AI TECHNOLOGY FROM ANTARES VISION GROUP AUTOMATES AND INCREASES QUALITY CONTROL AND COULD LOWER PRODUCTION COSTS
The software developed by Oròbix, the Tech Company of the Group, is based on AI and exploits image segmentation techniques that identify anomalies on the component and also graphically show where there are located.
Artificial intelligence has many applications across diverse industries. One exciting example is in Italy, where AI technology from Antares Vision Group member Oròbix is helping improve inspections at a manufacturing facility that’s part of the Germany-based ZF Group.
Aim of the ZF Group AI project:
ZF Group is a global technology company that supplies passenger cars, commercial vehicles, and industrial technology systems. The ZF Gardone Val Trompia plant produces steering systems, hydraulic pumps for automotive applications, airbags, shock absorbers for motorbike applications and components for the maritime sector.
To improve quality control to meet stringent safety requirements for the manufacture of sprockets used in steering systems, ZF Gardone Val Trompia sought to automate the inspection process, traditionally performed with magnetic particle inspection techniques.
Sprockets must be free from defects, such as pre-existing cracks in the steel rod used in production or cracks that form during heat treatment. An automated quality control process is vital, mainly because production includes the high-heat techniques of carburizing (i.e., hardening by adding carbon) and induction hardening, which can create a wider variety of defects.
AI project development:
To develop a solution, ZF Gardone Val Trompia worked with Oròbix, part of Antares Vision Group and one of the foremost Italian companies engaged in developing specialized AI solutions for the manufacturing sector, including software that automatically detects manufacturing defects. The software Oròbix developed is based on AI and exploits image segmentation techniques that identify anomalies on the component and also graphically show where there are located.
Also joining the effort was CGM CIGIEMME S.p.A., which specializes in non-destructive magnetoscopic testing and designed and manufactured the inspection system installed at ZF Gardone Val Trompia.
The project began with the development of a prototype system. Image analysis is conducted directly on the inspection machine. At the same time, the algorithm learning phase is carried out in the cloud, following the annotation (i.e., labelling) activity of example images, performed with the support of a metallurgist.
AI project results: Quality control, cost and environmental impact reduction
The project has yielded excellent results, the most important of which is increased quality control. Compared with visual control conducted by an operator, there was a significant decrease in the detectability value — an accuracy index measured on a scale of 1 to 10, with 10 being “no possibility of identifying the defect” in the Failure Mode that affects the analysis process.
With the samples used in the project evaluation phase, Oròbix’s AI solution had the potential to reach a detectability value of 2. In contrast, the detectability value achieved with an operator’s visual control was almost three times greater.
Other benefits of the system include potential savings on the total cost of production and compatibility with a new induction hardening system ZF Gardone Val Trompia has planned that will allow better management of energy resources and reduce the environmental impact of the production process.