Discrete computational modelling of twin screw granulation

许多工业加工操作依赖于作为细粉的饲料，这些细粉的处理特性很差，必须通过造粒进行精馏，以形成更粗的颗粒。通常采用湿法制粒，在批处理过程中，通常在混合器中将粘结剂添加到粉末中。连续双螺杆造粒(TSG)具有很大的潜力，例如在制药行业，因为它在生产能力和设备设计方面具有灵活性、重复性、停留时间短、液固比较小，而且还能够造粒难以加工的配方。然而，仍然有一些重大的技术问题限制了它的广泛使用，需要对这一过程有更多的了解，以满足监管要求。此外，封装的原料药(活性药物成分)越来越引起人们的兴趣，开发一种不损害此类胶囊的TSG工艺将大大扩展其应用范围。由于原料药的高成本，TSG的试验性优化是昂贵的，而且往往不是最优的，并且不会导致对该过程的更一般的理解。对过程中单个进料颗粒的行为进行计算建模将克服这些限制。离散单元法(DEM)是应用最广泛的方法，但很少应用于TSG挤出机中的颗粒数量(~5500万)，这种例子涉及简化的颗粒间相互作用，例如通过假设颗粒是光滑的和球形的，并且任何液体都以离散的桥的形式存在，而不是与造粒相关的更大的饱和态。该项目将以多尺度战略为基础，开发更能代表真实系统的高级相互作用规律。膨胀颗粒粘结剂(如微晶纤维素)的整体和界面特性将使用粗粒度分子动力学进行建模，以将输入导出到包括硬颗粒和粘性聚合物粘结剂(羟丙基纤维素)的配方的中尺度有限离散元模型中。具有粘性粘结剂配方的弹性颗粒(例如乳糖和胶囊)将使用快速多极边界元方法进行建模。这些微观和中观模型将被用来为TSG的DEM模型提供闭合。它将涉及与中国科学院的合作，中国科学院率先将具有GPU集群的大规模并行高性能计算应用于离散建模，如DEM，尽管现有的相互作用定律更简单。将部署一项广泛的实验方案，以衡量实物投入并验证模型。对于所考虑的配方，TSG的螺杆设计和操作条件将使用DEM进行优化，并对结果进行经验验证。优化标准将包括颗粒大小分布、由乳糖配方生产的颗粒的片剂质量以及对胶囊的损害降至最低。主要的好处将是为TSG提供一个建模工具箱，以实现更快、更具成本效益的优化，并允许处理封装的API。详细的数据后处理将阐明将用于制定制度绩效图的机械信息。多尺度模型将应用于广泛的多相体系，例如很大一部分消费品、挤压工艺的催化剂糊和农药等农产品。微观和介观方法在研究乳状液中硬粒子和软粒子以及液滴的整体和界面行为方面有广泛的应用。高级建模和在大规模并行高性能GPU集群上的实施相结合，将允许前所未有的应用程序应用于极其复杂的多阶段系统。

Many industrial processing operations depend on feed materials that are fine powders with poor handling characteristics, which have to be rectified by granulation to form coarser granules. Generally wet granulation is employed, in which a binder is added to the powder in a mixer usually in batch processes. Continuous Twin Screw Granulation (TSG) has considerable potential, eg in the pharmaceutical sector, because of the flexibility in throughput and equipment design, reproducibility, short residence times, smaller liquid/solid ratios and also the ability to granulate difficult to process formulations. However, there remain significant technical issues that limit its widespread use and a greater understanding of the process is required to meet regulatory requirements. Moreover, encapsulated APIs (Active Pharmaceutical Ingredients) are of increasing interest and the development of a TSG process that did not damage such encapsulates would significantly extend applications. Experimental optimisation of TSG is expensive and often sub-optimal because of the high costs of APIs and does not lead to a more generic understanding of the process. Computational modelling of the behaviour of individual feed particles during the process will overcome these limitations. The Distinct Element Method (DEM) is the most widely used method but has rarely been applied to the number of particles in a TSG extruder (~ 55 million) and such examples involve simplified interparticle interactions e.g. by assuming that the particles are smooth and spherical and any liquid is present as discrete bridges rather than the greater saturation states associated with granulation. The project will be based on a multiscale strategy to develop advanced interaction laws that are more representative of real systems. The bulk and interfacial properties of a swelling particulate binder such as microcrystalline cellulose will be modelled using Coarse-grained Molecular Dynamics to derive inputs into a meso-scale Finite Discrete Element Method model of formulations that include hard particles and a viscous polymeric binder (hydroxypropylcellulose). Elastic particles (e.g. lactose and encapsulates) with viscous binder formulations will be modelled using the Fast Multi-pole Boundary Element Method. These micro- and meso-scale models will be used to provide closure for a DEM model of TSG. It will involve collaboration with the Chinese Academy of Science, which has pioneered the application of massively parallel high performance computing with GPU clusters to discrete modelling such as DEM, albeit with existing simpler interaction laws. An extensive experimental programme will be deployed to measure physical inputs and validate the models. The screw design and operating conditions of TSG for the formulations considered will be optimised using DEM and the results validated empirically. Optimisation criteria will include the granule size distribution, the quality of tablets for granules produced from the lactose formulation and the minimisation of damage to encapsulates. The primary benefit will be to provide a modelling toolbox for TSG for enabling more rapid and cost-effective optimisation, and allow encapsulated APIs to be processed. Detailed data post-processing will elucidate mechanistic information that will be used to develop regime performance maps. The multiscale modelling will have applications to a wide range of multiphase systems as exemplified by a large fraction of consumer products, catalyst pastes for extrusion processes, and agriculture products such as pesticides. The micro- and mesoscopic methods have generic applications for studying the bulk and interfacial behaviour of hard and soft particles and also droplets in emulsions. The combination of advanced modelling and implementation on massively parallel high performance GPU clusters will allow unprecedented applications to multiphase systems of enormous complexity.