Radon Patrol

Detector

Camber detector unit: Measured volume of air will be pump to camber with scintillator detectors. Electronic will collect numbers of detected Radon isotope decays. Number of decay determines harmful Radon isotope density.

Develop small energy efficient IoT device “Radioactivity patrol” for areas with underground Radon-222 (222Rn) and Radon-220 ( 220Rn) source areas, which have sever task: 24/7 monitoring of ionization radiation based on scintillation crystal and semiconductor photo multiplier. This setup can measure with very high accuracy with low noise background. On this first stage of project, we would like to measure appearance of harmful Radon isotopes and doses of ionization radiation at all.

24/7 temperature, humidity and air pressure in surrounding environment based on common sensors with high energy efficiency. This setup of of sensors shall gave us view about weather condition in measured area, which will be use for “radioactivity forecast in later stage of project” especially for areas with 222Rn and 220Rn gas sources.
Collect and upload measured data to server, via IoT network SigFor or Lora. From this set of measured data, we would like to search for dependencies between radioactivity peaks and weather conditions, like air pressure changes or temperature.
“Radioactive patrol” IoT device will by build on scintillator based detector. Purpose of use this detector is high energy resolution of scintillator crystal + semiconductor photomultiplier set, which gave us guarantee of harmful Radon isotopes occurrence. This is important especially in high background radiation, which is typical for Uran deposit areas (source of 222Rn and 220Rn).
Output data of “Radioactive Patrol” IoT device will be Radon concentration in time intervals with outdoor environment conditions. In previous short term measurements, was found out,that concentration of 222Rn or 220Rn is very changeable (daily difference up to record value 106 Bq/m3). Our hypothesis is that 222Rn or 220Rn time concentration is highly dependent on local environment conditions as air pressure changes, air humidity and temperature.

This devices will be placed in selected areas in Kosice area Jahodna, where is large deposit of Uranium, which Radon is one element of Uranium decay product. With regard to the uranium pool near the residential areas, civilians are exposed to ionizing radiation from natural Uranium products. One is mainly Radon radioactive isotopes which is naturally unstable and it produces very dangerous alpha radiation. Alpha particles can travel in air only few centimetres or maximum meters, but very high mass of alpha particle is dangerous to human DNA during impact. If Radon molecules comes to lungs, there is high probability to release Alpha particle which can damage exposed human cells (especially DNA high probability of malignant tumor). From medic studies is common known, that only small rise of Radon activity (+ 100 Bq/m3) can increase the number of cases of lung cancer by up to 16%. That’s the main reason, we we would monitor Radon activity changes and predict high concentration due to weather forecast in monitored are.

Radon is naturally high density gas and as the gas flows landscape. Our hypothesis is that air pressure, winds, rains, temperature and humidity affects the amount of released gas from the underground sources and distribute it in the countryside.

Second stage:

After sufficient quantity of measured samples with help of AI, it will be possible to predict radiation peaks based on weather forecast in inquiry areas. Also, from energy analysis and activity of trapped particles/rays, it will be possible set Sievert units (measure of the health effect of low levels of ionizing radiation on the human body) in areas where will be Radiation patrol unit placed. It will be possible to predict Radon levels without placing high density net measuring device in given area.

AI Algorithms

In the beginning, while not having enough historical data recorded, we will use rather simpler baseline algorithms such that k-NN, SVM, for example. However, the data are of spatial and time context, thus, related algorithms will be selected and used: For example, predicting Radon levels might correspond to a task of time-series forecasting such that other, contextual data (e.g. weather conditions or landscape) are considered also. After gaining enough knowledge of the surface properties of the data, their types and characteristics (e.g. the noise in the data or the bias) we will be able to choose the right models and algorithms. We expect that there will be no single model but rather an ensemble of various models which will, most likely, lead to desirable results.

AI Frameworks

The work will be carried out in an exploratory manner, i.e. the team will try various AI frameworks, compare their results and choose the one which are better suited for the final platform. The intention is to use freely available AI frameworks.