrialization.

Connectivity and structured information are expected to be the cornerstones of Industry 4.0, but acquiring information is still complex and the associated costs must be predictable and as low as possible to optimize ROI.
In this context the HydroZest project has been awarded by the DIFIGED H2020 project to implement the Digital Challenge proposed by GE Hydro and its real-world industry defies.

Today, microcontrollers are evolving fast and become a great platform to collect, process and classify information at low cost. However, there is no “off the shelf” well industrialized solutions offering at the same time a panel of sensing capabilities, data processing and AI inference for the GE Hydro France’s harsh environment with the high level of today’s requirements.

The HydroZest solution has been designed in a partnership with GE to meet the needs specified in a very first specific use-case  for connecting a set of sensors, collecting data and running an embedded artificial intelligence algorithm to detect the operating mode of the Main Inlet Valve and its condition among a set of failures described by GE Experts to anticipate maintenance.

A modular hardware solution to connect the different sensors and quickly gather data has been developed. Modularity is important in such projects to be able to iterate and make the prototype evolve with the needs of the application (change microcontroller for processing needs, add a radio module, add more ADCs or add a Real Time Clock). Within HydroZest project,  4-20 mA analog sensors and RS232 data become available in the 6TRON range.
The system purpose is to collect and fusion these analog and digital data in order to extract discriminative information coming from smart sensors and classic sensors used to identify GE’s failure modes and anomalies. Then the system will be able to provide alarms to operators with associated information concerning the mode of operation and feature vectors.
Furthermore, the final artificial intelligence model will be integrated into an industrial version of the prototype.

Thanks to CATIE’s platform, the industrialization phase is accelerated because most of the work previously done can be reused (hardware and software). A pilot was installed in 2022 to start data collection and inference on customer sites. GE is collecting data but is also providing services to enhance availability using its APM Platform to better plan corrective actions and decrease maintenance costs.