Related Vendor

FAULHABER Drive Systems

The Human Hand Serves as a Model for an End-Effector

The specially developed end-effector (lead image) plays a central role in this process. During manual grinding, the human hand automatically compensates for small positional deviations. It aligns the grinding wheel with the surface while simultaneously maintaining the required contact force. These capabilities are mechanically replicated in the “AutoFIN” end-effector. The so-called Cube system enables pneumatic force and displacement compensation. This allows positional deviations to be compensated for and an adjustable contact force to be applied to the surface as consistently as possible. Beneath this lies the actual “robot hand.” It has two rotational degrees of freedom and allows the grinding wheel to align itself independently with the local surface orientation. In this way, minor inaccuracies and the effects of wear are compensated for directly at the tool. A purely software-based correction would be too complex for this purpose and could only partially account for the deviations that occur in the actual process. The modular concept thus combines the precision of the robot with mechanical compensation modeled after the flexibility and adaptability of the human hand. The functionality of this system has already been tested in several trials using actual components from BMW’s toolmaking department.

The Path from Tool Geometry to a Safe Grinding Path

Aside from the hardware, it is primarily the software that determines whether automation can be economically viable for a batch size of one. Manually programming a robot for each individual press die would be far too time-consuming. Therefore, over the course of the project, an end-to-end software chain is developed that automatically generates an executable and validated robot program based on the part geometry. First, the die is automatically captured using a 3D scanner and localized within the robot’s workspace. Next, a segmentation algorithm extracts the grinding surfaces from the CAD file and divides them into areas that are geometrically and process-technically related. In this process, distinctions are made between flat surfaces, radii, edges, and concave and convex areas. Appropriate machining strategies can then be selected for each of these segments. Based on this, the path planning generates the movements of the grinding tool across the tool surface.

Only when Nothing Can Go Wrong Anymore is it Time to Put it into Practice

In addition to simple parallel paths, the user can also select cross-grinding or machining directions that follow the contour. Cross-grinding (see previous image), for example, can level out milled surfaces from different directions and produce a more uniform grinding pattern. However, even before the paths are converted into an actual robot program, they are digitally validated. A kinematic solver developed specifically for this purpose analyzes the grinding paths and assesses whether certain areas are accessible or if critical robot positions could occur. A digital twin of the system then checks the motion for potential collisions between the robot, end-effector, tool, and surroundings. Only then is the path automatically translated into an executable robot program. The individual software modules are coordinated by a central orchestration system and made accessible to the user via a common user interface. The software is already in use as a beta version on the laboratory test bench and will eventually enable component setup and program creation in a matter of minutes.