Skip to Content
Client story

Improving Wieland’s QA with AI-driven computer vision

Client: Wieland Group
Region: Global
Industry: Metal Processing

To inspect products even more efficiently, Wieland Group and Capgemini set up AI-driven computer vision, a solution that ensures the highest quality and is managed in the cloud, enabling Wieland to scale it independently as required

Client Challenge: Wieland places not only the highest functional but also visual quality demands on its products. In order to guarantee this quality, manual visual inspection has always been of great importance. Both the fact that it is a round test object and the metallic, reflective copper surface pose challenges.
Solution: Wieland worked with Capgemini to develop, implement, and test a scalable computer vision solution based on a deep learning model to enable the automation of visual quality inspection for many use cases in production.
Benefits:

  • More efficient visual quality inspection with improved performance at the same time
  • Support for more than 20 test cells and 50 products with a scrap rate of 1.5%
  • Centralized management from the cloud
  • Easy scalability to further use cases

A leading supplier of semi-finished products

The Wieland Group is the global market leader for semifinished products and system solutions made of copper and copper alloys. Founded in 1820, the company now employs around 9,500 people at more than 80 locations and generates a turnover of 6.3 billion euros – with the aim of continuing to grow. To drive sustainability, efficiency, productivity and reliability, Wieland is focusing on digital transformation, automation and optimization. This is why the company turned to Capgemini, a global leader in transforming businesses through intelligent industry solutions, to collaborate on innovative digital solutions.

Following a thorough review of Wieland’s existing operations, including on-site visits, the partners set up a series of joint workshops that identified pain points that could be addressed with intelligent solutions. Through these engagements, Wieland and Capgemini identified that the organization had to invest a disproportionately high amount of effort to ensure quality demands were met. Meanwhile, the partners also recognized the myriad possibilities offered by image processing solutions based on AI. These complementary conclusions led Wieland and Capgemini to select AI-driven computer vision as their initial focus for innovation.

Determining the feasibility of an AI solution

With the primary objectives, challenges, and solution established, the project team then launched a feasibility study. This enabled Wieland and Capgemini to determine whether or not, when excluding real-world factors and constraints, they could train a deep learning model to detect purely optical defects on the surface of components.

However, this study clarified an additional challenge: a lack of specimens. This meant that the partners would need a smart idea and a clever extension of the training data set. Fortunately, Wieland and Capgemini utilized data augmentation to artificially expand the training dataset, significantly improving the trained model’s ability to detect the considered defects. Thus, the team could reliably conclude that such a solution could work if applied effectively.

Of course, the model had to be applied under realistic conditions. This meant finding a hardware setup that met real-time requirements, such as a reliable and resilient edge computing approach with fast response times, which was necessary for consistent operation in case the network or infrastructure failed. As the project team setup, the hardware, they took into consideration camera angles, background, and other critical factors that would impact the solution’s effectiveness. In the end, this showed that reliable recognition rates were possible under practical conditions.

Based on the OPC UA standard, a fast, reliable, real-time data exchange could be established between the Programmable Logic Controller of the conventional measuring system and the robot in order to enable material feeding for the edge device. With the hardware and edge computing in place, the partners then connected the setup to the cloud, which served as the basis for future data analysis and efficient management of the AI solution.

Setting up Wieland to scale independently

Following the successful feasibility study and PoC, Wieland and Capgemini took the next step: setting up Wieland to be able to self-reliantly carry out the solution’s large-scale deployment. This was done based on Capgemini’s own CLEA* solution, a pre-existing generic platform allowing the verification of controlled assets and triggering actions based on computer vision. This included the computer vision system and edge device’s hardware setup, as well as the cloud requirements and the necessary configurations and connections.

During this phase, close alignment and open communication between Wieland and Capgemini were the key to the project’s success. Ultimately, trusting cooperation laid the foundation that enables Wieland to continuously improve the solution and independently scale it to further use cases.

 As a result of this engagement, Wieland now has a solution that can be extended to more than 20 test cells and applied to more than 50 products. Computer vision can also inspect one million parts per week with a scrap rate of 1.5%.

Due to the applicability to a wide range of the company’s product portfolio, the solution is a significant building block of automation and digitization and represents a major step towards intelligent industry for Wieland. AI-driven computer vision enables Wieland to produce high-quality products for their customers in a highly efficient manner. The collected data in the cloud provides the potential to run additional evaluations to improve production and processes and enable the company to make data-driven decisions.

Meet our experts

Dr. Ing. Johannes Lechner

Senior Business Analyst
Johannes advises our clients in the field of Digital Manufacturing and collaborates with them to implement innovative new solutions, from the pilot phase right through to industrialization. Johannes focuses on achieving tangible benefits for our clients that go beyond simply using technology, such as increased efficiency and quality.

Stefan Steidle

 IT Transformation Director for Digital Manufacturing
Stefan supports our clients in the automotive and manufacturing sectors in the digital transformation of their development and production landscapes. He focuses on digital consistency, i.e. the integration of data, systems, processes and organizations along the entire value chain from product development to factory planning and production.

    Client stories