Transpiration cooling has been widely used in many components operating under extremely high temperature, due to its advantage of uniform cooling, high efficiency and low coolant consumption. Recently matured metallic additive manufacturing technologies pushed transpiration cooling to a new stage with its extensive precision in fabricating micro regular features for porous media. As a result, micro controlling of porous features and developing thermal-load-adaptive transpiration cooling became possible. However, the large number of holes inside the porous media could cause a parameter exploding for investigating additive manufactured transpiration cooling structures, if each hole was treated separately. Meanwhile, the coolant accumulation phenomena remained to be a complex problem to understand or model. The present study proposed a convolution modeling method, which took advantage of convolution functions to model the cooling effectiveness of transpiration cooling with complex hole distribution. The proposed convolution model included an input layer, a convolution layer and an output layer. A single logic matrix was used as the input to dramatically reduce the dimension of parameter space. As convolution used the weighted average logic value of a range of neighboring regions, it was capable to reproduce the internally links and interactions of different coolant jets. Based on a series of numerical results, the model was initialized and trained to achieve a relative error around 4.04%. The model was further validated by several more complex geometries and demonstrated good accuracy in predicting transpiration cooling efficiency. This effort established a novel method to model transpiration cooling that could be additively manufactured. Compared with traditional correlation formulas, this network-like model had a much higher resolution while still preserving the low computational cost. (C) 2019 Elsevier Ltd. All rights reserved.