Program 5 - AI-enabled design & automation

RP5-1
ai-based personalized product design and fabrication

 

The Project “AI-based Personalized Design and Fabrication” focuses on the research of 3D human modelling, parametric design, prototyping and non-traditional materials fabrication. The efforts are spearheaded by local and international experts with solid track records in academic research, technology transfer, and commercialization, specializing in the areas of smart manufacturing, solid modelling, advanced materials, AI design and automation. 

​

The Project enables its team of engineers to explore the depth of their research ideas through setting up of advance mechanism and equipment. These include, notably, a Full-Scale 3D Human Body Scanner in the laboratory that allows full body digitization of human subject with 0.2mm precision to support the formulation and analysis of a wide range of 3D human modelling schemes. The digital human can then be used as an avatar for virtual reality applications. Furthermore, going from digital human to personalized wearable fabrication, algorithms are developed to transform the target geometric shape into special representation map. Upon different encoding, such map can be converted into knitting instructions map or weaving map. Our knitting instructions map is compatible with the products from one of the biggest knitting machine manufacturers under collaborative development.

​

Our developments are multi-directional that innovates both software and hardware. We are independently developing an in-house 3D weaving machine that supports non-traditional materials weaving. The prototype has been developed and is under testing and fine tuning for improved performance. We are exploring the potential applications from the advance mechanical property provided by the 3D weaving fabrication on non-traditional materials.

RP5-2
Machine-learning
Enabled Automation Processes and Prototyping

Dr. Shih-Chi Chen’s team is focusing on  implementing advanced artificial intelligence (AI) and machine-learning-based methods to the field of precision machine control and advanced manufacturing. Specifically, we have custom-developed a precision multi-layer roll-to-roll (R2R) printing system and two micro-additive manufacturing platforms, i.e., femtosecond projection two-photon lithography system and digital holography-based 3D printing system, to implement different AI methods. We expect the proposed method may significantly improve the system performance in terms of fabrication rate and resolution by automated system training and parameter optimization, and lower the cost by retiring expensive metrology components such as capacitance probes in precision machines. For example, in our multi-focus 3D printing system, hundreds of laser foci are parallelly and independently controlled to fabricate a design object with optimal laser exposure doses and fabrication trajectories, which cannot be achieved previously without substantial empirical studies and trial and errors.  On the R2R platform, low-cost strain sensor arrays are installed on the system to realize deep-learning based multi-axis precision control, achieving 100 nm positioning precision without the need of parameter tuning.