A titanium alloys design method based on high-throughput experiments and machine learning

Abstract

In this work, the effect of Mo and Cr on microstructure and mechanical properties of newly titanium alloys (Ti–3Al–2Nb-1.2V–1Zr–1Sn-xCr-yMo) was investigated, and a composition-microstructure-properties relationship was established by diffusion multiple. The microstructure characterization (volume fraction, size of α phases) for alloys with different molybdenum equivalent (Mo[q]) was predicted by machine learning (BP neural network), and the result shows a good agreement between the predicted results and experimental values. Combining diffusion multiple and BP neural network, a Ti alloy (Ti–3Al–2Nb-1.2V–1Zr–1Sn–4Cr–4Mo) with outstanding mechanical properties was successfully designed. The mechanical test result shows that excellent balance of strength (YS~1200 MPa) and plasticity (El~12%) can be achieved after the solution treatment at 750 °C and aging at 550 °C for 6 h. During deformation, Primary globular primary α phases were elongated, and secondary acicular α phases resisted the dislocation slipping, which provides good plasticity and strength for the alloys, respectively.

Keywords

• High-throughput methods
• Ti alloys
• Machine learning
• Molybdenum equivalent