If you can predict when a production process will go wrong, you can save much time and money. VCST, a globally leading manufacturer of high-tech components for the automotive industry, developed an online monitoring system together with Flanders Make. This system, based on cheap hardware and smart algorithms, prevents failures during the production of gears and detects process anomalies. In this way, the quality of the end product is ensured.
Low-cost vibration monitoring allows for online gear manufacturing anomaly detection
Stringent requirements on engine noise emission and increasing pressure on production while maintaining a virtual zero scrap rate is pushing companies towards procedures that need a permanent monitoring of the production process.
Specifically for VCST, this requires strict monitoring during the final phase of the gear manufacturing process, i.e. a finishing operation shaping the microgeometry of the gear tooth flank and improving its surface quality. This decisive step has a direct impact on the operating quality of gears and in particular on the operating noise emission of the end product.
VCST primarily uses a grinding process for its high finishing quality. However, some anomalies during this process can lead to a serious deterioration of the quality of the manufactured parts. This deterioration in quality only comes to light after having performed a quality control on sampled parts of the production batch. The permanent monitoring of the grinding process enables to scrap individual pieces instead of the whole of their respective batches.
Fig. 1 Grinding process and quality control of gears.
Needs and Constraints
To maintain the aimed at workpiece quality, VCST wants to be able to identify these anomalies during the actual grinding process. The development of a tailored solution faced one major constraint: the process status should be known within the time frame needed for grinding one gear (i.e. max. 60 s). This requires fast and robust processing algorithms.
Within the scope of the VibMon project, Flanders Make and VCST developed an indirect, vibrations-based method. The algorithms allow to:
- detect the grinding pass and identify the vibration signal segments that are most relevant for extracting detection features;
- process the signal to assess the health status of the grinding process.
Tests and Results
The developed method was tested on data acquired on a grinding machine on which different anomalies leading to quality problems were introduced (non-flat and eccentric incoming parts, high radial feed and high axial feed rate). The high axial feed rate generates an increased grinding force because more material is removed. Vibrations recorded using two tri-axial accelerometers were processed to extract the monitoring features.
Fig. 2 Grinding machine used for tests
The results showed that it is possible to detect grinding malfunctions during the actual grinding process. The following figures illustrate for two cases, how an abnormal grinding process can be distinguished from a normal one (Fig. 3).
Fig. 3 Gear grinding anomalies detected following the introduction of certain features.
The execution of the algorithms meets the time constraints. Fault features can be extracted online during the grinding process. This enables implementing the algorithms on a low-cost embedded platform with limited computational power (Fig. 4). This embedded platform was also developed in the VibMon project by CoSys-Lab (University of Antwerp). It is capable of acquiring and processing data as well as generating alarms whenever a process anomaly is detected. The low-cost aspect of the platform will allow for its deployment on a fleet of production machines.
Fig. 4 Low-cost embedded platform
Together, VCST and Flanders Make are making online monitoring of gear quality possible through the detection of process anomalies. VCST plans to deploy online monitoring on all grinding machines in its production plant to maximise machine utilisation while at the same time warranting the quality of the end product and significantly reducing scrap rates.