Greybox model-based prediction of wear evolution of coated tools through experimental and model-driven identification of relevant loads

涂层硬质合金刀具（CCCT）广泛应用于机械加工，表现出复杂的瞬态磨损行为。在目前的技术水平下，这种磨损行为的分析和建模是通过确定性（白盒）和数据驱动模型（黑盒）的孤立使用进行的，每种模型都通过实验调查的帮助进行验证。然而，通过这种方式，由于安装磨损过程，CCCT的行为不能在整个使用寿命期间以足够的精度建模。在此背景下，本课题旨在通过实验和模型驱动的相关载荷水平识别，实现基于灰盒模型的磨损演变预测，从而大幅提高CCCT全寿命使用行为的预测质量。为了实现这一目标，首先在一台生产车床上进行了广泛的实验研究。通过记录原位和非原位测量数据，可以精确描述CCCT的初始状态，并对磨损过程进行详细监测，直到使用寿命结束。使用数据驱动的异常检测进行缺陷检测，在先前输入的训练数据的帮助下，对过程力和噪声排放进行现场监测和分析，有助于确定整个使用寿命（黑箱）中的关键负载范围。使用复杂的非原位测量技术，可以在关键载荷范围内显示原位测量数据与工具当前磨损状况之间的相关性。通过实验研究验证的数值切屑形成模拟（白盒）用于量化无法通过实验确定的状态变量（例如切屑和刀具之间的相对速度或法向接触应力）。基于临界载荷水平下CCCT的磨损情况，对数值模拟中的磨损率进行调整，从而整体实现灰盒建模。该程序可以预测CCCT在整个使用寿命期间的行为，以及在其使用寿命结束时的内在随机失效，其精度迄今尚未达到。

Coated cemented carbide tools (CCCT) used extensively in machining exhibit a complex, transient wear behavior. At the current state of the art, the analysis and modelling of this wear behavior is carried out through the isolated use of deterministic (white box) and data-driven models (black box), each of which is validated with the aid of experimental investigations. In this way, however, the behavior of CCCT cannot be modelled with sufficient accuracy over the entire service life due to the instationary wear progress. Against this background, the aim of this project is to realize a prognosis of the wear evolution based on a greybox model through the experimental and model-driven identification of relevant load horizons and thus to achieve a considerable improvement in the prognosis quality of the service behavior of CCCT over their entire service life. In order to achieve this goal, extensive experimental investigations are first carried out on a production lathe. By recording in-situ and ex-situ measurement data, a precise description of the initial condition of the CCCT as well as a detailed monitoring of the wear progress until the end of the service life is achieved. The use of a data-driven anomaly detection for defect detection, which carries out an in-situ monitoring and analysis of the process forces and noise emissions with the help of previously fed training data, serves to identify critical load horizons over the entire service life (blackbox). The use of complex ex-situ measurement techniques then allows correlations between the in-situ measurement data and the current wear condition of the tool to be revealed at the critical load horizons. Numerical chip formation simulations, which are validated by experimental investigations (white box), are used to quantify state variables that cannot be determined experimentally (e.g. relative speed between chip and tool or normal contact stress). Based on the wear condition of the CCCT characterized at the critical load horizons, an adjustment of the wear rate in the numerical simulations is provided, thus implementing the greybox modelling as a whole. This procedure makes it possible to predict the behavior of the CCCT over its entire service life as well as the immanent stochastic failure at the end of its service life with a precision that has not been achieved so far.