Abstract A neural-network machine called “βLow” enables a high-throughput recommendation for new β titanium alloys with Young’s moduli lower than 50 GPa. The machine was trained by using a very general approach with small data from experiments. Its efficiency and accuracy break the barrier for alloy discovery. βLow’s best recommendation, Ti-12Nb-12Zr-12Sn (in wt.%) alloy, was unexpected in previous methods. This new alloy meets the requirements for bio-compatibility, low modulus, and low cost, and holds promise for orthopedic and prosthetic implants. Moreover, βLow’s prediction guides us to realize that the unexplored space of the chemical compositions of low-modulus biomedical titanium alloys is still large. Machine-learning-aided materials design accelerates the progress of materials development and reduces research costs in this work.

中文翻译：

---

机器学习推荐具有类骨模量的经济实惠的新型钛合金

摘要 一种名为“βLow”的神经网络机器能够对杨氏模量低于 50 GPa 的新型 β 钛合金进行高通量推荐。该机器是通过使用非常通用的方法和来自实验的小数据进行训练的。它的效率和准确性打破了合金发现的障碍。βLow 的最佳推荐，Ti-12Nb-12Zr-12Sn（重量百分比）合金，在之前的方法中是出乎意料的。这种新合金满足生物相容性、低模量和低成本的要求，有望用于骨科和假肢植入物。此外，βLow的预测指导我们认识到，低模量生物医用钛合金化学成分的未开发空间仍然很大。机器学习辅助材料设计加速了材料开发的进程并降低了这项工作的研究成本。