Artificial intelligence solves decades of medical problems – it controls self-driving cars – it defeats the biggest chess masters – you can read such and similar news in the press every day. But it’s worth knowing that it can also provide a lot of help in performing everyday tasks more efficiently, whether it’s administrative tasks or document management, so engineers can focus on design and creating new things.
It was on an „innovation day” when the engineering department of Knorr-Bremse Rail Systems Budapest first got to know practical utilisations of AI with the help of Neuron Solutions.
Several project ideas on the application of this technology made the engineers inspire to apply AI in the day-to-day work of mechanical design activities. As a next step, Neuron Solutions began to co-operate with the engineering team and identify use cases for simplifying and speeding up the work of design engineers or for resolving problems that are not or just with great difficulty can be addressed. In the course of this joint thinking a number of excellent AI use case ideas were created.
Finally, an idea was implemented that allows you to select tens of thousands of technical drawings for parts that have accumulated over the decades quickly and easily. In the case of Knorr-Bremse, the challenge is to find a large number of drawings of different ages and formats, of varying quality, mostly not yet digitally designed (even hand-drawn, ink-drawn) based on a particular feature, namely the surface treatment mark.
The practical benefit of automated processing is that in the event of a change in standard (statutory) regulations, the relevant drawings (and thus parts) can be found in a resource-saving manner, which is a huge help for documents made decades earlier.
The challenge was to generate structured (e.g., alphanumeric) data from unstructured (e.g., visual) data. AI is well suited for this task, as one of its most common applications is the processing of visual information (images, videos), which was previously not possible or only to a very limited extent with traditional data analysis tools. Specifically for image processing, so-called for object detection, AI “suppliers” of AI platforms and frameworks have already created many tools, i.e. a specially constructed and “trained” neural network, so if someone wants to build such a system, the work does not have to start from scratch.
The engineering project team of Knorr-Bremse Budapest and Neuron Solutions team achieved that AI was able to recognize the surface treatment signal by extending and modifying the available methods. Thanks to the effective collaboration, an experimental technical drawing selection solution was created in just three months. The word “experimental” here rather suggests that this was the first AI application project at the engineering site at this site, but of course the system itself was not only suitable for experimentation. It can be used with great efficiency in everyday work, considering that it can recognize the drawings in question and the technical drawing marks you are looking for with more than 95% accuracy. In addition, the scope of this application can be further expanded by teaching the system to recognize other drawing symbols and information on drawings, but it can also be used to check the quality of newly created drawings by expanding its functions by recognizing any missing drawing symbols or other information.
The implementation and effectiveness of the project demonstrated very well that the application of AI is no longer just the prerogative of high-budget research institutes: in a short time, using reasonable resources, engineering tools can be created to speed up and facilitate engineering work in almost any company.
The Neuron Solutions team is happy to work with the engineering team of Knorr-Bremse Budapest to create a solution that brings practical benefits to everyday work.
We’ve helped Knorr-Bremse to make artificial intelligence work in their Budapest Research & Development Center
