AIGAD (Artificial Intelligence for Graph Neural Network-based Anomaly Detection) is a set of ontological GNN-based models that leverages the correlations between different sensor measurements in IoT Industrial Networks in order to spot potential operation anomalies in advance and act accordingly. Put differently, AIGAD’s main purpose is to capture complex inter-sensor relationships and detect and explain anomalies that deviate from usual functions by using AI-based technologies that can exploit underlying, non-visible properties of sensors that typically correlate when these unexpected behaviours show up. This will allow industrial processes to anticipate potential operation downturns and minimize their negative effects.
Usage of Standards for data interoperability:
With AIGAD our goals include but are not limited to, fostering interoperability by creating a WoT (Web of Things) ontological resolution middleware that will be used to map IoT sensor data to AIGAD GNN anomaly-detection resolution models. Our clients add their devices to the database of IoT devices available in the platform, also including a TD (Things Descriptor) instance describing the device they included. This action will allow the models to clearly establish in advance the type of devices, how data is formatted, etc. thus fostering scalability in AIGAD.