ASSOCIATE PROFESSOR & GRADUATE PROGRAM CHAIR Rowan University, United States
Introduction: Polymer nanofiber yarns fabricated from biocompatible and biodegradable polymers can be employed in biomedical applications such as tissue engineering, biosensors, implants, sutures and many more. Nanofibers have a high surface area to volume ratio which results in excellent mechanics and great intractability in biological media. A new nanoyarn manufacturing platform was developed in our lab facilitates great control over manufacturing parameters and lead to improvements in mechanical tenacity compared to nanoyarns fabricated via self-assembly electrospinning and cone electrospinning. Nanoyarns made using the described method can be fabricated from any electrospunable polymer. This also means that nanoyarn production rate is directly tied to the production rate of electrospun nanofibers. One of the main obstacles in scaling up the production of electrospun nanofibers is the relatively low yields. Thus, scaling up nanoyarn production requires improving electrospinning efficiency. Electrospinning efficiency depends on many factors including pump rate, collector material, needle height, track material. We hypothesize that said parameters affect the electric field of spinning system, thus indirectly tying the electric field system efficiency. Finding optimal values for electrospinning parameters is paramount in maximizing electrospinning efficiency. This will be done using a machine learning approach . Efficiency is measured at various spinning parameters. Said efficiencies along its corresponding parameters are used to generate a training set for a machine learning model. Said model is then used to predict optimal spinning parameters for maximizing efficiency.
Materials and
Methods: Electrospinning and fiber collection and yarn fabrication is done using a parallel track collection system Figure 1. Nanofibers are spun and collected on a collection belt. The collected fibers form an interconnected mesh of aligned and discrete nanofiber segments Figure 1C. Said fibers are transferred to a yarn spinning device (Figure 1E) and spun into yarn. An electrospinning system consists of tunable parameters. A 3D electrical field model is generated using COMSOL to match the parameters of each spinning process. Spinning efficiency is assessed by measuring the fiber weight post collection to the starting polymer weight. The electric field and efficiency alongside their parameters are used to generate a dataset to train a convolutional neural network (CNN) classifier. Data processing is done using the Pandas library and the PyTorch library is used to construct, optimize the assess the CNN. Model results will then be used to modify electrospinning system parameters and obtain more efficiency values. Those values will be used as new data point for the CNN.
Results, Conclusions, and Discussions: Discussion Current electrospun yarn fabrication techniques allow for little parameter tunability because fiber spinning and yarn spinning occur simultaneously. A new manufacturing platform was developed separates the fiber spinning and yarn spinning process. This approach allows for better control over manufacturing parameters. We call those fibers multi-fiber twisted yarns (MFTY). Polycaprolactone (PCL) MFTYs were manufactured using the system described in Figure 1. PCL MFTY tenacities outperformed nanoyarns made via electrospinning self-assembly and cone electrospinning (Table 1) [1].
The main obstacle to scaling up nanoyarn production is the low yields associated with electrospinning. To solve this problem, we will investigate a machine learning approach to optimize the various parameters that contribute to an electrospinning system. The main driving force behind electrospinning is the electric field. The electric field is affected by all the parameters that make up the electrospinning system such as polymer pump rate, track material, needle height, track speed, track width, etc. A parameter sweep is used to model the electric field using COMSOL. The same sweeps is used to construct electrospinning systems and measure their yield. The yield, electric field and their associated parameters will be used to construct a training set for a convolutional neural network classifier with the goal of predicting and optimizing the yield of the electrospinning system. References
[1] Göktepe, F. and B.B. Mülayim, Long path towards to success in electrospun nanofiber yarn production since 1930’s: a critical review. Autex Research Journal, 2018. 18(2): p. 87-109.
