Modeling of Soft sensor based on Artificial Neural Network for Galactic Cosmic Rays Application

Conference of Theoretical Physics and Nonlinear Phenomena 2014: "From Universe to String's Scale"

Suparta, W.^1 ; Putro, W.S.^1

Space Science Centre (ANGKASA), Institute of Climate Change, Universiti Kebangsaan Malaysia, Bangi, Selangor Darul Ehsan

43600, Malaysia^1

关键词: Artificial neural network models;    Galactic cosmic rays;    Input variables;    Levenberg-Marquardt algorithm;    Multi layer perceptron;    Neutron monitors;    Root mean square errors;    Space radiations;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/539/1/012021/pdf
DOI  :  10.1088/1742-6596/539/1/012021

来源: IOP

【 摘 要 】

For successful designing of space radiation Galactic Cosmic Rays (GCRs) model, we develop a soft sensor based on the Artificial Neural Network (ANN) model. At the first step, the soft sensor based ANN was constructed as an alternative to model space radiation environment. The structure of ANN in this model is using Multilayer Perceptron (MLP) and Levenberg Marquardt algorithms with 3 inputs and 2 outputs. In the input variable, we use 12 years data (Corr, Uncorr and Press) of GCR particles obtained from Neutron Monitor of Bartol University (Fort Smith area) and the target output is (Corr and Press) from the same source but for Inuvik area in the Polar Regions. In the validation step, we obtained the Root Mean Square Error (RMSE) value of Corr 3.8670e-004 and Press 1.3414e-004 and Variance Accounted For (VAF) of Corr 99.9839 % and Press 99.9831% during the training section. After all the results obtained, then we applied into a Matlab GUI simulation (soft sensor simulation). This simulation will display the estimation of output value from input (Corr and Press). Testing results showed an error of 0.133% and 0.014% for Corr and Press, respectively.