請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/89974
標題: | 鎳鈦形狀記憶合金分子動力學模擬勢能比較與微結構分析 Comparison of Molecular Dynamics Simulation Potential and Microstructure Analysis of NiTi Shape Memory Alloys |
作者: | 馮紹祐 Shao-Yu Feng |
指導教授: | 陳志軒 Chih-Hsuan Chen |
關鍵字: | 形狀記憶合金,分子動力學模擬,鎳鈦,神經網路原子間勢能,麻田散體相轉變, Shape Memory Alloys (SMAs),Molecular Dynamics Simulation,Nickel Titanium,Neural Network Interatomic Potential (NNIP),Martensitic Transition, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 鎳鈦形狀記憶合金(NiTi)是一種特別的材料,具有形狀記憶和超彈性特性,使其在許多應用中,如生物醫學和航空航天領域,都有廣泛的應用。這種合金的特性來自其能在不同的溫度和應力條件下,會進行特殊的相變換。分子動力學模擬是一種功能強大的工具,可以用來研究這種相變換的微結構變換機制。透過模擬,我們可以觀察到原子級別的結構變化,並理解這些變化如何影響材料的宏觀行為。例如,我們可以模擬合金在受到外部壓力時的反應,並觀察其在冷卻和加熱過程中的形狀變化。這些模擬結果不僅可以幫助我們更好地理解鎳鈦形狀記憶合金的性質,也可以指導我們設計更有效的應用策略。然而分子動力學模擬中,傳統的原子間勢能函數難以準確模擬實際物理現象,可能導致預測誤差大。因此,本研究主要探討分子動力學模擬中的神經網路分子間勢能(NNIP)與傳統勢能函數的比較。我們將NNIP應用於鎳鈦形狀記憶合金的分子動力學模擬,並與第二近鄰修正嵌入原子勢能(2NN-MEAM)進行比較。我們使用由本研究團隊開發的麻田散體兄弟晶識別方法(MVIM),全面分析和評估兩者的準確性和相變化結果。我們建立一系列模擬流程,包括由溫度和應力誘導的相變化,以及不同晶向的壓印試驗。我們觀察NNIP在模擬不同晶向模型受到溫度或應力下誘發的麻田散體微結構演變的表現,並研究了模型在麻田散體相轉變過程中的兄弟晶分布和微結構變化。
本研究結果揭示,不同的模型邊界設定、變形邊界設定和晶向具有明顯的影響力,不僅會影響相轉變溫度,還會改變麻田散體兄弟晶的種類與分布。特別是週期性邊界、具有自由表面缺陷的開放式邊界與混合式邊界,它們在麻田散體相轉變溫度、兄弟晶種類及結構上展現出迥然不同的特性。最後,我們發現,NNIP勢能模型相對於2NN-MEAM勢能,能更準確地描繪鎳鈦形狀記憶合金的熱力學行為和結構變形行為,如塑性變形或自由表面缺陷。然而在特定條件下,NNIP勢能卻也展現出模擬結果的不確定性與應用上的限制。這些結果對於理解形狀記憶合金的性質,尤其是在不同條件下的變化,擁有深遠的理論和實用價值。 Nickel-titanium shape memory alloy (NiTi) is a unique material with shape memory and superelastic properties, making it widely applicable in numerous fields such as biomedicine and aerospace. The characteristics of this alloy stem from its special phase transformations under varying temperature and stress conditions. Molecular dynamics simulation is a powerful tool for studying the microstructural transformation mechanisms of these phase changes. Through simulation, we can observe atomic-level structural changes and understand how these changes affect the macroscopic behavior of the material. For instance, we can simulate the alloy's response to external pressure and observe its shape changes during cooling and heating processes. These simulation results not only help us better understand the properties of nickel-titanium shape memory alloys but also guide us in designing more effective application strategies. However, in molecular dynamics simulations, traditional interatomic potential functions struggle to accurately simulate actual physical phenomena, potentially leading to significant prediction errors. Therefore, this study primarily investigates the comparison of Neural Network Interatomic Potential (NNIP) and traditional potential functions in molecular dynamics simulations. We applied NNIP to the molecular dynamics simulation of nickel-titanium shape memory alloy, and compared it with the Second Nearest-Neighbor Modified Embedded Atom Method (2NN-MEAM) potential. We used the Martensite Variants Identification Method (MVIM), developed by our research team, to comprehensively analyze and evaluate the accuracy and phase change results of both. We established a series of simulation processes, including phase changes induced by temperature and stress, and indent tests of different crystal orientations. We observed the performance of NNIP in simulating the evolution of martensitic microstructures induced by temperature or stress in different crystal orientation models, and studied the distribution of twin variants and microstructural changes in the model during the martensitic phase transition. Our research results reveal that different model boundary settings, deformation boundary settings, and crystal orientations have a significant impact, not only affecting the phase transition temperature but also changing the types and distribution of martensitic variants. Especially the periodic boundaries and open boundaries with free surface defects, they show distinctly different characteristics in martensitic phase transition temperature, variants types, and structure. Finally, we found that the NNIP potential model can more accurately depict the thermodynamic behavior and structural deformation behavior of NiTi shape memory alloys compared to the 2NN-MEAM potential, such as plastic deformation or free surface defects. However, under specific conditions, NNIP is also demonstrated to have uncertainties in simulation results and limitations in application. These results have profound theoretical and practical value for understanding the properties of shape memory alloys, especially changes under different conditions. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/89974 |
DOI: | 10.6342/NTU202303544 |
全文授權: | 未授權 |
顯示於系所單位: | 機械工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-2.pdf 目前未授權公開取用 | 15.43 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。