پديد آورندگان :
فولادي ، نعمت‌الله نويسنده Fouladi, Nematollah , دربندي ، مسعود نويسنده Darband, Masoud
كليدواژه :
جست‌وجوي سريع , ساختار اطلاعاتي , مرتب‌سازي جهت‌دار , مرتب‌سازي لايه‌‌يي , شبكه‌هاي بي‌سازمان
چكيده فارسي :
در اين نوشتار روشي جديد در زمينه‌ي دسترسي سريع به زيرساختهاي شبكه‌هاي بي‌سازمان با دو ويژگي حافظه‌ي اشغالي كمينه و سرعت محاسباتي بالا ارايه مي‌شود. ايده‌ي اصلي روش حاضر، مرتب‌سازي لايه‌‌يي ساختار اطلاعاتي شبكه‌هاي بي‌سازمان با شماره‌گذاري جهت‌دار المان‌ها و گره‌هاست، كه منجر به ايجاد نظم لايه‌‌يي منسجم در ساختار اطلاعاتي يك شبكه‌ي بي‌سازمان با اتصالات دلخواه المان‌ها مي‌شود. در اين روش با محدود‌كردن عمليات جست‌وجو در يك شبكه‌ي بي‌سازمان، و جست‌وجوي جهت‌دار در داخل لايه‌هاي مرتب‌شده، دسترسي به زيرساختهاي آن شبكه تسريع مي‌شود. نتايج حاصل از ارزيابي عملكرد روش حاضر در مرتب‌سازي ساختار اطلاعاتي شبكه‌ي بي‌سازمان بر روي مقطع طولي يك زيردريايي حاكي از قابليت زياد روش حاضر در مرتب‌سازي ساختار اطلاعاتي شبكه‌هاي بي‌سازمان دوبعدي است. آزمون‌هاي انجام‌شده با هدف دسترسي سريع به زيرساختهاي شبكه‌هاي بي‌سازمان مختلف، مويد سهولت و سرعت بالاي آدرس‌دهي با ساختار اطلاعاتي جديد هستند.
چكيده لاتين :
Contrary to structured grid methods, the unstructured grid provides sufficient flexibility to generate grid in very complicated geometries and eases grid adaptation, where it is required. Huge effort have been attributed towards improving the unstructured grid computations and despite achieving remarkable advancements, this field of research is still open to be explored by new contributors. Basically, most grid generators number unstructured grid substructures randomly. Therefore, no global directional features can be readily achieved in a regular unstructured grid data structure. This can cause serious trouble for grid users, such as those in CFD, where the knowledge of node, face, and cell neighborhood is essential. Since the neighborhood numberings are quite random, the node/face/cell numbers need to be stored explicitly in large vectors and matrices. Therefore, converting this random data arrangement into a layer data structure can enhance the computational efficiency in unstructured grid applications. In this research, a very simple and computationally low cost numerical procedure is developed to construct a layered data structure for a 2D triangular unstructured grid. The key point in this pattern is that each layer has a quasi-structured data structure performing ordered element number and node number patterns. To achieve this purpose, we develop a layer-by-layer ordering and renumbering algorithm. In this algorithm, the elements and nodes of the unstructured grid are renumbered in a manner to produce a new data structure very similar to those for a structured one. This algorithm benefits from the ordered elements data and nodes data produced in the preceding layer to construct the current neighboring layer and to reorder its element and node indices properly. The procedure starts from the first ordered node layer (usually the inner boundaries), and consecutively constructs the next neighboring element and node layers until reaching the outer boundaries of the domain. In this method, the overall patterns of the achieved node layers are mostly determined by the primitive chosen node layer. If we choose the nodes located at the target object as the first node layer, the resulting node layers would schematically take a pattern very similar to lines distributed around the target object in an O-type structured grid. However, if the initial node layer contains both the nodes located at the target object and the grid lines connecting the object to the boundaries, the overall pattern of the constructed node layers will be schematically similar to a set of grid lines around the target object in a C-type structured grid. Furthermore, if the target object consists of several components, the first node layer must include, not only the nodes located at the bodies of all components, but, also, the nodes located at the grid lines, which connect these components to each other. According to our investigation, the current algorithm generates a directional layered data structure very robustly, even in complex mesh domains. However, in the case of a complex element connection inside layers, it does not guarantee the physical neighborhood condition for some pairs of adjacent elements or nodes in the achieved data structure. These exceptions mostly require negligible search inside layers to find neighbor elements or nodes. Additionally, the current method facilitates the addressing procedure, because the search is confined to some neighboring layers instead of the entire grid data. To show the robustness of the method, the new data structure is constructed for an unstructured grid around a submarine cut out along its axial direction. The current evaluations show that the developed algorithm works very successfully, even in domains with complicated geometries and/or unstructured mesh distributions.
