Unlocking Smart Manufacturing with Digital Twins and Machine Learning

The use of digital twins is becoming more widespread in the manufacturing sector due to applications such as product development, preventive/prescriptive maintenance, performance optimization, and fault detection. Due to the complexity of the manufacturing process and large datasets involved, a digital twin requires the use of advanced modelling techniques and machine learning to create a dynamic virtual representation. In the realm of smart manufacturing, the fusion of digital twins and machine learning stands as a beacon of progress. Spearheading this frontier is a groundbreaking study by Professor Javaid Butt that is probing the depths of additive manufacturing with a focus on the fused filament fabrication (FFF) process using thermoplastics.

While FFF offers many advantages such as ease of operation and abundance of materials, two critical challenges persist, leading to considerable time and material wastage through trial and error. These challenges revolve around unexpected maintenance and identification of optimal printing parameters. With the accelerated attempts to enhance the mechanical properties of FFF-printed parts, composite filaments are being manufactured with additives and fibers that tend to clog the printing nozzle or result in under/over extrusion to affect the quality of the printed parts. Furthermore, identification of optimal processing parameters to achieve desired properties is also crucial for product development. Therefore, this study focuses on the design and deployment of a digital twin to help manufacturers in solving these critical problems related to the FFF process.

A digital twin framework has been proposed and implemented for the extruder assembly of the FFF system that can be used for different materials and process parameters; thereby, limiting the wastage of time and resources through trial-and-error to ensure sustainability of operation. Parameters can be changed in the virtual digital twin to ensure that proper material flow can be achieved in the printing nozzle without clogs or obstructions; thus, preventing blockages and enhancing the lifespan. To represent the intricacies of its real-world counterpart, the digital twin comprised three key simulations: conjugate convective heat transfer, multiphase material melting, and non-Newtonian micro-channel dynamics. This digital twin mirrors the intricacies of its real-world counterpart, capturing even the most minute details for material flow. Through rigorous experimentation and validation, the digital twin emerges as a true reflection of reality, offering a reliable portal into the world of smart manufacturing.

The utilisation of sophisticated machine learning algorithms further enhanced the capabilities of the digital twin by providing predictions based on simulation and experimental data for product performance. Together, they form a dynamic framework poised to revolutionize the additive manufacturing landscape. The implications of this work are profound. By harnessing the predictive power of the digital twin, manufacturers can pre-emptively tackle two perennial challenges: unexpected maintenance and the identification of optimal printing parameters. With this digital twin at their disposal, manufacturers can now traverse the intricate landscape of additive manufacturing with enhanced clarity and precision.

The digital twin heralds a new era of manufacturing innovation by providing a realm where parameters are fine-tuned in the virtual domain to ensure seamless material flow and pristine product quality. This work ushers in an era of efficiency, sustainability, and unparalleled innovation by replacing trial and error with calculated precision for manufacturing excellence.

For further information, please contact Professor Javaid Butt at [email protected]

You can also visit the research paper link: https://doi.org/10.3390/met13010024