Research project KOGNIA: design aided by AI
The University of Augsburg and CADENAS have worked together on an assistance system for engineering which is supported by Artificial Intelligence
The University of Ausgburg and CADENAS are part oft he research project KOGNIA in order to develop an assistance system for Computer Aided Design (CAD), supported by Artificial Intelligence (AI). The researchers took advantage of the resemblence of CAD to the human language.
The research project KOGNIA was founded to support designing processes by AI and machine learning. Sponsored by the Bavarian State Ministry of Economic Affairs, Regional Development and Energy, researchers of the Institute for Software & Systems Engineering (ISSE) at the University of Augsburg and employees of CADENAS have worked on an AI-supported assistance system for CAD.
Theory and practice
The project’s aim was to reuse knowledge and solutions from earlier constructions for new use cases. “When a company designs a product, for instance a smartphone, for the first time, the construction plan is gathered“, explains Prof. Dr. Wolfgang Reif, Director of ISSE and head of KOGNIA. In this manner huge bulks of construction plans and data emerge. In theory, designers who are developing a successor can draw on this.
But in practice, especially young and unexperienced engineers are facing multiple problems: On the one hand they have to learn the company’s standard construction plans and on the other hand they have to search for the right part in countless product calatogs. “This takes 60 to 70 percent of the time working on a CAD construction“, knows Carola Lenzen, project staff and research fellow at ISSE.
An AI which searches for components
KOGNIA researched on simplifying the search for parts with the help of AI. CADENAS is a predestined support for the project, because the company has more than 30 years of experience with the software PARTsolutions. The solution for Strategic Parts Management helps to make the search for parts easier and therefore to speed up design processes. On this basis ISSE and CADENAS have developed an AI which gathers every part in its place within the overall structure and its interaction with other components.
Construction plans resemble compound sentences
“We perceived, that CAD and the human language are very similar. Like the word order in a sentence represents a linguistic structure, you can understand a product’s CAD construction as an order of parts“, outlines Reif. So the researchers faced a new challenge: existing artificial neural networks, the algorithms behind the AI, are used to process pictures or language. “We had to find a way to translate parts into words, to understand the complexity“, illustrates Lenzen.
„The AI analyses every existing design of a company to learn the components' connectivity. That way it acquires company-specific knowledge. Because of that KOGNIA doesn’t give a generic hint but provides a definite suggestion, that fits closely the internal use case. So it takes experienced employees less time to find the right parts and new designers learn about the company’s engineering conditions“, says Frank Epple, CTO and KOGNIA project manager at CADENAS.