Quality assured flow production of light UHPC bar elements using artificial neural networks

建筑仍然受到工艺的强烈影响。建设项目的价值主要发生在建设过程中。由于天气等外部条件的影响，施工过程容易出错。此外，在许多情况下，对建筑工人手工技能的依赖导致了结构质量方面的不满意结果。施工时间长是就地制造的另一个后果。预制混凝土构件可以解决部分现场生产问题。然而，与现场混凝土施工类似的复杂生产步骤（准备钢筋和模板，剥离，清洁模板）导致类似的问题。模块化设计应该会导致更快和更便宜的建设。通过将可能的组件简化为元素目录，可以实现更高的相似性，从而实现经济的流动生产过程。带有节点和杆的桁架应允许在短时间内以高精度安装支撑结构。提出的研究重点是用挤压法生产棒材元件。由建筑材料研究所（IfB）开发的用碳强化的超高强度混凝土（UHPC）制成的空心圆柱体被挤出。该元件的特点是特别轻的重量，高屈曲稳定性和延展性。挤出UHPC需要高水平的质量保证。在拟议的研究项目中，将由静力学与动力学研究所（ISD）开发的人工神经网络（ANN）形式的连续、自我学习质量保证。其目的是开发一种UHPC挤出工艺，该工艺“从第一天开始”就包含传感器概念，即从制造的第一步开始。人工神经网络使用测量数据来监控生产过程。为此，人工神经网络通过从“健康”挤压过程中获得的测量数据进行训练。此外，在某些过程中故意引入错误，以“训练”人工神经网络识别这些错误并推荐对策。最后，在统计分析中，通过提出的质量保证框架对安全增益进行量化。工作计划提供了以下工作包：-开发一种挤压uhpc的方法-开发一个传感器概念- UHFB挤压实验-人工神经网络的训练-错误过程的模拟。-安全收益的定量评估

Construction is still strongly influenced by workmanship. The value of a construction project is mainly taking place at the construction. Due to external conditions such as weather, the construction process is prone to error. Furthermore, the dependence on the manual skill of construction workers leads in many cases to unsatisfactory results in terms of quality of the structures. Long construction times are yet another consequence of in-situ manufacturing. Precast concrete elements can solve a part of the production problems on site. However, similar complex production steps as with in-situ concrete construction (preparing reinforcement and formwork, stripping, cleaning formwork) lead to similar problems.A modular design should lead to faster and cheaper construction. By reducing the possible components to a catalog of elements, higher similarities can be possible, allowing an economic flow production process.Trusses with their nodes and bars should allow the erection of supporting structures with high precision in a short time. The focus of the proposed research is the production of bar elements by extrusion. Hollow cylinders made of ultra-high-strength concrete (UHPC) with carbon reinforcement, developed at the Institute for Building Materials (IfB), are extruded. This element is characterized by a particularly low weight, a high buckling stability and a ductile behavior. Extruding UHPC requires a high level of quality assurance. In the case of the proposed research project, a continuous, self-learning quality assurance in the form of an artificial neural network (ANN), which is to be developed by the Institute of Statics and Dynamics (ISD).The aim is to develop an extrusion process for UHPC that incorporates a sensor concept "from day one", i.e. from the first steps of manufacturing. The measured data is used by the ANN to monitor the production process. For this purpose, the ANN is trained by measured data obtained from “healthy” extrusion processes. Furthermore, errors are deliberately introduced in some processes to “train" the ANN to recognize such and recommend countermeasures. Finally, in a statistical analysis, the safety gain is quantified by the proposed quality assurance framework. The work program provides for the following work packages:-Development of a method of extruding UHPC-Development of a sensor concept- Experiments to UHFB extrusion- Training of the ANN- Simulation of erroneous processes.-Quantitative assessment of the safety gain