Increasing the throughput of laser-induced shock wave indentation testing by an adapted measurement strategy and data evaluation based on machine learning

点击翻译按钮获取中文摘要

It has already been proven in the past that material properties can be efficiently determined by means of laser-induced shock wave indentation testing. The aim of the project at hand is to increase the throughput of the measurement process on the one hand and to expand the range of material properties that can be determined on the other. The basis of the measurement technology is a confocal sensor that measures the indentation geometry, whereby in particular forming speed-dependent effects can also be recorded. A measurement strategy to be developed will enable the measurement method as a whole to be used for high throughput. In addition, further suitable descriptors are to be identified. An integral part of the measuring system is the evaluation of the results of the confocal sensor by means of neural networks. Here, the recorded data is the input variable and the material parameters are the output. The neural networks perform this task very efficiently based on an appropriate configuration to be worked out in the project. However, their training requires a very large number of available data sets. In order to avoid having to determine these experimentally at great effort, a highly efficient finite element simulation of the indentation process is being developed, which provides the required data. At the same time, the simulation enables various new evaluation approaches, some of which also include neural networks. The evaluation of these different approaches, e.g., with regard to accuracy and computing time, is another component of the project at hand. Finally, the developed measurement methodology will be tested and evaluated using a benchmark.