New strategies in model order reduction of trajectory piecewise‐linear models

Abstract

Model order reduction based on trajectory piecewise linearization (TPWL) is a beneficial technique for approximating nonlinear models. One efficient method for building projection matrix in TPWL reduction is by aggregation of projection matrices of linearization points (LPs). However, in this method, the size of projection matrix will also grow up by increasing the number of LPs, which yield the increment of the size of reduced model. In other words, the size of reduced model will depend on the number of LPs. In this paper, we will address this issue and propose two new strategies for obviating this problem. Contrarily to former works in TPWL modeling, we established a model via TWPL based on output weighting of parallel linear models. Then, we proposed two reduction strategies for suggested TPWL model. The first algorithm inspires from former works in this field but in a parallel structure that enable segregation of projection matrices whereas the second algorithm remedies the problem by considering the high-order TPWL model as a unit linear model and reduces this model like a linear model but uses back projection method for constructing different outputs. The efficiency of methods is shown by comparison with former TPWL methods through vast simulations.

12 Model order reduction and digital twins

“…For products with temperature-dependent material properties, the heat equation (121) becomes nonlinear due to κ = κ(T). In this case nonlinear algorithms (e. g., [8,114,69]) need to be applied.…”