Experimental tests of Gamma-ray Localization Aided with Machine-learning (GLAM) capabilities

Gamma-ray Localization Aided with Machine-learning (GLAM) utilizes an array of four rectangular NaI(Tl) detectors, and data processed a k-nearest neighbor machine learning model, to predict the location gamma-ray sources in stationary scenarios. This work demonstrates GLAM capabilities real source locations when trained purely on simulated data. To account for efficiency differences not captured simulations, simple calibration methodology that improves agreement between experimental multi-detector arrays is also introduced. The approach was tested using two different sets input features. first set features consisted total counts each detector. second from detector photopeak Compton continuum, which resulted better performance than simply counts. For one-minute measurements 40 μCi 137Cs up 500 cm away, mean absolute errors (MAEs) angular radial predictions were 3.5°and 10.0%, respectively. measurement times 6 60Co 250 MAEs 2.8°and 12.7%,