IC封裝研發階段導入模流分析(Mold flow analysis)有助於進行事前分析與尋求最佳化設計。在進行分析前，需要將模型轉換至實體網格(Solid mesh)，提供求解器(Solver)進行計算。實體網格型式中，六面體網格(Hexahedron)為公認品質最佳之網格型式，但是在自動化建立具有較高的難度，需以人工方式建構。四邊形網格(Quadrilateral mesh)為建構六面體網格之基礎，因此網格建構著重於表面網格(Surface mesh)。由於IC封裝模型具有多個內輪廓，在自動建構四邊形網格上較為困難，本研究針對IC封裝模型開發自動化區域劃分方法(Automated partitioning method)，結合自動化四邊形網格建構技術。此外，在模型劃分後於各區域建構不同密度的網格，減少整體網格數量。本研究首先針對封裝模型所有內輪廓進行計算，於所有內輪廓周圍建立邊界框(Bounding box)。接著，由模型本身輪廓與邊界框產生區域邊界並建立資料，區域邊界將用於組合成區域。最後，將區域邊界以指定條件組合成區域，再記錄區域建構網格所需資料，區域資料則可用於自動網格建構。本研究以19個模型測試，包含載體封裝模型、嵌入式晶圓級封裝(Embedded wafer level packaging)等，幾乎都可以成功劃分並建構網格。;The adoption of mold flow analysis during the packaging development stage can help in pre-process analysis and design optimization. Before conducting the analysis, the model needs to be converted to a solid mesh and provided to the solver. The solid mesh types include tetrahedrons, prisms, and hexahedrons. Among them, hexahedral mesh is recognized as the highest quality mesh type, but its automated construction is more challenging and needs to be constructed manually. The construction of quadrilateral surface meshes is critical as they form the basis for hexahedral meshing. Building a quadrilateral mesh automatically becomes more difficult when the IC packaging model contains multiple internal contours. In this study, an automated region partitioning method to divide the IC packaging model into multiple regions and combined it with automatic quadrilateral meshing techniques. Additionally, to enhance computational efficiency, different node spacing can be allocated during point generation, allowing for varying mesh densities between regions. Multiple size parameters were also adopted in this study to achieve a distribution of high and low-density meshes across the entire mesh and reduce the overall number of meshes. The research initially performed calculations on all internal contours of the packaging model and established bounding boxes around them. Subsequently, region boundaries were generated based on the model′s contours and bounding boxes, and data was created for each region boundary. These region boundaries were then combined into regions based on specified criteria, and the necessary data for constructing grids in each region was recorded. This region data was utilized for automated grid generation. The proposed method was tested on 19 models, including different types of packaging models, and achieved successful partitioning and grid construction for nearly all of them.
