DeepSens

Name of the project: Modelling the behaviour and calibration of gas sensors based on Deep learning (DeepSens) (ACE053/22/000029)

Principal Investigator (PI): Petia Radeva

Funding entity: ACCIÓ. Agència de Suport a l'Empresa Catalana Type of entity: Not defined City funding entity: Barcelona, Spain

Duration: 27/07/2022 - 26/07/2025

In recent years, there have been various accidents, including fatalities, caused by toxic gas leaks in different plants and facilities both in Catalonia and in the rest of the world. The last one at the company Aigües de Girona, where poor handling of chemical components caused a toxic cloud on July 14 that caused the confinement of all people located less than a kilometre from the water treatment plant Bescanó water (Girona); likewise, on January 14, 2020, there was a serious accident at the IQOXE plant in La Canonja (Tarragona) when an ethylene oxide process reactor exploded, causing 3 deaths and 8 serious injuries.

Populations close to the largest petrochemical hub in southern Europe have grown many times towards industries, causing hundreds of thousands of people to live very close to complexes at risk of gas leakage. From time to time, accidents or incidents can occur that cause part of these gases to leave the perimeter of the industries.

Modelling the behaviour and calibration of gas sensors based on Deep learning (DeepSens). The vision is to develop and integrate algorithms based on machine learning (Machine learning (ML) and in particular Deep Learning (DL) represented by NNs) for the correction of zero drift and sensitivity of gas detectors for emergency safety applications and of SmartCities that is safe, accurate and robust, using distributed and heterogeneous networks of sensors.

Members involved

Research lines