Artificial Intelligence
Efficiency Booster for the Digital Development Process
Our offering in the field of artificial intelligence makes the difference and is unique on the market. We developed, trained and specialised our AI ourselves long before ChatGPT and the like. As a medium-sized technology company, we knew over ten years ago that big data and the associated data volumes could no longer be analysed manually by users or experts. In the meantime, we have proven many times over that we can offer real added value with our intelligent technology for many use cases in the DMU and 3D environment.
Our range currently comprises five products and is aimed at all companies that want to use cutting-edge technology to make their processes more efficient, their 3D data more transparent and their digital product development faster.
In the vast majority of cases, the investment and effort required to introduce an AI solution is very high. Huge amounts of data have to be available without knowing what the AI will do with it and whether there will be any usable results at all. With our generic AI, we make it easier for you to get started. You can use the AI immediately via our Quality Monitor and receive your first results straight away. The technology is already being used productively by numerous customers.
When we integrate our AI products and the artificial intelligence learns along with them, it is very important to us that you have full control at all times. This is why the process role of the AI can be expanded step by step and you decide which tasks the AI takes on in the overall process of virtual geometric validation. The possibilities are many and varied: our solution can, for example, provide support in the day-to-day business of assessing collisions, distance violations and other conflicts.
Our AI for image generation automatically generates information that users need to analyse and previously had to create themselves. This also includes photorealistic images that can be used to capture and assess conflict areas in just a few seconds. We also work on a use-case basis for component recognition and service.
AI Development at invenio Virtual Technologies
Our AIs at a Glance:
Generic AI for Virtual Geometric Validation
The AI is integrated into our standard product Quality Monitor (QM) and is already being used successfully in production. Many of our customers use QM to calculate several million component pairings for geometric consistency every day. The generic AI automatically prepares potential problem areas so that all relevant information is available at a glance and users can assess the conflicts quickly and efficiently. This significantly reduces manual effort. The AI can be used in a variety of ways, from preparing the data and enriching it with additional information to independently assessing the conflicts.
Advantages:
- already trained by us
- independent of customer data
- Can be used across all sectors
- Can be used immediately thanks to plug and play
- Has been in productive use since 2020
- ideally suited for getting started
Co-learning AI for Virtual Geometric Validation
The co-learning AI also supports you in virtual geometric validation to assess conflict points in digital prototypes. Just like the generic AI, the solution is integrated into Quality Monitor. It builds on the generic AI and adapts to customer-specific data and special features. The co-learning AI looks over the user's shoulder during the assessment and thus learns continuously.
Our solution can also suggest evaluations of problem areas and thus support you in the efficient assessment of geometric conflicts. The co-learning AI always works hand in hand with the users, but depending on the level of training and confidence, it can also make decisions independently to provide even more relief.
The self-learning AI continues to learn when users evaluate collisions. We include many different factors in the training, such as component relationship, metadata and confidence level. This allows our solution to recognise different assessments for the same problem. The AI provides suggestions by drawing on what it has learnt so far and comparing them on the basis of similarities.
Advantages:
- trains itself, i.e. the AI automatically looks over the user's shoulder when making judgements
- learns more and more and becomes even more efficient
- adapts to customer-specific data / special features
- Has been in productive use since 2022
- recognises different assessments for the same problem
- if the AI learns incorrectly, the error can be traced and removed retrospectively
AI Image Generation
The AI for image generation is integrated in the VT-DMU module 'photo-inVT' and in the material editor. This means it can be easily integrated into customer processes and systems. The AI technology supports the rendering of images; the graphics are generated directly from the 3D data. AI is used by our customers to generate images from large amounts of data or to render 3D scenes in real time. The results can be used for marketing materials, photorealistic product presentations or the exact visualisation of product details and much more.
The VT solution guarantees perfect image quality with minimal calculation time and low hardware requirements. During the rendering process, the AI uses the trained mesh, optimises noisy image areas and finalises the image.
Advantages:
- AI can be easily integrated into customer systems
- photorealistic images can be generated directly from the 3D/CAD data
- Automatic generation in batch
- even large amounts of data are calculated with high performance
- shorter calculation time (without AI 900 sec., with AI 45 sec.)
- Low hardware requirements
- No external costs for image creation
- possible applications: mass image generation and real-time rendering of 3D scenes, precise visualisation of product details, marketing documents, photorealistic product presentation, documentation
AI Component Recognition
In addition to generic and co-learning AI, we are constantly looking for further use cases to successfully utilise artificial intelligence in virtual geometric validation. As part of a research project, we have successfully developed our AI for automatic component recognition.
The input is once again 3D data, which is recognised and classified by the trained neural network. The network can be adapted to customer-specific data. For large components, the AI uses other component properties so that components of any size can be recognised. The recognised component types are output with probabilities.
The AI provides valuable information that would not be available without it. These details can in turn be used in component handling systems, e.g. for targeted collision handling and calculation in virtual geometric validation, for production planning or for purchasing processes. The function is already integrated and available in Quality Monitor via 'Similar Relations'.
Advantages:
- integrated in the modular software toolkit VT-DMU
- is first fed with 3D geometries in order to learn
- then correctly identifies over 90% of small parts
- Indicates component type and probability
- Small parts could be included or excluded separately
- for larger components, in addition to the neural network, also uses other component properties
AI for Service
In digital service, service processes for a new product are simulated during development. Even before production begins, it must be clear how the result can be serviced or repaired. This means that all workshop information such as repair instructions, spare parts concepts or repair times must already be available at this stage.
More and more product variants are being developed under high time pressure. Physical prototypes, on which repair concepts could be developed, have been in decline for years. Due to these conditions, workshop information is increasingly being created completely digitally.
The mix of huge amounts of data, daily changes, high time pressure and many expert opinions is an ideal starting point for using artificial intelligence efficiently. Our "agents" automatically register every change in the development data. As soon as a geometric deviation is identified, it is presented to the AI prototype for evaluation. The AI immediately recognises what impact this deviation has on the repair concept and makes a recommendation as to whether the service process needs to be checked again. This process noticeably reduces the burden on users in their daily work, as they no longer have to look at every change in the data manually before making a decision.
Advantages:
- Workshop information is created digitally and in parallel to development
- 'Agents' of the VT solution automatically register every geometric change in the development data
- As soon as a deviation is identified, it is presented to the AI for evaluation
- The AI immediately recognizes the impact of the change on the repair concept and makes recommendations
- The previous values are either retained or the repair process is reassessed
- The technology takes the pressure off the experts, who still always have the decision-making authority
- Initial tests have shown: The entire process is up to 60% more efficient
