Even in the highly technical world of industrial manufacturing, where precision and efficiency are top priorities, one crucial problem often remains unsolved: despite optimized processes, parts are not always optimally designed. The financial impact is significant and any efforts to optimize are often time-consuming, costly and difficult for customers to access.
At Optimate, we recognized this problem and actively addressed this challenge. Our goal was to demonstrate that innovative algorithms are capable of identifying the optimization potential of sheet metal parts in real time in a fully automated manner and to optimize the parts immediately.
An idea became a solution
An innovative idea has now become a market-ready cloud solution: our patented optimization algorithm.
But developing a custom solution proved challenging because there were no existing implementations that addressed the development of a geometric optimization algorithm for sheet metal parts. In addition, it was necessary to generate a large dataset of test parts to further explore the patterns of design defects and predict the optimization potential of a sheet metal part based on its characteristics using machine learning algorithms.
Let us now take a closer look at "optimization“
Part optimization plays a crucial role in effectively minimizing cost factors. On the one hand, it enables the early detection of manufacturing problems, such as non-compliance with design guidelines. On the other hand, part optimization can help reduce significant cost drivers by identifying cost-intensive manufacturing processes or inefficient use of materials and replacing them with more cost-effective alternatives. A good example of this is costly process steps integrated into the design, such as welding. Welds are detected and can often be replaced by bends, eliminating the need for the costly welding process.
How did we approach the problem technically?
The optimization pipeline
In the first step, our algorithm analyzes the part in three-dimensional space, recognizing key features such as welds, shape, internal contours and material. Subsequently, the part is abstracted into a graph. Now a Constraint Satisfaction Problem (CSP) is formulated, whose solutions subsequently represent the generated variants of the optimized components. The last step of the optimization algorithm is the translation of the variants into 3D geometries: The part has been successfully optimized and better variants of it have been generated. The use of a geometric modeling kernel enables a high-performance and precise analysis of the CAD designs.
The optimization pipeline was developed in a step-by-step process from the initial proof-of-concept phase to the Minimum Viable Product (MVP) and finally to a production system ready for deployment. Proven software development methods such as Test-Driven Development and Continuous Integration were consistently applied.
The development of a complex optimization pipeline
During development, the challenge was to automatically identify the deficiencies of the designs and the optimization of the parts. This required the formulation of geometric designs as an optimization problem to be solved within an acceptable amount of time. Machine learning was required to provide a solution to this challenge.
A major focus of the development was the use of state-of-the-art software architectures that transformed our system into a cloud application. This cloud application is characterized by its ability to scale at will and operate extremely efficiently. This ensures that our solution remains stable and powerful even as requirements and data volumes grow.
During development, we also focused on innovative development methods. These included Test Driven Development (TDD), where tests are written before the actual implementation to ensure quality and functionality. Pair Programming allowed developers to work together in teams and come up with solutions in real time, which encouraged creative ideas. Continuous Integration also ensured that changes were continuously integrated and tested into the overall system, which accelerated bug detection and resolution.
The development of our algorithm was done internally in close collaboration with the Motius team. This agile collaboration approach allowed us to benefit from the combined experience and expertise of both teams.
With our Automated Optimization algorithm, we are taking a big step towards our vision of revolutionizing the sheet metal part design process. If you wish to deepen your understanding of Artificial Intelligence in sheet metal processing, including its functionality and application areas, click here to explore further in our blog post.
At Optimate, we are not guided by the stauts quo. As pioneers in our field, we curiously test boundaries and dare to be and think differently. We are always looking for the optimum and critically questioning ourselves and the world, always looking for ways to make it better.
Do you feel inspired by the challenges we overcome at Optimate? If you're interested in becoming part of our team and collaborating on innovative solutions, check out our jobs page to find out how you can join us!