Hierarchical hybrid performance prediction method for customized product in Internet + environment

Abstract

The invention discloses a hierarchical hybrid performance prediction method for a customized product in an Internet + environment, and the method comprises the steps: constructing a single fidelity layer hybrid model by a calculation simulation data set, and constructing an optimal BP neural network model by a historical actual measurement data set; and constructing a double-fidelity-layer hybridmodel based on the single-fidelity-layer hybrid model and the optimal BP neural network model by adopting a root-mean-square error strategy. Based on the hybrid agent model, association fusion of high-fidelity historical measured data and low-fidelity calculation simulation data is realized through hierarchical mixing of different fidelity levels, so that customized product performance predictionbased on multi-fidelity data information is realized; the prediction precision and generalization ability of the performance prediction model are improved, and efficient and credible prediction of customized product performance in the design stage is achieved.

Description

technical field [0001] The invention belongs to the field of performance prediction of customized products under the "Internet+" environment, and in particular relates to a layered and mixed performance prediction method of customized products under the "Internet+" environment. Background technique [0002] Under the "Internet +" environment, users' individual needs for products are becoming increasingly prominent, and at the same time, higher requirements are put forward for users' deep participation in the design process. For the personalized customized product design driven by user needs, the performance prediction of customized products helps to achieve efficient and rapid response to user needs in the design stage, and at the same time can effectively reduce design costs and shorten the design cycle, thereby effectively promoting deep user participation in design process. [0003] There are generally two types of traditional product performance prediction methods. The ...

Images