Amiral Technologies

Our value proposition is based on the four key competitive elements

• Address the lack of historical data

Industry suffers from lack of historical data of defects. DiagFit addresses this issue/shortcoming with the accuracy of DiagFit’s health indicators. Unsupervised learning allows training the algorithm during normal operations to build the “good behaviour space” and detect deviation with unprecedented precision levels.

Customer Case: Non-Destructive Testing in the energy sector with unsupervised learning
- Defect detection rate: 100%
- False alerts rate: 17%

• Maximize prediction performance

Whether it is the defects detection rate or the false alerts rate, DiagFit allows reaching unprecedented precision levels. False alerts rate is minimized thanks to the accuracy of DiagFit’s health indicators that allow to build voting mechanisms reducing the false alerts rate.

Customer Case: Detection of defects in a railway point system:
- Supervised mode: 96,5% detection rate (against 91,5% without DiagFit).
- Unsupervised mode: 94,2% detection rate with 23,3% false alerts.


• Address complex problems for critical equipment

Thanks to the extraction of State of Health indicators, DiagFit allows detecting weak and transitory signals and to address fine correlation issues between variables.

- Successful Proofs of Concept in Aeronautics, Automotive and Energy
- Design of « Software Sensors* » in Automotive and Manufacturing

* a software sensor is a type of software that, given the available information, processes what a physical sensor otherwise would. It learns to interpret the relationships between the different variables and observes readings from the different instruments. Think of it as a kind of “ghost” of the physical sensor.

• Shorten time to build predictive models

This criterion is particularly of importance in aeronautics for example where a very large number of models needs to be designed to cover numerous equipment, systems and subsystems on different aircrafts (potentially hundreds of models per aircraft version).

Modelling of wear in aeronautic equipment
Model designed in few hours