Roll 'n' Roll
Intelligent Roll-to-Roll Processing of Advanced Carbon Fabrics based on Machine Learning and Non-Linear Automatic Control
In the world of cutting-edge composite materials, AmaDema (Host Organization) has proven itself by introducing NanoWeld®, a revolutionary method of enhancing the characteristics of technical fabrics, such as carbon fiber. In the quest of pushing the production of NanoWeld® to fullfil the needs of the modern composites industry, AmaDema Ltd along with PHOEBE Research & Innovation Ltd started the development of a Roll-to-Roll (R2R) modular machine, befittingly named Roll 'n' Roll. Roll 'n' Roll presents an industrial, innovative solution for scaling up NanoWeld® into mass production and thus, reaching TRL 9, while at the same time offers a novel industrial solution for R2R processing control, which can be further commercialized and implemented in any industrial R2R process line.
A key process in the production of NanoWel d® is the consolidation of the enhancing nano-fabric to the composite material of choice. This consolidation has thus far been performed using Ultrasonic Welding (UW), which for polymers, due to the ultrasonic frequencies, is able to locally increase the mobility of the polymeric material (polymer nano-fabric in the case of NanoWel d®) molecules, resulting in melting points in front of the sonotrode head area. Although UW is a well-proven method for the consolidation of different materials, it presents an issue when scaling up. Multiroll sonotrodes have not been developed yet and therefore, this crucial sub-process is considered a bottleneck in the mass prodcution of NanoWel d®. The solution to this problem is offered by the development of the Roll 'n' Roll module.
At the heart of the Roll 'n' Roll module is a machine-learning-based non-linear automatic feedback control system, developed by PHOEBE Research & Innovation Ltd. The use of machine-learning-based non-linear control has been decided due to the very nature of the problem; the underline industrial process (relation of surface temperature to the electric energy input) has a non-linear behavior, which is not described through mathematical formulas. Therefore, the use of classical control strategies that are build on top of linear processes would not suffice; even in the case of linearization of the process around a stability point, the closed-loop system would not offer the required robustness, since there is no guarantee that the process state will remain in a predefined range.
In the case of NanoWel d® processing, as well as many other products in the market, the material "to be processed" does not always have the same characteristics (polymer material, thickness, etc. may vary) and therefore, key parameters, such as welding temperature, welding force applied and roll speed need to be adjusted. As a result, the PD or PID controls would have to be reset each time to comply with the characteristics of the material "to be processed" The Roll 'n' Roll solution offers the machine continuous training, through which the machine learns how to quickly adapt across the whole of its domain of operation and be able to learn any changes in the process dynamics and adjust the control decisions accordingly during operation. The benefits from such an adaptive system to industrial applications of R2R processing are:
Lower manufacturing costs
Increased production speed and volume
Faster product adjustability to market needs
Production energy savings and lower carbon footprint
Particularly for NanoWel d® the benefits from such an adaptive system are:
Overcoming the bottleneck of UW process scale-up
Increase the production speed from 0.16 m/min to at least 8 m/min