Use cases
Defect detection in fibreglass production
Glass fibre products are manufactured using a continuous process that involves several steps, including batch mixing, glass melting and mixing, molten glass transportation and distribution, fibre forming, fibre winding, drying & curing and packaging. The fibrisation process is designed to produce glass filaments with a defined and controlled diameter, operating at very high temperatures (up to 1600°C).
It is a very sensitive process that is subject to breaks, a situation where the process is downgraded, leading to an average rate of defects and waste of up to 10%. After fibrisation, the glass fibres are shaped into bobbins using winding machines. However, the melting, fibrisation and winding process steps are heavily dependent on people’s expertise.
- to detect breaks as early as possible, allowing for fast operator intervention to minimise downtime and destabilisation of the glass fibre production process
- to automate the classification of breaks by type to understand the cause and identify the preliminary indicators of break occurrence
- to predict breaks by correlating their occurrence with upstream parameters monitored at the glass melting level
Currently, operators use a microscope to manually analyse glass beads created at the moment of a glass fibre break, which is detected by observing the bushing. There are no other indicators of a break. The COGNIMAN project aims to develop a system that uses sensors and machine learning to automatically monitor the bushing and detect breaks. With this new system, operators can react more quickly to breaks and minimise manufacturing downtime. The automated system will decrease waste and improve the cost structure of the product, making recycling glass waste more efficient.
Precision machining for deburring of large metal parts
Deburring is a critical process that ensures machined parts are free of unwanted material, called burrs, that
could cause problems during assembly or use. Traditional deburring methods like manual or
machine-based removal are only sometimes effective or efficient. Robotic deburring presents an attractive alternative, but current solutions are limited by their inability to handle large or complex parts and their lack of adaptability.
The goal of this project is to develop a smart, reactive, and safe robotic solution for deburring large metal parts in small batches. The solution will reduce the physical burden on human operators, improve the quality of the finished parts and increase efficiency.
- Develop a robot that can work collaboratively with humans, avoiding collisions and following easy commands
- Equip the robot with cognitive capabilities, including new sensors and machine learning to enable autonomous deburring and quality checking
- Train the robot to learn from its own experience allowing it to perform deburring on parts it has never seen before
- Ensure that the robot can move around the part safely and autonomously
The COGNIMAN project aims to develop an advanced deburring robot that can handle large and complex parts in a safe, efficient, and adaptable way. The robot will use machine learning and multimodal sensors to understand the deburring task and its environment and will be connected to a digital twin for AI training and optimisation. The robot will have mapping, navigation and guidance capabilities which will be tested on parts from wind-power, naval and machine-tool sectors at GOIMEK in Spain. The outcome will be a cognitive deburring robot that can interact with humans safely and confidently.
Additive manufacturing for medical implants
Post-processing of components produced through Additive Manufacturing (AM), particularly in metal, can be a significant burden in terms of cost and time. This pilot will develop solutions to decrease this burden, reducing costs, lead time and development time for a new device. This will enable more complex components to be manufactured faster. Moreover, it will improve worker satisfaction and help AM have a larger impact on the medical device industry. Areas of particular concern are the support material (sacrificial material required during the printing process but unnecessary for the final application), selective surface finish and component distortion.
For medical products, the printed surface can be advantageous for osteointegration and bone growth into the implant, however, it can also be too rough where the implant is in contact with soft tissue. The requirement for different surface finishes on a single component and the removal of support materials whilst maintaining the component’s quality is what makes the post-processing complex. Hence this pilot is focused on its automation. This is especially interesting for patient-specific implants where a patient has unique requirements or trauma that needs immediate solutions.
- Development of a flexible, collaborative robotic system that is easily reconfigurable for post-processing of metal AM components
- Use of AI-enhanced sensors and automation for component inspection and measurement
- Application of Digital Twinning for improved process understanding and advancing design for additive manufacture
- Increased working satisfaction and upskilling for advancing AM technologies
AI-enhanced robots can automate the finishing process for customised medical implants and components, saving time and money while maintaining high-quality finishes. Currently, there are some automated and semi-automated technologies available for surface finishing, but they lack self-adaptive robotics for automated support removal. This pilot will focus on additive manufacturing and precision machining of medical components/implants and will print various example components with developed support structures. The goal is to develop algorithms that can automatically guide robots for support structure removal and subsequent finishing and polishing of surfaces. The resulting robot will be safe, highly flexible, reconfigurable and self-adaptive, reducing the risk of powder exposure to the operator while allowing quick reconfiguration for new geometries and feedback from the operator. Using collaborative/cognitive robots, the demonstrator will show significant improvements over the existing support removal and finishing processes used, while demonstrating a meaningful and seamless social collaboration between humans and robots.
Flexible manufacturing – Digital library for batches
Acciaierie Bertoli Safau S.p.A. (ABS’s) internal logistics deals with finished and semi-finished products in batches. Batches are the smallest handling units. The products come in different quality grades and large sizes due to customisation.
The challenge is to operate and maintain a large library of product batches in a crowded storage place. To address this challenge, the goal is to establish a digital library with real-time access that will allow full visibility of the internal logistic process for finished/semi-finished products and automate the process.
- Introduce implicit identification of batches with drones
- Replace manual barcode scans of labels and tags and mobile device operations
- Know the location of the product, its size and the remaining space spot in the plant
- Optimise space, transport and processing time
- Manage the height in stacks and the complexity of unpackaged products
- Evaluate the use of vehicles as sensors to detect critical issues or safety hazards
- Automate material inventory
The presence of metal in products makes using some technologies, like Radio Frequency Identification (RFID) expensive and complicated. The system must also manage the complexity of non-packaged products and flows, handle materials inside and outside of buildings and use different vehicles. There is no single technology that works for all contexts, so a combination of technologies must be used. Mission creation and routing optimisation are also challenging and must be managed by a software program that considers different constraints. The COGNIMAN application will interface with other data servers and systems to optimise the flow of products. This pilot will focus on automating storage management and digitalising storage information for quick product identification. The outcome will be a digital library accessible in real-time to optimise and manage diverse customer products and reduce processing time. The system will be demonstrated at ABS premises.