Investigation of a metamodel-based methodology to optimise the component properties of discontinuous long fibre reinforced plastics by systematic design of manufacturing processes

不连续长纤维增强塑料(DLP)目前是纤维复合材料领域中最重要的材料类别之一。对于这些材料的加工，一次成形工艺通常用于生产近净形状的部件。在处理过程中，DLP发生流动。这会影响纤维取向，从而决定零件的机械性能。通过模拟可以预测由特定工艺配置产生的力学性能。然而，由于这些都是计算昂贵和耗时的，并不是所有可能的配置都能被模拟，这限制了达到期望的机械性能轮廓。该项目的目标是一种使用热固性和热塑性DLP压缩模塑中的初始占用状态的例子来优化DLP加工过程的方法。由于这一过程的高度复杂性和长的作用链，采用了成形对最终零件性能影响的元模型方法。通过元建模，可以复制通过绘制复杂模型或上下文的样本确定的关系。为了针对项目中使用的优化目标(扭转和三点弯曲下的零件变形最小化)优化工艺，在遗传算法的帮助下迭代地评估元模型，并通过附加工艺配置和所产生的变形来扩展。研究了优化方法的收敛行为对元模型抽象程度、零件几何形状、载荷情况和温度引起的材料效应的依赖关系，从而在两个不同的抽象层次上进行建模。在第一个抽象级别上，对输入参数和所产生的元件挠度之间的直接关系进行建模。这种方法也称为“黑箱”，它忽略了DLP在加工过程中的行为，以及流动和机械部件特性之间的复杂关系。在第二个抽象级别上，耦合了两个元模型。首先，建立了DLP过程中不连续流动的元模型，从而建立了材料行为本身的元模型。其次，使用黑盒方法对由此产生的组件行为进行建模。因此，对近过程元模型的影响进行孤立的分析是可能的。元模型的验证将通过交叉验证来执行。此外，还将对优化方法的普适性进行评估。

Discontinuous long fibre reinforced plastics (DLP) currently represent one of the most important material classes in the field of fibre composites. For the processing of these materials, primary forming processes are generally used to produce near-net-shape components. During processing, flowing of the DLP occurs. This influences the fibre orientations, determining the mechanical properties of the part. The mechanical properties resulting from a specific process configuration can be predicted by simulation. However, since these are computationally expensive and time-consuming, not all possible configurations can be simulated, which limits reaching a desired mechanical property profile.The aim of the project is a methodology for the optimisation of DLP processing processes using the example of the initial occupancy state in the compression moulding of thermoset and thermoplastic DLP. Due to the high complexity of this process and the long chains of action, the metamodeling approach of the shaping influences on the final part properties is pursued. By metamodeling, relationships can be replicated that were determined by drawing samples of a complex model or context. In order to optimise the process with regard to the optimisation goals used in the project (minimisation of part deformation under torsion and three-point bending), the metamodels are evaluated iteratively with the aid of a genetic algorithm and extended by additional process configurations and resulting deformations. The convergence behaviour of the optimisation approach is investigated with respect to its dependence on the degree of metamodel abstraction, part geometry, load case and temperature induced material effects.The modelling is therefore carried out on two different abstraction levels. On the first abstraction level, the direct relationship between input parameters and the resulting component deflection is modelled. This approach, also called "black box", neglects the behaviour of the DLP during processing as well as the complex relationships between flow and mechanical component properties. On the second abstraction level, two metamodels are coupled. Firstly, a metamodel of the discontinuous flow of the DLP during the process and thus of the material behaviour itself is created. Secondly, a black box approach used for modelling of the resulting component behaviour. Therefore, an isolated analysis of the influence of near-process metamodeling is possible. Validation of the metamodels will be performed by cross-validation. Additionally, an evaluation of the generalizability of the optimisation approach will be carried out.