Sudden machine downtimes in production are a nightmare for companies. The longer the disruption lasts, the more expensive it ends up being. It also affects manufacturers' delivery reliability and reduces their competitiveness. "These are usually relatively minor defects or signs of wear," explains Ulrich Moser, Head of Marketing at IEF-Werner from Furtwangen in the Black Forest. "However, if these are not recognized in time, the degree of wear increases and the machine or component no longer works precisely. At some point, it will fail completely."

What if it could independently monitor its status, recognize problems and weak points and inform the responsible employee in good time? Or even better: if it could rectify the effects of problems itself or communicate with other components? Production would continue without expensive machine downtimes and service technicians could schedule maintenance measures in good time. Wear and tear could be reduced and the service life of the components significantly extended. The automation specialist has been working on precisely this topic as part of the EU research project Selsus, which was led by the Fraunhofer Institute for Manufacturing Engineering and Automation IPA in Stuttgart.

The project is called "Health Monitoring and Life-Long Capability Management for Self-Sustaining Manufacturing Systems", which describes the ability of machines and components to monitor and optimize themselves to a certain extent and thus maintain production. The IPA scientists worked together with partners from industry and science on a technology that uses intelligent software and sensor networks to detect weak points and signs of wear at an early stage and thus predict failures. Selsus was funded by the European Commission with almost 5.4 million euros.

Selcomp - the intelligent unit

"With the ability to optimize themselves, the components in production lines no longer perform an isolated task. Instead, they are a dynamic, versatile and specialized part of a structure of intelligent work cells," explains Ulrich Moser. As part of the project, these logical units are called Selcomps: They collect, store and analyze process data. As autonomously operating systems, they are therefore able to evaluate their own condition and performance and recognize the degree of wear and tear without external influence.

But how can such components be designed? "To find out, we developed our own Selcomp from a toothed belt axis and installed it in a demonstrator," says Moser. IEF-Werner deliberately opted for the linear unit because it has proven itself in numerous handling tasks. Driven by a servo motor, it works efficiently and cost-effectively. The integrated guide can also be used in slightly contaminated environments and the guide elements work almost maintenance-free for up to 10,000 kilometers. With this component, the toothed belt enables high acceleration and speeds with short cycle times. For the construction of a collaborative demonstrator, the axis was mounted on a stable frame on which an intelligent welding unit from partner HWH is also integrated. The HWH-Selcomp consists of a welding controller and welding gun - with sensors for measuring temperature, current, voltage and force. This exemplary scenario represents the widely occurring application of a moving component in cooperation with a production process.

High demands lead to wear

"The data specified by the manufacturer regarding the service life of a machine component are usually only guide values that can vary greatly during use," explains Moser. Different loads, such as those caused by rapid starting and stopping, can lead to abrasive, adhesive or fatigue wear on the contact surfaces of the slides and guideways or in the drive bearings. Accuracy continues to decrease and the risk of failure increases. To ensure that the axle can always run in optimum condition, it was necessary to pinpoint the exact points where wear occurs. The developers initially installed additional sensors. "This enabled us to identify critical points in the motor and carriage bearings, among other things," explains Moser. After extensive tests, the IEF technicians realized that the necessary information could not be collected with additional sensors alone. It is easier and, above all, more cost-efficient to do this by indirect determination on the basis of known data, such as the motor current.

The task now was to develop a powerful model that would calculate the practical service life of the timing belt axis from the permanently collected data. "To solve this task, we worked closely with the ICT provider Hugin Expert," says Moser. The Danish company specializes in Bayesian networks - a mathematical method that can be used to calculate the probability of a certain event or condition occurring. This method takes into account several variables and the possibilities associated with them, for example that a certain cable subject to high stress will break in the near future.