Continuous generating gear grinding is a well-established and widely used process in the industry for large-scale production gears. It offers an economic/efficient process for finishing gears, which shapes the micro-geometry of the gear tooth flank and improves its surface quality. The resulting quality of ground gears depends on several factors, namely the tool performance, the machine stability as well as the correct clamping/positioning of the workpiece. The grinding step is very crucial since it has a direct impact on the operating quality of gears and in particular on the running noise behaviour of the end product. The potential of online vibration-based gear grinding monitoring has been explored and demonstrated in the previous work (Gryllias et al., 2017) as a means of quality control that could lead to the overall reduction of production losses and to the prevention of sending defective parts to customers. Several features which could be used to monitor the grinding processes and to identify a specific type of defects have been proposed & experimentally validated to some extent. The types of faults include (i) high feed rate, (ii) high infeed, (iii) non-flat workpiece, and (iv) eccentric workpiece. However, a further investigation on a new test campaign revealed that none of the features developed in (Gryllias et al., 2017) was robust enough to detect eccentric workpieces during the grinding process. It is worth mentioning here that an eccentric workpiece fault is unlikely to happen, but it is analogous to a runout on the incoming workpiece quality. In this paper, a qualitative model to predict the vibration signature due to eccentric workpieces/runouts is developed and discussed. Based on the qualitative understanding, a novel feature to detect eccentric workpieces/runouts during gear grinding processes based on vibration signals has been developed. The newly developed feature has been validated on real vibration signals captured during the emulation of process malfunctions on an industrial gear grinding machine. In this study, it is shown that the novel feature seems sensitive and robust for detecting workpiece eccentricities/runout errors of about more than 60 µm. It is also shown in this study that the feature is insensitive to other types of gear grinding faults, which is important for diagnostics/root-cause analysis purposes.

连续发电齿轮磨削是大规模生产齿轮行业中公认的且广泛使用的过程。它为齿轮的精加工提供了经济/高效的方法，可改变齿轮齿面的微观几何形状并改善其表面质量。最终齿轮的质量取决于几个因素，即刀具性能，机器稳定性以及正确的工件夹紧/定位。磨削步骤至关重要，因为它直接影响齿轮的运行质量，特别是最终产品的运行噪音。在线振动基于齿轮磨削监控的潜力已在先前的工作中进行了探索和展示（Gryllias等，2017年）作为一种质量控制手段，可以导致生产损失整体减少，并防止将有缺陷的零件发送给客户。已经提出了几种可用于监视磨削过程并识别特定类型缺陷的功能，并在一定程度上进行了实验验证。故障类型包括（i）高进给速度，（ii）高进给，（iii）非扁平工件，以及（iv）偏心工件。但是，对新测试活动的进一步调查显示（Gryllias等人，2017）中开发的功能都没有足够强大的功能来检测磨削过程中的偏心工件。在这里值得一提的是，偏心工件故障不太可能发生，但它类似于进入工件质量的跳动。在本文中，开发并讨论了定性模型来预测由于偏心工件/跳动引起的振动信号。基于定性的理解，已开发出一种新颖的功能，可基于振动信号在齿轮磨削过程中检测偏心工件/跳动。新开发的功能已经通过在工业齿轮磨床上模拟过程故障期间捕获的真实振动信号进行了验证。在这项研究中，表明该新颖功能对于检测大约60 µm以上的工件偏心率/跳动误差似乎很敏感且很可靠。在这项研究中还表明，该功能是对其他类型的齿轮打磨故障不敏感，这对于诊断/根本原因分析很重要。