Research on radon concentration measurement value correction based on FASTLOF and NPSO-BP neural network model

Abstract

To address the issue of decreased measurement accuracy in radon measurement devices due to the effects of temperature and humidity, a method has been proposed for correcting radon measurement readings based on a FASTLOF (Fast Local Outlier Factor) and NPSO-BP (Normalized Particle Swarm Optimization-Back Propagation) neural network model. The study employed the RAD7 portable radon detector and utilized the FASTLOF, NPSO, and BP neural network algorithms to perform data detection and correlation analysis on the environmental temperature, humidity and instrument readings. A correction model for the measurement data was established and trained to enhance the validity of the instrument's readings. Validation and analysis were conducted using data sets, stable radon concentration measurements in HD-6 multifunctional self-controlled radon chamber, and indoor radon measurement experiments. The experimental results indicate that the model can effectively correct radon concentrations, improve the accuracy and stability of the measurement data, with the maximum relative error not exceeding 8.6%, thus meeting monitoring requirements.