ANN-ACCO手法による数値モデルの検定に関する研究

Abstract

In order to properly simulate the natural phenomena using numerical model, model parameters have to be estimated by an appropriate manner. Here, new approach using ACCO and artificial neural network is proposed for the calibration of numerical simulation model. ANN works as an error estimator in this proposed model. From the comparison of results with ACCO, although the number of function evaluation time is larger than that of ACCO, it is shown that the optimization by ANN-ACCO is reasonably carried out with better accuracy and stability. Model calibration was also successfully established by ANN-ACCO, then the some degree of its applicability to model calibrations were shown.