Manufacturing (WP3)
WP3 focuses on improving the accuracy, performance and reliability of production machinery and processes. It achieves this by developing adaptive AI models trained on a unique combination of real experimental data and synthetic data from physical simulations (FEM).
A key part of WP3 involves developing models to predict the behaviour and characteristics of manufacturing machines based on sensor data, and utilising these for feedback control of manufacturing processes. This also includes the creation of datasets and simulations of critical operating conditions, which will help to reduce the costs of experimental testing and accelerate development. WP3 will result in adaptive manufacturing systems capable of dynamically responding to changes in operating conditions and increasing both overall productivity and production quality.
What problem does the WP solve, and why is it important?
01
A lack of, or low variability in, real-world training data from operational sources
02
The considerable time and financial costs involved in large-scale physical experiments and the manual annotation of data
03
The limited accuracy of existing physical models when real-world conditions deviate
Examples of use / areas of application and benefits
Machine Tools and Mechanical Engineering
- improved machining accuracy and process stability
- reduction in scrap rates and operational deviations
- faster and more reliable prediction of spatially and temporally dependent machine characteristics
Optomechanical systems and precision engineering
- optimisation of the assembly of high-tech systems
- greater accuracy and long-term stability of the equipment
- elimination of faults in monitoring equipment caused by environmental factors
Additive manufacturing (3D printing)
- proposal of default technical and print parameters using an AI assistant
- detection of areas with sub-optimal temperature on the track, with a proposed correction
- improved stability, repeatability and geometric accuracy of the manufacturing process
Management of iron and steelworks
- the deployment of robust machine vision in the demanding environment of a steelworks
- automatic capture of CCTV images with anomalies or uncertain interpretation
- real-time optimisation of process parameters and ongoing model training by the operator
Key people
Participating institutions
Použité technologie a postupy
- Advanced physical modelling of machine and structural properties using the finite element method (FEM)
- Synthetic data and data augmentation to expand training datasets and cover a wide range of operating conditions
- Industrial machine vision, advanced metrology systems and in-process sensor technology
- Compensatory and adaptive algorithms for real-time process control
- Training AI models on hybrid datasets (a combination of experimental measurements and synthetic data)
- Transfer learning for the efficient transfer of trained models between different machines, processes and application domains
- Advanced measurement methods for characterising machine properties and validating models
Used technologies and procedures
2027
Automatic annotation software based on the human-in-the-loop principle, designed for computer vision
R- software
A digital model of the accuracy of an opto-mechanical assembly based on an AI model
Gfunk – funkční vzorek
2028
AI assistant for designing the process parameters for PEM extrusion
R- software
An AI model of a machining centre with enhanced thermal-mechanical accuracy across the working range
Gfunk – funkční vzorek
2029
A high-precision digital model of a device for quality control of opto-mechanical assemblies based on the application of an AI model
Gfunk – funkční vzorek
2031
A tool for managing data pipelines, training models and deploying them into production
R- software
AI assistant for detecting problem areas in 3D printing
R- software
An AI model for a machining centre to improve production efficiency within the working area
Gfunk – funkční vzorek