Autonomous Self-Optimising Continuous Flow Reactors for Precision Polymer Synthesis

该项目旨在开发一个利用机器学习算法和连续流的平台；为各种高价值聚合物的高通量合成找到最佳条件。这种平台在小分子中得到了很好的研究，但在聚合物等大分子中却没有得到很好的研究。聚合的主要优化目标是低分散性和高转化率的聚合物，在过去的十年中，RAFT等方法的反应时间已经从几小时缩短到几分钟，使其适用于高通量流动系统。这是通过在某些溶剂中使用半衰期短的引发剂来实现的。许多复杂因素使得聚合难以在流动中进行，例如，层流速度流剖面影响分散性并引起污垢。该项目的目的是建立在当前多变量，多目标反应器的基础上，该反应器目前受到可用于优化的变量数量的限制。目前的反应器使用一锅试剂溶液，通过加热线圈泵送到在线GPC和台式NMR中，该算法找到了一组低分散性和高转化率的条件。由于该算法在一组训练点上工作以产生代理函数，因此不需要先验知识。该平台的开发将使用多个泵来进行优化，以找到理想的试剂比例，从而可以研究分子量和反应时间等其他目标，目的是形成一个“Dial-A-Polymer”平台，允许生产定制聚合物，并找到最佳条件，只需点击几下即可快速精确地制造它们。

This project aims to develop a platform that utilises machine learning algorithms and continuous flow; that can find optimum conditions for a variety of high value polymers for high-throughput synthesis. Such platforms are well studied for small molecules but less so for macro-molecules like polymers. The main optimisation objectives for polymerisations are low dispersity and high conversion polymers, in the last decade, the reaction times of methods such as RAFT have been reduced down from hours to minutes making them suitable for high-throughput flow systems. This is done by using initiators with short half-lives in certain solvents. Many complications render polymerisation difficult to carry out in flow, such as, laminar velocity flow profiles affecting dispersity and causing fouling. The aims of this project are to build upon a current multi-variable, multi-objective reactor that is currently limited by the number of variables that can be used in the optimisation. The current reactor uses a one-pot solution of reagents which are pumped through a heated coil into an inline GPC and benchtop NMR, the algorithm finds a set of conditions that give low dispersity and high conversion. As the algorithm works on a set of training points to produce a surrogate function no a priori knowledge is needed. The development of this platform would use multiple pumps to carry out optimisations to find the ideal reagent ratios so other objectives such as molecular weight and reaction time can be investigated - with an aim to form a "Dial-A-Polymer" platform, allowing custom polymers to be produced and to find the best set of conditions to make them rapidly and precisely with a few clicks.