Tool Condition Monitoring in Gear Hobbing

Due to its high productivity, gear hobbing is one of the most frequently used manufacturing processes in the soft machining of cylindrical gears. One of the major objectives of an optimized manufacturing process is to maintain the required component quality while reducing the manufacturing costs and/or increasing productivity. In both cases, the knowledge of the tool wear condition is of great importance. On the one hand, the component quality decreases with increasing tool wear. On the other hand, the tool has to be replaced before a critical wear condition leads to an unusable state. However, changing a tool too early might also result in increased costs, as non-productive time increases. Knowledge of tool condition is therefore indispensable for an optimized utilization of tool life and component quality. Tool Condition Monitoring (TCM) offers a methodical approach to monitor tool wear during the process.

Figure 1
Source: WZL RWTH Aachen University

The first prerequisite for a valid implementation of TCM is a detailed understanding of the process. Based on the geometry of the workpiece and tool as well as the process parameters, the chip geometries in the process can be determined with penetration calculations. In

addition to the chip characteristics, the theoretical tool engagement frequencies and other process information such as tool substrate and material must be taken into account.

In order to investigate the next steps in the implementation of TCM, different sensors were investigated, both on the tool and on the workpiece side. When selecting and mounting the sensors, it is important to ensure that the frequency ranges of the sensors can detect the frequencies that occur. The assembly location of the sensors should be as close as possible to the cutting point, so that the attenuation of the signals by intervening machine parts is not too great.

The second major point in addition to the necessary process understanding is the correct signal analysis and interpretation. First, it is necessary to filter the signals for any interfering signals that do not come from the cutting process itself. Once a clean signal is available, it can be analyzed using many different characteristics in the time or frequency domain. Finally, the correct influence parameters must be assigned to the various signal characteristics. The assignment is either done manually or can be found by using artificial intelligence (AI) methods. Specifically designed neuronal networks record all influencing factors and constantly optimize the accuracy of the results with increasing amounts of data.

For the investigations, a gear hobbing machine Liebherr LC180 was equipped with various sensors. The sensors included tool- and workpiece-side acceleration and acoustic emission sensors. An airborne sound sensor was placed in the installation space of the machine. In addition to the sensors, the performance diagnostics of the machine control system were also evaluated.

The influence of wear, process parameters and material on the signal data was investigated.

The results showed that the combination of several sensors is useful for recording all relevant influences. Changes in the signal data due to various influences could be detected for all sensors, with the exception of the structure-borne sound sensor on the workpiece side. The results of this sensor could not be used due to the large distance between the mounting location and the cutting point.

An example result is shown in Figure 2. The AE signal on the tool side is shown for a new and for a damaged tool. The data is plotted in the time domain at the top and in the frequency domain at the bottom. In the time domain, no difference can be observed between the new and the damaged tool in the time domain. In the frequency domain, two distinct peaks can be seen. For the new tool, these peaks are at approx. f = 80 kHz and f = 120 kHz. With the damage, the first peak shifts forward to a frequency of about f = 70 kHz. The second peak remains at the frequency of f = 120 kHz. However, the second peak is larger in magnitude after the damage than without. From this, it can be concluded that the damage has an influence especially on higher frequencies.

Figure 2
Source: WZL RWTH Aachen University

For a stable tool condition monitoring, further research is needed in the future to investigate the correlation between the sensor signals and the tool condition. With the knowledge of the interdependencies in combination with a reliable algorithm, tools can be monitored for different gear use cases.


Steffen Hendricks, M.Sc., Research Assistant Gear Technology, Laboratory for Machine Tools and Production Engineering WZL of RWTH Aachen University

Nico Troß, M.Eng., Team Leader Gear Soft Machining, Laboratory for Machine Tools and Production Engineering WZL of RWTH Aachen University

Dr.-Ing. Jens Brimmers, M.Sc., Chief Engineer Gear Technology, Laboratory for Machine Tools and Production Engineering WZL of RWTH Aachen University

Professor Dr.-Ing. Thomas Bergs, Head of Chair of Manufacturing Technology, Laboratory for Machine Tools and Production Engineering WZL of RWTH Aachen University, Fraunhofer Institute of Production Technology