Prediction of surface conditions for robust control of a turning process based on in-process data acquisition and data driven soft sensor approach

表面决定了工件的功能、阻力和寿命。一方面，它确定了工件的几何形状，并结合其形貌确定了与其他工件的接触面。由于表面也暴露在环境中，因此它同时受到腐蚀过程的影响。此外，裂纹的形成和扩展以及抗塑性变形是由表面层的性能（残余应力、硬度和显微组织）决定的。计划项目的主要目标是将加工工件的地形演变，残余应力和硬度与工艺参数，过程中的干扰以及表面层的初始条件联系起来。待开发的模型将基于硬车削产生的数据。加工过程中测量加工力和局部温度。现有的工艺后表征技术，如XRD残余应力分析，将被微磁3MA方法系统地取代，从而能够在工艺中确定表面层的性质。必须使用一套合适的样品进行校准。该传感器对各种材料特性的独特灵敏度使得校准要求很高。利用系统辨识的方法，在获取数据的基础上，生成非线性经验过程模型。由于所有相关现象的物理建模非常复杂，因此提供了不适合控制设计的模型结构，因此这些模型不容易转移。最后，刀具磨损的建模和在线估计对于第二个资助期的设想控制设计是必要的。因此，软传感器设计是第二个建模任务。软传感器将能够根据工艺变量和初始材料特性来预测表面层的特性。软传感器将从从不同硬度的工件和几个定义的刀具磨损阶段获得的数据中得出，这样在预测表面层性能时就会考虑到这两者。通过模型反演，可以计算出被操纵变量的所需值。

The surface determines functionality, resistance and lifespan of a workpiece. On the one hand it defines the workpiece geometry and in combination with its topography the contact surface with other workpieces. As the surface is also exposed to the environment, it is at the same time subjected to corrosive processes. Furthermore, crack formation and propagation as well as resistance against plastic deformation are determined by the properties (residual stress, hardness and microstructure) of the surface layer.The primary goal of the planned project is to correlate topography evolution, residual stress and hardness of a machined workpiece to the process parameters, the disturbances within the process as well as the initial condition of the surface layer. The model to be developed will be based on data generated by hard turning. Machining forces and local temperatures will be measured in-process. Established post-process characterization techniques such as XRD residual stress analysis will be systematically substituted by micro-magnetic 3MA approach to enable in-process determination of the surface layer properties. Calibration by means of a suitable set of specimens is necessary. The sensor’s distinct sensitivity to various material properties makes calibration highly demanding.A non-linear empirical process model will be generated on basis of acquired data using methods of system identification. As physical modeling of all relevant phenomena is very complex and, thus, provides for model structures inappropriate for control design, these models cannot easily be transferred. Finally, modelling and online estimation of tool wear is necessary for envisaged control design in the second funding period. Thus, soft sensor design is the second modelling task. The soft sensor will be able to predict surface layer properties based on process variables and initial material properties. The soft sensor will be derived from data obtained from workpieces of varying hardness and several defined stages of tool wear, such that both are considered when surface layer properties are predicted. By means of model inversion the required values of the manipulated variables can be computed.