Development of a large-scale parallel and dispersed computation system of complexed heat transfer problems

传热传质问题通常是由各种不同的传递过程组成的复杂系统。为了得到它们的计算解，既要求解各个相关过程，又要将它们的解耦合成一个完整的系统。为此，应根据传导、对流和辐射等输运过程将问题分解，并交换边界条件来求解。在本研究中，过程的各个领域由该领域的研究人员和专家独立求解，并通过传递边界条件的信息，使分离的解决方案成为一个完整的解决方案。每个研究者的子系统也被分解成更多的子系统，以加速使用晶片机的计算。他们集成所需的计算结果通过大学之间的通信系统JUNET进行交换。可交换边界条件下离散的个体计算和不同解的集成是两个需要解决的课题。作为问题的一个例子，半导体晶体生长的过程中的直拉法进行了研究，因为它是高度非线性耦合的基本过程的热量和质量传递。结果表明，子系统边界条件的交换和施加是控制整个系统集成速率的重要因素。

The heat and mass transfer problems usually consist of complexed systems of various different transport processes. In order to obtain computational solution of them, it is necessary to solve the individual related processes as well as to couple their solutions into an integrated system. For this purpose, it should be solved by dividing the problem into parts depending on the transport process such as conduction, convection and radiation and by exchanging the boundary conditions each other. In the present research, the individual fields of the process are solved independently by the investigators, specialists in the field, with communicating the information of the boundary conditions which make the separated solutions into an integrated one. The subsystems of each investigator are also made to be decomposed into further subsystems in order to accelerate the calculation by using the transputers. Their computed results necessary for the integration are exchanged through the communication system between universities, JUNET. Two subjects to be solved are the individual computation dispersed under the communicable boundary conditions and the integration of different solutions. As an example of the problem, the process of semiconductor crystal growth by the Czochralski method is studied because it is highly non-linearly coupled with fundamental processes of heat and mass transfer. As the results, it is found that exchanging and imposing the boundary conditions of subsystem should be the most important factor for rate-controlling of the integration of the whole system.