An automatic tool to generate and solve zone models of chemically reacting flow

流体流动与化学或物理转化结合在一起在工程中无处不在，例如燃烧、结晶和反应器设计，但也模拟人体内的药物摄取。由于空间全分辨模拟往往过于昂贵(很容易达到几天或几周的求解时间)，分区模型是一个重要的成熟工具。它们将几何图形(如管道或间歇反应器)分解为几个区域。对于每个区域，化学和物理现象的解决与跨区域之间的材料传输相结合。化学反应通常是复杂的，涉及数十或数百种物质以多种方式进行反应。由于反应发生在非常不同的时间尺度上，所产生的系统在数学上是僵硬的，难以求解，现有的标准方法往往非常低效，使得近实时的模拟非常困难。该项目将开发一种语言来高效地描述区域模型，并结合一个工具来自动创建变换和传输方程，并使用从一系列可能的选择(Single Implative、Rosenbrock、IMEX等)中自动选择合适的数值求解器来求解这些方程。该项目的目的是创建一个工具的原型，该工具可以自动生成分区模型网络的方程，并选择一个有效的求解器来近实时地生成解。将完成以下目标：1.创建一个可以自动生成分区模型的方程并生成相应C代码的工具。回顾现有数值求解器的文献，并根据已报道的方程模拟化学反应的性能对它们进行排名。选择最有前景的五种方法，并将它们与代码生成工具集成。设计启发式算法为生成的模型选择最有效的方法，以便在保持可接受的精度的同时尽可能快地求解该模型。

Fluid flows coupled with chemical or physical transformations are ubiquitous in engineering, examples being combustion, crystallisation and reactor design, but also modelling drug uptake in the body. Since spatially fully resolved simulations are often too expensive (easily reaching solution times of days or weeks), zonal models are an important established tool.They decompose the geometry, e.g. a pipe or a batch reactor, into zones. For each zone, chemical and physical phenomena are solved in combination with transport of materials across interfaces between zones.The chemical reactions are often complex, involving dozens or hundreds of species reacting in a multitude of ways. Because reactions happen on vastly different time scales, the resulting systems are mathematically stiff, difficult to solve and existing standard methods are often very inefficient, making near real-time simulations very difficult.The project will develop a language to describe zonal models efficiently in combination with a tool that automatically creates transformations and transport equations and solves these using an automatically selected appropriate numerical solver from a range of possible choices (single implicit, Rosenbrock, IMEX, etc). The aim of the project is to create the prototype of a tool that can automatically generate the equations for a network of zonal models and select an efficient solver to produce solutions in near real-time.The following objectives will be completed to achieve the aim:1. Create a tool that can automatically generate the equations for a zonal model and generates corresponding C code.2. Review literature of available numerical solvers and rank them based on reported performance for equations modelling chemical reactions.3. Select the five most promising methods and integrate them with the code generation tool.4. Devise heuristics to select the most efficient method for a generated model to solve it as fast as possible while maintaining acceptable accuracy.