Wu C.-TLin P.-HHuang S.-YTseng Y.-JChang H.-TLi S.-YYen H.-W.HUNG-WEI YEN2022-03-222022-03-22202225891529https://www.scopus.com/inward/record.uri?eid=2-s2.0-85122621771&doi=10.1016%2fj.mtla.2021.101313&partnerID=40&md5=6ecf9f66df419b461833737941a26e3dhttps://scholars.lib.ntu.edu.tw/handle/123456789/598383An ensemble neural networks model is developed to detangle the complex effects of alloying elements on phase stability and Young's modulus, enabling design guidelines for low-modulus Ti alloy. This computational framework, which is named βLow 2.0, is validated and examined to understand the sensitivity of the modulus to the alloy composition for material selections. The exercise provides design guidelines for future material designs considering metastability of β phase and effects of Nb, Zr, Mo, Sn, and Ta. To evaluate this data-driven model, a basic uncertainty quantification function is applied to understand the model in the present data space. The results of the model validation are also presented with experimental data different from that of the calibration dataset. This work enables comprehensive metallurgical principles for alloy design by using neural network-based machine learning. ? 2022 The Author(s)Artificial neural networkMachine learningPhase stabilityTitanium alloyYoung's modulusAlloying elementsDesignElastic moduliTitanium alloysAlloy designsComplex effectsComputational frameworkEnsemble neural networkLow-modulusNeural network modelTi alloysTitanium (alloys)Young modulusβ Ti alloyNeural networksjournal article10.1016/j.mtla.2021.1013132-s2.0-85122621771