Modified GMDH-NN algorithm and its application for global sensitivity analysis

Abstract Global sensitivity analysis (GSA) is a very useful tool to evaluate the influence of input variables in the whole distribution range. Sobol' method is the most commonly used among variance-based methods, which are efficient and popular GSA techniques. High dimensional model representation (HDMR) is a popular way to compute Sobol' indices, however, its drawbacks cannot be ignored. We show that modified GMDH-NN algorithm can calculate coefficients of metamodel efficiently, so this paper aims at combining it with HDMR and proposes GMDH-HDMR method. The new method shows higher precision and faster convergent rate. Several numerical and engineering examples are used to confirm its advantages.

[1]  Saltelli Andrea,et al.  Global Sensitivity Analysis: The Primer , 2008 .

[2]  Tadashi Kondo,et al.  Medical image diagnosis of lung cancer by a revised GMDH-type neural network using various kinds of neuron , 2011, Artificial Life and Robotics.

[3]  Bruno Sudret,et al.  Global sensitivity analysis using polynomial chaos expansions , 2008, Reliab. Eng. Syst. Saf..

[4]  A. Ivakhnenko.  Heuristic self-organization in problems of engineering cybernetics , 1970 .

[5]  A. Saltelli,et al.  Making best use of model evaluations to compute sensitivity indices , 2002 .

[6]  H. Rabitz,et al.  Practical Approaches To Construct RS-HDMR Component Functions , 2002 .

[7]  Herschel Rabitz,et al.  High‐dimensional model representations generated from low order terms—lp‐RS‐HDMR , 2003, J. Comput. Chem..

[8]  S. J. Farlow.  The GMDH Algorithm of Ivakhnenko , 1981 .

[9]  Nilay Shah,et al.  Global sensitivity analysis using sparse high dimensional model representations generated by the group method of data handling , 2016, Math. Comput. Simul..

[10]  Nilay Shah,et al.  Metamodelling with independent and dependent inputs , 2013, Comput. Phys. Commun..

[11]  I. Sobola,et al.  Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates , 2001 .

[12]  H. Rabitz,et al.  Random sampling-high dimensional model representation (RS-HDMR) and orthogonality of its different order component functions. , 2006, The journal of physical chemistry. A.