Development of innovative machine and deep learning: algorithms for the detection, classification, and monitoring of CBRNe agents

In today's world, the potential occurrence of events related to CBRNe (Chemical, Biological, Radiological, Nuclear, and Explosive) risks is increasingly relevant. Given the current level of globalisation and technological advancement, alongside recent events in human history, phenomena such as the emergence or re-emergence of biological agents and the subsequent risk of pandemics are now recognised as real possibilities. If it has been considering the potential for the intenTIonal or accidental release of chemical or radiological agents, the development of tools and algorithms for the identification and risk assessment of these agents becomes essential for ensuring safety and security. Aim of this work is to explore the use of Machine Learning (ML) and Deep Learning (DL) methodologies as tools to mitigate CBRNe events. Thus, that work investigate several ML and DL algorithm for several applicaTIons. That has been allowed starting from the investigation about the use of Physics-Informed Neural Network to extract useful information by uncalibrated data, when the physics law that describe a phenomenon are known, resolving the necessity of a classical supervised calibration approach that cannot always possible (e.g. due to the complexity of the experiments through which the data are acquired, because of the cost, or the time demand, as well as the hazardous nature of handling certain substances), thus a problem linked with the entire research world linked with the CBRN. Passing through the investigation about a novel Light (or Laser) Induced Fluorescence Technique in combination with a DL algorithm to detect and measure the concentration’s values of more than one Biological Agent (BA) in a mixture sample (linked with the Biological (B) Threats). Concluding with DL algorithms’ applications linked with the Chemical (C) and Radiological/Nuclear (RN) threats, through a numerical or simulation point of view. In the first case researching about the possibility to use DL algorithm to reconstruct the dispersion map of a chemical substance in the environment using local data e weather condition’s data, supposing to have a static or mobile grid of chemical sensors. In the second case investigating the use of DL algorithm for different scope connected with the RN threats, but at the least to identify a radiological source only through the analysis of energy emission spectra.