APHELIO

Cyber-physical systems (T° sensors, presence sensors, noise sensors, position sensors (people and objects), fluid consumption sensors, actuators, automatic machinery, regulator, etc.) piloted by the IoT are being used increasingly. Being capable of detecting operational anomalies in these systems, whether due to breakdowns, malfunctions, incorrect use or cyber attacks, is crucial to the control of these systems. The AI-based techniques developed in the CSL project enable us to answer these questions.

The central element of CSL is a machine learning algorithm patented and designed to capture the dynamic behaviour of a cyber-physical system. The algorithm works in unsupervised mode and exploits the specific features of the systems connected to the physical world to function with as little data as possible by using pre-configured relationships and an indicator of the completeness of the learning.

CSL is able to connect to cyber-physical systems by using existing protocols or via dedicated wireless sensors without interfering with the existing installation. It makes it easier to understand the normal behaviour of the system and simpler to configure the exceptional behaviours so as to offer intelligent surveillance 24 hours a day.

The basic CSL algorithm has been evaluated for the detection of anomalies in heating systems (performance losses, abnormal usage, cyber attack). A configurable modular IoT agent is under development.

The technology can be applied to cyber-physical systems in smart buildings, industrial systems and healthcare systems. Examples of detectable anomalies are: Abnormal presence/absence of people, abnormal use of energy (time, quantity), cyber attacks, breakdown or deterioration of equipment, irregular or inappropriate maintenance in relation to the condition of the equipment.