Innovations 2022 - End-to-end Design operation
Explore our latest "end-to-end design operation" innovations
Optimising the design process is an important element for many companies. Through the innovations below, your product and process designers will be able to work out faster, more efficient and cheaper design processes.
Our other 2022 Innovations:
Stiffening of thin-walled structures
The best structural design by optimising its structural topology. That is what we demonstrate with this demonstration.
With structural topology optimisation, designers often think of 3D printed parts. Yet there are other solutions. We investigated ways to make thin-walled designs better and faster via stiffeners.
STIFFENERS IN INJECTION-MOULDED COMPONENTS
To design stiffeners in thin-walled structures, as often found in injection-moulded components, we developed a scale-based topology optimisation that:
- reduces design time,
- improves stiffness compared to manual design,
- reduces the optimisation time required for comparable performance,
- and offers many more producible designs compared to regular topology optimisation.
From data and implicit knowledge to actionable production insights
In Industry 4.0, data is the new oil. Manufacturing companies are therefore capturing huge amounts of data. Yet many companies struggle to turn their data into actionable insights.
One reason is that the data comes in different forms, comes from different systems, is stored in different data silos and requires technical knowledge to retrieve it.
A second challenge is that in addition to the data, domain knowledge is needed to turn this data into actionable insights. This knowledge is often spread across different knowledge sources, from subject matter experts to machine design documentation, standards or laws of physics.
KNOWLEDGE GRAPHS AS A SOLUTION
Knowlegde graphs are an emerging technology that addresses both challenges. For instance, every question you ask on Google is answered by a knowledge graph.
Together with manufacturing companies, we adapted the technology to be applicable in the context of smart and sustainable manufacturing. This opens the door to fast and flexible root cause analysis, prediction and optimisation of production parameters. So instead of asking "Hey Google, who is the king of Belgium?", you can ask "Hey knowledge graph, what causes the defect in my products?".
Make faster controllers using a virtual 3D environment
Working on real machines to develop new operating software is not only impractical. It is also expensive. To reduce both development time and costs, we created a new offline model and operator-in-the-loop approach for developing and validating controllers. All in a virtual 3D environment.
3D AGRICULTURAL SIMULATOR
We built a photo-realistic 3D agricultural simulator, which we use to do model and operator-in-the-loop testing as well as develop new controllers. By simulating the machine and its environment, we receive offline operator feedback. We use this to fine-tune the control software. This saves time as well as costs. Win-win.
Hybrid AI applied on a foil processing machine
With this demo, we show how hybrid AI combines the strengths of data and models to identify unknown parameters faster. As a result, we improve the operation, tuning and performance of your machines. To do this, we use a physics-based backbone that continuously improves the hybrid model.
IDENTIFYING UNKNOWN PARAMETERS
In response to the high demand to identify unknown parameters with less data, we offer a tool to identify those parameters in real-time.
This hybrid estimation technique requires less data than traditional AI and provides accurate results in less time.
Preliminary research shows that a hybrid model requires up to six times less data/time to achieve the same model accuracy as an uninformed neural network.
Reduce controller commissioning time via AI hotstart
Suppose your company has a fleet of production machines and business is going well. So you want to expand your machine fleet with 1 (or more) similar machines. Easier said than done, as commissioning and adjusting machine and controller are time-consuming tasks.
To improve this process, we developed an AI method that uses available historical data from the other (similar) machines in the fleet.
FASTER COMMISSIONING VIA AI
Typically, (experienced) machine operators configure a new machine. Yet in doing so, they regularly encounter problems and often have to spend more time than planned fine-tuning the settings. For machines performing multiple tasks, this fine-tuning is even more difficult because each task requires specific tuning.
Through our AI method, we save an average of 35% in iterations and increase performance during initial testing by 70%. As for systems performing a number of similar tasks, we expect to reduce commissioning time from several days to just one day.
As commissioning is accompanied by improved initial performance, we also expect that you can run the tests with slightly less secure settings, further reducing the commissioning time.
'First time right' profile design through extraction and evaluation of CAD features
'First time right' product design can significantly reduce costs. To help the designers in your design department with their tasks, we have developed a software tool for CAD files. This feature extraction allows designers to:
- Automatically evaluate production constraints;
- Reduce the number of design cycles;
- Access expert manufacturing knowledge.
WHAT DOES OUR SOFTWARE DO?
The software automatically detects the areas to be checked, verifies the constraints and visualises the result. This in just seconds and with an accuracy equivalent to an expert's evaluation.
At Reynaers Aluminium, where they use this software, this has reduced the design cycle per profile by 2 weeks.
Formalising manufacturing knowledge accelerates flawless CAD designs
How can you secure valuable manufacturing knowledge to ensure the long-term success of your business?
We have developed a new methodology for this. It allows you to formally capture and preserve your unique manufacturing knowledge in a generic, process-independent way. This makes your organisation less vulnerable to losing production expertise when your process experts leave your organisation.
In addition, our methodology ensures:
- A faster learning process for junior process experts.
- Fast and independent designs without manufacturing defects by your CAD design engineers
This is possible because we automatically interpret formalised knowledge through the connected CAD evaluation software we have also developed. The total solution thus streamlines the interaction between CAD designers and process engineers.
Maximum performance with minimal energy
Through this technology, we show how you can easily and accessibly optmalise a machine design to:
- minimise energy consumption;
- maximise performance and;
- reduce the total cost of ownership.
We demonstrate our algorithms on a prototype breathing machine that clearly visualises the design parameters. This makes it possible to enter the design parameters into a similar simulation, on the one hand, and to immediately see a visual representation and measurement of energy consumption, on the other.
Result: In a 12-hour simulation, our algorithm reduced RMS torque by 67%.
Digital Twin-based estimators for hybrid vehicles
In this demo, you will see how we support low-cost vehicle sensors with virtual sensors in the extraction of dynamic variables. This results in more reliable safety-critical information such as tire forces and sideslip angle.
METHODOLOGY BASED ON DIGITAL TWINS
To achieve this increased reliability, we have developed a new methodology based on digital twins that:
- we implement in real-time on the embedded hardware;
- significantly reduces the development time of the virtual sensors;
- increases flexibility for different applications.