Abstract
In this study, the Finite Element Method (FEM) [1-3] was used to build a 3D model, and to simplify the curing reaction of the epoxy molding the experimentally verified Equivalent CTE was used. The resulting chemical effects such as volume shrinkage have finally successfully established a model for Fan-Out Panel-Level Packaging (FOPLP). Although FEM can save a lot of time and cost compared with experiments, the results are often different due to human error, and the model must be re-established when the structure size is changed. Therefore, in order to reduce the above situation, this study uses an ensemble learning algorithm of Extremely Randomized Trees (ET) to predict the amount of warpage at different packaging scales. First, a training database for fan-out panel-level packaging models of different geometric sizes will be established through the finite element method. After learning, ET can quickly predict any geometric size Warpage value for the panel level package. However, from many literatures, it can be found that the warpage of the fan-out panel level package is an asymmetric shape. There may be many reasons for affecting this result. For example, the thickness of the package can vary due to uneven shrinkage of the epoxy, which can lead to opportunities for asymmetric warpage. Therefore, this study further explores the causes of asymmetric warpage, such as the influence of the process and so on. At the same time, a threedimensional model is established by the finite element method to simulate asymmetric warpage caused by process differences.