Towards spatial property control of the glass transition temperature and microstructure of 3d printed shape memory polymers via set-point tuning of the extrusion temperature and extraction of the underlying physical model with gaussian process regression

This research focuses on the spatial property control of Shape Memory Polymers (SMPs), which can enhance their performance and functionality, thereby expanding their applications in various fields. Existing methods for controlling generic material and shape memory specific properties do not offer a deterministic approach for selective tuning over the part's volume during its synthesis. This work bridged that gap, with an experimental study that explored how the process parameter of extrusion temperature can be adjusted to spatially control the polyurethane's (PU) glass transition temperature (Tg) and underlying microstructure. The results indicated a highly non-linear characteristic, yet proving that the SMP's Tg could be tuned by 17.8 % around its nominal value of 45 °C, ranging from 42 to 50 °C, while having a fixed composition filament. The coupled process-structure-property effect was modeled with a Gaussian Process Regression to quantify the experimental uncertainty and extract the shape function. The root cause of the resulted different shape memory properties was investigated with Atomic Force Microscopy (AFM), and was found that with an increased extrusion temperature the distribution of the microstructure domains is evolving, with the hard domains of the polymer chains coalescing with the soft ones. This confirmed that the spatial material control goes beyond the one property of Tg, and further affects the material at the micro-level and its crystallinity amount. The latter was confirmed from analyzing the Differential Scanning Calorimetry (DSC) at a temperature window of 105–155 °C. It revealed that the SMP samples printed below 230 °C extrusion temperature did not achieve full crystallinity while their microstructure could sustain further development. The findings of this work demonstrated that shape memory properties can be tuned in-process, allowing the assignment of different actuating temperatures at a volume element size of 0.4 ∗ 0.2 ∗ 10 mm. This enables the design and manufacturing of active structures that perform complex multi-sequence actuations. A final contribution of this work was the development of the conditional framework that uses experimental data and statistical analysis to extract the underlying physical model. This methodology can be applied to sparse data material discovery scenarios as it determines the experimental termination point.