Babu Anand, Abraham B Moses, Naskar Sudip, Ranpariya Spandan, Mandal Dipankar
Quantum Materials and Devices Unit, Institute of Nano Science and Technology, Knowledge City, Sector 81, Mohali, 140306, India.
Departament de Ciència de Materials i Química Física & Institut de Química Teòrica i Computacional (IQTCUB) Universitat de Barcelona, c/ Martí i Franquès 1-11, Barcelona, 08028, Spain.
Small. 2024 Dec;20(51):e2405393. doi: 10.1002/smll.202405393. Epub 2024 Oct 14.
Engineered poly(vinylidene fluoride) (PVDF) with its diverse crystalline phases plays a crucial role in determining the performance of devices in piezo-, pyro-, ferro- and tribo-electric applications, indicating the importance of distinct phase-detection in defining the structure-property relation. However, traditional characterization techniques struggle to effectively distinguish these phases, thereby failing to offer complete information. In this study, multimodal data-driven techniques have been employed for distinguishing different phases with a machine learning (ML) approach. This developed multimode model has been trained from empirical to theoretical data and demonstrates a classification accuracy of >94%, 15% more noise resilience, and 11% more accuracy from unimodality. Thus, from conception to validation, an alternative approach is provided to autonomously distinguish the different PVDF phases and eschew repetitive experiments that saved resources, thus accelerating the process of materials selection in various applications.
具有多种晶相的工程聚偏二氟乙烯（PVDF）在决定压电、热释电、铁电和摩擦电应用中器件的性能方面起着关键作用，这表明在定义结构-性能关系时进行独特的相检测非常重要。然而，传统的表征技术难以有效区分这些相，因此无法提供完整信息。在本研究中，采用了多模态数据驱动技术，通过机器学习（ML）方法区分不同的相。这个开发的多模态模型已经从经验数据训练到理论数据，其分类准确率>94%，抗噪声能力比单模态高15%，准确率比单模态高11%。因此，从概念到验证，提供了一种替代方法来自动区分不同的PVDF相，并避免了节省资源的重复实验，从而加速了各种应用中材料选择的过程。
