Application of BP network to nonlinear rectification of non-closed LRAD monitoring pipe interiors

Background: The accuracy for multi-parameter of long range alpha detector (LRAD) system needs to be improved. Purpose: To overcome the nonlinear effect on accuracy and have the measurements corrected. Methods: Key-parameter characteristics that affect open LRAD monitoring inside on-site pipes were analyzed. Based on the 352 groups of data from key-parameter characteristics analyzing, this paper adopted BP neural network to train and predict the measurements. Results: The predicted result has a good consistence with measurement, and the mean percentage error of the result is less than 5%. Conclusions: Using BP to predict the analysis result of LRAD has a good accurate, and besides, overcomes the impact of the system of nonlinear basically on condition of satisfying the radioactivity measurements of the law of statistical fluctuations. (authors)