Abstract:The current research of the decomposition methods of complex optimization model is mostly based on the principle of disciplines, problems or components. However, numerous coupling variables will appear among the sub-models decomposed, thereby make the efficiency of decomposed optimization low and the effect poor. Though some collaborative optimization methods are proposed to process the coupling variables, there lacks the original strategy planning to reduce the coupling degree among the decomposed sub-models when we start decomposing a complex optimization model. Therefore, this paper proposes a decomposition method based on the global sensitivity information. In this method, the complex optimization model is decomposed based on the principle of minimizing the sensitivity sum between the design functions and design variables among different sub-models. The design functions and design variables, which are sensitive to each other, will be assigned to the same sub-models as much as possible to reduce the impacts to other sub-models caused by the changing of coupling variables in one sub-model. Two different collaborative optimization models of a gear reducer are built up separately in the multidisciplinary design optimization software iSIGHT, the optimized results turned out that the decomposition method proposed in this paper has less analysis times and increases the computational efficiency by 29.6%. This new decomposition method is also successfully applied in the complex optimization problem of hydraulic excavator working devices, which shows the proposed research can reduce the mutual coupling degree between sub-models. This research proposes a decomposition method based on the global sensitivity information, which makes the linkages least among sub-models after decomposition, and provides reference for decomposing complex optimization models and has practical engineering significance.

Global sensitivity analysis for multiple outputs and their solutions

Online Probabilistic Assessment of Operating Performance Based on Safety and Optimality Indices for Multimode Industrial Processes

A generalized sensitivity analysis method based on variance and covariance decomposition of summatory functions for multi-input multi-output systems

Analytical Global Sensitivity Analysis with Gaussian Processes

A comparison between Sobol's indices and Shapley's effect for global sensitivity analysis of systems with independent input variables

A new paradigm for global sensitivity analysis

Global sensitivity analysis of computer models with functional inputs

Variance Based Global Sensitivity Analysis For Uncorrelated And Correlated Inputs With Gaussian Processes

Derivative-based global sensitivity measures and their link with Sobol sensitivity indices

Efficient global sensitivity analysis method for dynamic models in high dimensions

Global Sensitivity Analysis for Field Response Based on the Manifold of Feature Covariance Matrix

Combination Weighting Approach Based on Deviation Function and Joint-Entropy

Global sensitivity metrics from active subspaces

Mixed-Variable Global Sensitivity Analysis For Knowledge Discovery And Efficient Combinatorial Materials Design

Extended Newton Methods for Multiobjective Optimization: Majorizing Function Technique and Convergence Analysis

Sensitivity Analysis of the Cost Coefficients in Multiobjective Integer Linear Optimization

Calculations of Sobol indices for the Gaussian process metamodel

Decomposition Method of Complex Optimization Model Based on Global Sensitivity Analysis

Kernel-based ANOVA decomposition and Shapley effects -- Application to global sensitivity analysis

Global optimization using Sobol indices

Non-probabilistic sensitivity analysis method for multi-input-multi-output structures considering correlations