Predicting radon measurements for a radon control system

Radon is a colorless, odorless and tasteless radioactive gas formed by the decay of radium in nature. It accounts for about half of the radiation dose taken from natural sources. Found naturally in soil, rocks and groundwater, this gas can accumulate in weakly ventilated indoor environments and can become a significant health risk. Radon is recognized by the World Health Organization as the second most dangerous cause of lung cancer after smoking. Therefore, it is crucial for public health to monitör and predict indoor radon levels before they reach dangerous levels. This thesis present the IRL-RTA (Isparta Radon Laboratory-Radon Prediction Network) system, which is developed for online radon monitoring and predicting future measurements. The system is capable of taking hourly indoor radon measurements from various locations and predicting the next measurements based on previous data collected. This radon prediction network was built on IRL-RMS (Isparta Radon Laboratory Radon Monitoring System), which was originally designed for indoor use and has demonstrated reliable performance in monitoring radon levels effectively. A preliminary study was carried out to determine the form of the prediction algorithm to be used for IRL-RTA. The study compared three formulations: linear regression (T1), a temperature-dependent approximation (T2) and a humidity-dependent approximation (T3). Among these, the linear regression approach (T1) was found to be the most successful in producing consistent estimates that aligned with the pre-measurement data. Building upon the T1 formulation for the prediction algorithm, we integrated our C# coding into the IRL-RTA design. To evaluate the success of this system for predicting radon levels, we carried out test measurements over 210 days. Test-measurements were carried out at three different measurement points, which have different radon levels at Süleyman Demirel University (B1, B2 and B3 are units where low, medium and high radon concentrations were detected in previous studies, respectively). At each measurement point, a measurement unit from the IRL-RMS system and a measurement unit from the IRL-RTA systems were placed and the agreement between the two identical detectors and the agreement between the test-measurements and estimations by the detector of the IRL-RTA system were analyzed by calculating the relative error values between the data group. When 44,500 radon measurement and prediction data were analyzed, it was found that for all the measurement points examined, the agreement levels of radon predictions obtained by the IRL-RTA architecture were higher than the agreement levels between the measurements of two detectors collecting data from the same environment. The percentages of agreement between radon measurements and predictions of the IRL-RTA system were 91.5% for measurement point B1 (radon level <100 Bq/m3), 95.6% for measurement point B2 (radon level 100-300 Bq/m3) and 94.9% for measurement point B3 (radon level >300 Bq/m3). Furthermore, the IRL-RTA system demonstrated a high data transmission performance, achieving a rate of 98.1%. The findings of the study clearly showed that the proposed design can successfully generate data that aligns closely with actual measurements and can be reliably used to estimate indoor radon concentrations.