Find out more about current projects:
Machine learning for PCD cutting edges
When special surface quality is needed, tools with diamond cutting edges are usually used, as they also meet the highest demands in terms of cutting parameters and longevity. Polycrystalline diamond (PCD) cutting edges are extremely wear-resistant. Nevertheless, they have to be replaced or reground again and again in order to maintain the full performance of the respective tool. Since such components are subject to 100% inspections at MAPAL, it is naturally a very time-consuming work step. Each cutting edge must be examined individually for possible production defects in the microscopic range. At MAPAL, this was done exclusively by hand and with a trained eye via the eyepiece – until recently. "One of many traditional processes in which deep learning technologies can provide decisive support," Joshua Roth proudly reports on his co-development.
"One of many traditional processes in which deep learning technologies can provide decisive support."
Targeted feedback for users
"The employee no longer has to put every single part under a microscope. Now, he has a tray that can be loaded with a certain number of cutting edges. The cutting edges are then automatically recorded and uploaded to the Microsoft Azure cloud. This is where an object detection model is that is trained to recognise microscopic material defects and to provide the user with targeted feedback," says Joshua Roth, explaining how the implemented pipeline works. The imaging process is intended to support breakout detection and creates considerable added value for the production process by pre-sorting. Even a computer doesn't always find everything. That's why you need a person to make the final decision at the end of the quality control.
Be an important driver of digitisation and discover the exciting opportunities awaiting you in our IT department.
