Multi-fidelity Data Aggregation using Convolutional Neural Networks

Multi-fidelity data exist in almost every engineering and science discipline, which can be from simulation, experiments, a hybrid form. High fidelity are usually associated with higher accuracy expense (e.g., high resolution experimental testing or finer scale simulation), while low-fidelity on the opposite side terms of cost. aggregation (MDA) this study refers to process combining two multiple sources different have estimation low computational MDA has wide range application science, such as multiscale multi-resolution imaging, simulation-testing. This paper presents novel framework named Data Aggregation using Convolutional Neural Networks (MDA-CNN) for multi-fidelity modeling. The MDA-CNN architecture three components: compiling, perceptive field convolution, deep neural network mapping. captures utilizes implicit relationships between any high-fidelity datum all available defined local convolution. Most existing strategies rely collocation method interpolation, focuses single point relationship. proposed several unique benefits. First, treats image processes them CNN, is very scalable dimensional more than fidelities. Second, flexibility nonlinear mapping facilitates does not need assume specific among Third, that at same order physical mechanisms assumptions needed some error estimation-based model). Thus, handle across scales, derivatives other correlated phenomenon framework. validated extensive numerical examples multi-source data. Discussions given illustrate benefits limitations Conclusions future work presented based observations current study.