Aarhus University Seal

MADE FAST

MADE FAST is the third major project coordinated by the MADE consortium in relation with manufacturing in Denmark. FAST is an acronym for Flexibility, Agility, Sustainability and Talent and the full budget is more than a quarter of a billion DKK, where 80MDKK comes from the Danish Innovation Foundation. It is a collaboration with more than 50 companies, 5 universities and 3 GTSs.

MADE FAST is divided into five streams of work where the Department of Engineering at Aarhus University (ENG/AU) participate in Work Stream 3 and 4 (WS3, WS4). Each of the WSs are divided into a series of part projects each led by at least one company providing the case for the research to be undertaken. Each part project as such are independent, but naturally there is also collaboration across the different part projects in order to gain findings at a higher level. WS3 deals with “Agile Production Systems” whereas WP4 deals with “Digitalization of Manufacturing Processes”. The common theme for all the part project that ENG/AU is involved in are that they involve research related to digital twins of either the production systems themselves or the manufacturing processes as a whole. Inside the digital twins the target is to obtain predictive capabilities using a combination of machine learning and co-simulation combining different kinds of models. The foci of the different part projects depend upon the needs from the case study in question, and each of them will have one dedicated PhD student.


The different part projects that ENG@AU are leading are:

3.01

 Digital twin with Co-simulation for Packing and Assembly Lines in Manufacturing

A digital twin will be comprised of modular simulation models of the manufacturing equipment, produced by different modelling and simulation tools. The behavior of the digital twin will be computed in real-time by combining the behavior of each simulation model, using a co-simulation engine. The behavior of the digital twin will be used to predict if the physical manufacturing (the physical twin) performs as predicted by the co-simulation and product outputs for decision support in the form of descriptive and prescriptive signals to the operator and/or production manager. In a real-world situation, discrepancies between the physical and digital twin will occur and often develop or change over time due to changes in the physical environment. The cause of such discrepancies can either be an indication that the physical manufacturing is not performing optimally, and actions can be taken to avoid reduction in productivity of the production line, or a need for re-calibration between the digital and physical twin. In order to give the best possible advice to the operator, when discrepancies exist, the digital twin will need to be able to identify the cause of the discrepancy (e.g. using machine learning) and, when needed, transfer these results into suggested actions.

Collaborating organizations: LEGO, Danfoss Drives, Technicon, TI and SDU

4.07

Improving Filter Insert Performance and Quality using Simulation and Data Analytics

This part project aims at detecting defects and controlling the variability of oil filter inserts, and therefore improving filter performance and reducing manufacturing costs. This will be done by modelling the flow of oil through the filter insert (Filter Modelling) as well as relevant parts of the manufacturing process (Production Modelling) and correlating the simulation results with data collected from physical evaluations of a selected batch of filter inserts.

Collaborating organizations: CC Jensen, Universal Robots, Technological Institute, FORCE and DTU

4.08

Digital Twin of Movable Factory

In the last 15 years, the wind turbine has only gone one way, namely, bigger and bigger. Rotor diameter has grown with approximately 3.8m/year, and as the rotor gets bigger, the rest of the turbine’s components grow similarly. This means that the value chain is becoming increasingly challenged, and transport of large items is becoming a very big issue. Different places in the world there are restrictions on what times of the day where it is allowed to transport the components, there is restrictions on which roads we must use and what kind of transport, whether it is by truck, train etc. The cost of such transports becomes very expensive, and in a market where price per MW, is decreasing, we get increasingly challenged to find a way to reduce the cost in our value chain. One of the options to cope with these challenges is to move the manufacturing of the modules closer to installation site, and in a controlled environment to assemble such modules locally. The vision of this part project is to build a digital twin that can demonstrate establishing a movable factory solution that can be moved around the world and configures so the assembly processes for a specific case study.  

Collaborating organizations: Vestas, FORCE and Alexandra Institute

4.09

Enabling Real-Time Release Testing using Digital Twin in Medical Device Assembly

Since 2004 the US Food and Drug Administration (FDA) has encouraged pharmaceutical industry to implement a modern, science- and risk-based quality assessment system. In 2019 FDA released a draft guideline on Continuous Manufacturing to support the use of modern manufacturing technology and to simplify the manufacturing processes by, for example: using an integrated process with fewer process steps and shorter processing times; supporting an enhanced development approach (e.g., quality by design and use of process analytical technology); enabling real-time product release; and providing flexible operation. The vision of this part project is to develop, test and implement Continuous Manufacturing with special focus on real-time release testing to assembled medical devices using digital twins. It is expected that adopting real-time release testing in assembly of medical devices will increase device quality, lower manufacturing costs, decrease the lead time from assembly to release of device and in the end improve availability of quality devices to the patients. The digital twin is expected further to secure availability of the necessary regulatory documentation of quality for realizing real-time release testing.

Collaborating organizations: Novo Nordisk, FORCE, TI, Alexandra and DTU/MEK

4.10

Modular Digital Twins enhancing integration speed

Modular and/or collaborative automation applications has a significant competitive advantage as it enables key characteristics of future manufacturing demands. This includes simple tasks as machine tending and pick n’ place, but also process specific applications as path-specific applications as gluing, sanding and deburring, As well as advanced pose-specific applications as screwdriving, force-based insertion etc. Utilizing digital twins to decrease the integration time of system integrators and increase the accessibility and availability of such automation application to the untrained end-user operator, is a key success criterion of this project. This will be achieved by addressing digital models of components, to create application configurators and application specific co-simulation in a web-based environment to create high quality application integrations and predict performance of configurations as well as mitigate risk of variance.

Collaborating organizations: Technicon and SDU

4.18

Online process control and optimization using X-ray and AI

Manufacturing of mineral wool is a non-linear process in the sense that with the current knowledge and technologies it is not possible to control the output quality 100%. To improve efficiency and quality, an in-line quality system which continuously can scan the inside of the wool lane for quality issues related to density, homogeneity and foreign objects is of high importance. To enable automatic analysis of inline data, a deep learning-based solution for X-ray data classification/semantic segmentation will be developed along with a processing pipeline that can employ annotated inline data for training a deep learning model. This model will (after training) be able to classify new inline data to one of the classes indicated by the labels (human-provided annotations) included in the data used during training.

Collaborating organizations: Rockwool and FORCE