Artificial Intelligence (AI) algorithms for evaluation of optical fiber scintillation detector performance

Abstract

Optical fiber based scintillation detector performance is critical to determine life of detector and accuracy of measurements. To do this, very hard experimental works is required. Alternatively, there are much easier computer-based decision method including artificial intelligence (AI) algorithms. The use of AI based automatic detector performance systems could bring practicality to radiation based experimental measurements when accurate and rapid decision on scintillation detector performance has become necessary. In this study, a data set including 808 experimental measurements was used to predict performance of optical fiber. Application dose, wavelength (430 nm, 470 nm, 500 nm and 600 nm) and application time were given as input to AI algorithms (Linear Regression (LR), Lasso Regression (LASSO), ElasticNet (EN), Classification and Regression Trees (CART), Support Vector Machines (SVM), k-Nearest Neighbors (k-NN), Artificial Neural Network (ANN), Ensemble Algorithms, AdaBoost (AB) Algorithm, Gradient Boosting (GBM) Algorithm, Random Forests (RF) Bagging Algorithm and Extra Trees (ET) Bagging Algorithm) and attenuation, transmission and light intensity were automatically estimated. Performance of AI algorithms were tested for both the raw data (rd) and normalized data (nd), separately.Among the proposed expert systems, the GBM showed the best performance with the values of R2 = 0.9760 and R2 = 0.9756 for the rd and nd data, respectively. As a result, it was concluded that performance of fiber optic scintillation detector can be estimated accurately, rapidly, and dynamically using AI based expert system.