Wet spinning of silkworm silk has the potential to overcome the limitations of the natural spinning process, producing fibers with exceptional mechanical properties. However, the complexity of the extraction and spinning processes have meant that this potential has so far not been realized. The choice of silk processing parameters, including fiber degumming, dissolving, and concentration, are critical in producing a sufficiently viscous dope, while avoiding silk's natural tendency to gel via self-assembly. This study utilized recently developed rapid Bayesian optimization to explore the impact of these variables on dope viscosity. By following the dope preparation conditions recommended by the algorithm, a 13% (w/v) silk dope was produced with a viscosity of 0.46 Pa·s, approximately five times higher than the dope obtained using traditional experimental design. The tensile strength, modulus, and toughness of fibers spun from this dope also improved by a factor of 2.20×, 2.16×, and 2.75×, respectively. These results represent the outcome of just five sets of experimental trials focusing on just dope preparation. Given the number of parameters in the spinning and post spinning processes, the use of Bayesian optimization represents an exciting opportunity to explore the multivariate wet spinning process to unlock the potential to produce wet spun fibers with truly exceptional mechanical properties.
Keywords: adaptive experimental optimization; mechanical properties; regenerated silk; wet spinning.