General rightsThis document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/pure/about/ebr-terms
A B S T R A C TFused filament fabrication (FFF) is a 3D printing technique which allows layer-by-layer build-up of a part by the deposition of thermoplastic material through a nozzle. The technique allows for complex shapes to be made with a degree of design freedom unachievable with traditional manufacturing methods. However, the mechanical properties of the thermoplastic materials used are low compared to common engineering materials. In this work, composite 3D printing feedstocks for FFF are investigated, wherein carbon fibres are embedded into a thermoplastic matrix to increase strength and stiffness. First, the key processing parameters for FFF are reviewed, showing how fibres alter the printing dynamics by changing the viscosity and the thermal profile of the printed material. The state-of-the-art in composite 3D printing is presented, showing a distinction between short fibre feedstocks versus continuous fibre feedstocks. An experimental study was performed to benchmark these two methods. It is found that printing of continuous carbon fibres using the MarkOne printer gives significant increases in performance over unreinforced thermoplastics, with mechanical properties in the same order of magnitude of typical unidirectional epoxy matrix composites. The method, however, is limited in design freedom as the brittle continuous carbon fibres cannot be deposited freely through small steering radii and sharp angles. Filaments with embedded short carbon microfibres (∼100 μm) show better print capabilities and are suitable for use with standard printing methods, but only offer a slight increase in mechanical properties over the pure thermoplastic properties. It is hypothesized that increasing the fibre length in short fibre filament is expected to lead to increased mechanical properties, potentially approaching those of continuous fibre composites, whilst keeping the high degree of design freedom of the FFF process.
Natural Language Processing (NLP) has recently achieved great success by using huge pre-trained models with hundreds of millions of parameters. However, these models suffer from heavy model sizes and high latency such that they cannot be deployed to resourcelimited mobile devices. In this paper, we propose MobileBERT for compressing and accelerating the popular BERT model. Like the original BERT, MobileBERT is task-agnostic, that is, it can be generically applied to various downstream NLP tasks via simple fine-tuning. Basically, MobileBERT is a thin version of BERT LARGE , while equipped with bottleneck structures and a carefully designed balance between self-attentions and feed-forward networks. To train MobileBERT, we first train a specially designed teacher model, an invertedbottleneck incorporated BERT LARGE model. Then, we conduct knowledge transfer from this teacher to MobileBERT. Empirical studies show that MobileBERT is 4.3× smaller and 5.5× faster than BERT BASE while achieving competitive results on well-known benchmarks. On the natural language inference tasks of GLUE, MobileBERT achieves a GLUE score of 77.7 (0.6 lower than BERT BASE ), and 62 ms latency on a Pixel 4 phone. On the SQuAD v1.1/v2.0 question answering task, MobileBERT achieves a dev F1 score of 90.0/79.2 (1.5/2.1 higher than BERT BASE ). * This work was done when the first author was an intern at Google Brain.
The aim of this research is to manufacture intermingled hybrid composites using aligned discontinuous fibres to achieve pseudo-ductility. Hybrid composites, made with different types of fibres that provide a balanced suite of modulus, strength and ductility, allow avoiding catastrophic failure that is a key limitation of composites. Two different material combinations of high strength carbon/E-glass and high modulus carbon/E-glass were selected. Several highly aligned and well dispersed short fibre hybrid composites with different carbon/glass ratios were manufactured and tested in tension in order to investigate the carbon ratio effect on the stress-strain curve. Good pseudo-ductile responses were obtained from the high modulus carbon/E-glass composites due to the fragmentation of the carbon fibres. The experimental results were also compared with an analytical solution. The intermingled hybrid composite with 0.25 relative carbon ratio gave the maximum pseudo-ductile strain, 1.1%, with a 110 GPa tensile modulus. Moreover, the initial modulus of the intermingled hybrids with 0.4 relative carbon ratio is 134 GPa, 3.5 times higher than that of E-glass/epoxy composites. The stress-strain curve shows a clear "yield point" at 441 MPa and a well dispersed and gradual damage process
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.