The dynamics of biological systems, from proteins to cells to organisms, is complex and stochastic. To decipher their physical laws, we need to bridge between experimental observations and theoretical modeling. Thanks to progress in microscopy and tracking, there is today an abundance of experimental trajectories reflecting these dynamical laws. Inferring physical models from noisy and imperfect experimental data, however, is challenging. Because there are no inference methods that are robust and efficient, model reconstruction from experimental trajectories is a bottleneck to data-driven biophysics. In this Thesis, I present a set of tools developed to bridge this gap and permit robust and universal inference of stochastic dynamical models from experimental trajectories. These methods are rooted in an information-theoretical framework that quantifies how much can be inferred from trajectories that are short, partial and noisy. They permit the efficient inference of dynamical models for overdamped and underdamped Langevin systems, as well as the inference of entropy production rates. I finally present early applications of these techniques, as well as future research directions.
{"title":"Learning dynamical models from stochastic trajectories","authors":"Pierre Ronceray","doi":"arxiv-2406.02363","DOIUrl":"https://doi.org/arxiv-2406.02363","url":null,"abstract":"The dynamics of biological systems, from proteins to cells to organisms, is\u0000complex and stochastic. To decipher their physical laws, we need to bridge\u0000between experimental observations and theoretical modeling. Thanks to progress\u0000in microscopy and tracking, there is today an abundance of experimental\u0000trajectories reflecting these dynamical laws. Inferring physical models from\u0000noisy and imperfect experimental data, however, is challenging. Because there\u0000are no inference methods that are robust and efficient, model reconstruction\u0000from experimental trajectories is a bottleneck to data-driven biophysics. In\u0000this Thesis, I present a set of tools developed to bridge this gap and permit\u0000robust and universal inference of stochastic dynamical models from experimental\u0000trajectories. These methods are rooted in an information-theoretical framework\u0000that quantifies how much can be inferred from trajectories that are short,\u0000partial and noisy. They permit the efficient inference of dynamical models for\u0000overdamped and underdamped Langevin systems, as well as the inference of\u0000entropy production rates. I finally present early applications of these\u0000techniques, as well as future research directions.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"34 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ella M. King, Megan C. Engel, Caroline Martin, Alp M. Sunol, Qian-Ze Zhu, Sam S. Schoenholz, Vinothan N. Manoharan, Michael P. Brenner
Accurate interaction potentials between microscopic components such as colloidal particles or cells are crucial to understanding a range of processes, including colloidal crystallization, bacterial colony formation, and cancer metastasis. Even in systems where the precise interaction mechanisms are unknown, effective interactions can be measured to inform simulation and design. However, these measurements are difficult and time-intensive, and often require conditions that are drastically different from in situ conditions of the system of interest. Moreover, existing methods of measuring interparticle potentials rely on constraining a small number of particles at equilibrium, placing limits on which interactions can be measured. We introduce a method for inferring interaction potentials directly from trajectory data of interacting particles. We explicitly solve the equations of motion to find a form of the potential that maximizes the probability of observing a known trajectory. Our method is valid for systems both in and out of equilibrium, is well-suited to large numbers of particles interacting in typical system conditions, and does not assume a functional form of the interaction potential. We apply our method to infer the interactions of colloidal spheres from experimental data, successfully extracting the range and strength of a depletion interaction from the motion of the particles.
{"title":"Inferring interaction potentials from stochastic particle trajectories","authors":"Ella M. King, Megan C. Engel, Caroline Martin, Alp M. Sunol, Qian-Ze Zhu, Sam S. Schoenholz, Vinothan N. Manoharan, Michael P. Brenner","doi":"arxiv-2406.01522","DOIUrl":"https://doi.org/arxiv-2406.01522","url":null,"abstract":"Accurate interaction potentials between microscopic components such as\u0000colloidal particles or cells are crucial to understanding a range of processes,\u0000including colloidal crystallization, bacterial colony formation, and cancer\u0000metastasis. Even in systems where the precise interaction mechanisms are\u0000unknown, effective interactions can be measured to inform simulation and\u0000design. However, these measurements are difficult and time-intensive, and often\u0000require conditions that are drastically different from in situ conditions of\u0000the system of interest. Moreover, existing methods of measuring interparticle\u0000potentials rely on constraining a small number of particles at equilibrium,\u0000placing limits on which interactions can be measured. We introduce a method for\u0000inferring interaction potentials directly from trajectory data of interacting\u0000particles. We explicitly solve the equations of motion to find a form of the\u0000potential that maximizes the probability of observing a known trajectory. Our\u0000method is valid for systems both in and out of equilibrium, is well-suited to\u0000large numbers of particles interacting in typical system conditions, and does\u0000not assume a functional form of the interaction potential. We apply our method\u0000to infer the interactions of colloidal spheres from experimental data,\u0000successfully extracting the range and strength of a depletion interaction from\u0000the motion of the particles.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141253349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mahindra Rautela, Alan Williams, Alexander Scheinker
Particle accelerators are time-varying systems whose components are perturbed by external disturbances. Tuning accelerators can be a time-consuming process involving manual adjustment of multiple components, such as RF cavities, to minimize beam loss due to time-varying drifts. The high dimensionality of the system ($sim$100 amplitude and phase RF settings in the LANSCE accelerator) makes it difficult to achieve optimal operation. The time-varying drifts and the dimensionality make system parameter estimation a challenging optimization problem. In this work, we propose a Variational Autoencoded Latent Regression (VALeR) model for robust estimation of system parameters using 2D unique projections of a charged particle beam's 6D phase space. In VALeR, VAE projects the phase space projections into a lower-dimensional latent space, and a dense neural network maps the latent space onto the space of system parameters. The trained network can predict system parameters for unseen phase space projections. Furthermore, VALeR can generate new projections by randomly sampling the latent space of VAE and also estimate the corresponding system parameters.
{"title":"Accelerator system parameter estimation using variational autoencoded latent regression","authors":"Mahindra Rautela, Alan Williams, Alexander Scheinker","doi":"arxiv-2406.01532","DOIUrl":"https://doi.org/arxiv-2406.01532","url":null,"abstract":"Particle accelerators are time-varying systems whose components are perturbed\u0000by external disturbances. Tuning accelerators can be a time-consuming process\u0000involving manual adjustment of multiple components, such as RF cavities, to\u0000minimize beam loss due to time-varying drifts. The high dimensionality of the\u0000system ($sim$100 amplitude and phase RF settings in the LANSCE accelerator)\u0000makes it difficult to achieve optimal operation. The time-varying drifts and\u0000the dimensionality make system parameter estimation a challenging optimization\u0000problem. In this work, we propose a Variational Autoencoded Latent Regression\u0000(VALeR) model for robust estimation of system parameters using 2D unique\u0000projections of a charged particle beam's 6D phase space. In VALeR, VAE projects\u0000the phase space projections into a lower-dimensional latent space, and a dense\u0000neural network maps the latent space onto the space of system parameters. The\u0000trained network can predict system parameters for unseen phase space\u0000projections. Furthermore, VALeR can generate new projections by randomly\u0000sampling the latent space of VAE and also estimate the corresponding system\u0000parameters.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"16 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mahindra Rautela, Alan Williams, Alexander Scheinker
Addressing the charged particle beam diagnostics in accelerators poses a formidable challenge, demanding high-fidelity simulations in limited computational time. Machine learning (ML) based surrogate models have emerged as a promising tool for non-invasive charged particle beam diagnostics. Trained ML models can make predictions much faster than computationally expensive physics simulations. In this work, we have proposed a temporally structured variational autoencoder model to autoregressively forecast the spatiotemporal dynamics of the 15 unique 2D projections of 6D phase space of charged particle beam as it travels through the LANSCE linear accelerator. In the model, VAE embeds the phase space projections into a lower dimensional latent space. A long-short-term memory network then learns the temporal correlations in the latent space. The trained network can evolve the phase space projections across further modules provided the first few modules as inputs. The model predicts all the projections across different modules with low mean squared error and high structural similarity index.
{"title":"Towards latent space evolution of spatiotemporal dynamics of six-dimensional phase space of charged particle beams","authors":"Mahindra Rautela, Alan Williams, Alexander Scheinker","doi":"arxiv-2406.01535","DOIUrl":"https://doi.org/arxiv-2406.01535","url":null,"abstract":"Addressing the charged particle beam diagnostics in accelerators poses a\u0000formidable challenge, demanding high-fidelity simulations in limited\u0000computational time. Machine learning (ML) based surrogate models have emerged\u0000as a promising tool for non-invasive charged particle beam diagnostics. Trained\u0000ML models can make predictions much faster than computationally expensive\u0000physics simulations. In this work, we have proposed a temporally structured\u0000variational autoencoder model to autoregressively forecast the spatiotemporal\u0000dynamics of the 15 unique 2D projections of 6D phase space of charged particle\u0000beam as it travels through the LANSCE linear accelerator. In the model, VAE\u0000embeds the phase space projections into a lower dimensional latent space. A\u0000long-short-term memory network then learns the temporal correlations in the\u0000latent space. The trained network can evolve the phase space projections across\u0000further modules provided the first few modules as inputs. The model predicts\u0000all the projections across different modules with low mean squared error and\u0000high structural similarity index.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"47 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252721","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Asphalt concrete's (AC) durability and maintenance demands are strongly influenced by its fatigue life. Traditional methods for determining this characteristic are both resource-intensive and time-consuming. This study employs artificial neural networks (ANNs) to predict AC fatigue life, focusing on the impact of strain level, binder content, and air-void content. Leveraging a substantial dataset, we tailored our models to effectively handle the wide range of fatigue life data, typically represented on a logarithmic scale. The mean square logarithmic error was utilized as the loss function to enhance prediction accuracy across all levels of fatigue life. Through comparative analysis of various hyperparameters, we developed a machine-learning model that captures the complex relationships within the data. Our findings demonstrate that higher binder content significantly enhances fatigue life, while the influence of air-void content is more variable, depending on binder levels. Most importantly, this study provides insights into the intricacies of using ANNs for modeling, showcasing their potential utility with larger datasets. The codes developed and the data used in this study are provided as open source on a GitHub repository, with a link included in the paper for full access.
{"title":"Predicting the fatigue life of asphalt concrete using neural networks","authors":"Jakub Houlík, Jan Valentin, Václav Nežerka","doi":"arxiv-2406.01523","DOIUrl":"https://doi.org/arxiv-2406.01523","url":null,"abstract":"Asphalt concrete's (AC) durability and maintenance demands are strongly\u0000influenced by its fatigue life. Traditional methods for determining this\u0000characteristic are both resource-intensive and time-consuming. This study\u0000employs artificial neural networks (ANNs) to predict AC fatigue life, focusing\u0000on the impact of strain level, binder content, and air-void content. Leveraging\u0000a substantial dataset, we tailored our models to effectively handle the wide\u0000range of fatigue life data, typically represented on a logarithmic scale. The\u0000mean square logarithmic error was utilized as the loss function to enhance\u0000prediction accuracy across all levels of fatigue life. Through comparative\u0000analysis of various hyperparameters, we developed a machine-learning model that\u0000captures the complex relationships within the data. Our findings demonstrate\u0000that higher binder content significantly enhances fatigue life, while the\u0000influence of air-void content is more variable, depending on binder levels.\u0000Most importantly, this study provides insights into the intricacies of using\u0000ANNs for modeling, showcasing their potential utility with larger datasets. The\u0000codes developed and the data used in this study are provided as open source on\u0000a GitHub repository, with a link included in the paper for full access.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"22 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252729","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Camila Pazos, Shuchin Aeron, Pierre-Hugues Beauchemin, Vincent Croft, Martin Klassen, Taritree Wongjirad
The unfolding of detector effects in experimental data is critical for enabling precision measurements in high-energy physics. However, traditional unfolding methods face challenges in scalability, flexibility, and dependence on simulations. We introduce a novel unfolding approach using conditional Denoising Diffusion Probabilistic Models (cDDPM). Our method utilizes the cDDPM for a non-iterative, flexible posterior sampling approach, which exhibits a strong inductive bias that allows it to generalize to unseen physics processes without explicitly assuming the underlying distribution. We test our approach by training a single cDDPM to perform multidimensional particle-wise unfolding for a variety of physics processes, including those not seen during training. Our results highlight the potential of this method as a step towards a "universal" unfolding tool that reduces dependence on truth-level assumptions.
{"title":"Towards Universal Unfolding of Detector Effects in High-Energy Physics using Denoising Diffusion Probabilistic Models","authors":"Camila Pazos, Shuchin Aeron, Pierre-Hugues Beauchemin, Vincent Croft, Martin Klassen, Taritree Wongjirad","doi":"arxiv-2406.01507","DOIUrl":"https://doi.org/arxiv-2406.01507","url":null,"abstract":"The unfolding of detector effects in experimental data is critical for\u0000enabling precision measurements in high-energy physics. However, traditional\u0000unfolding methods face challenges in scalability, flexibility, and dependence\u0000on simulations. We introduce a novel unfolding approach using conditional\u0000Denoising Diffusion Probabilistic Models (cDDPM). Our method utilizes the cDDPM\u0000for a non-iterative, flexible posterior sampling approach, which exhibits a\u0000strong inductive bias that allows it to generalize to unseen physics processes\u0000without explicitly assuming the underlying distribution. We test our approach\u0000by training a single cDDPM to perform multidimensional particle-wise unfolding\u0000for a variety of physics processes, including those not seen during training.\u0000Our results highlight the potential of this method as a step towards a\u0000\"universal\" unfolding tool that reduces dependence on truth-level assumptions.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"67 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252759","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
High-energy physics experiments rely heavily on precise measurements of energy and momentum, yet face significant challenges due to detector limitations, calibration errors, and the intrinsic nature of particle interactions. Traditional unfolding techniques have been employed to correct for these distortions, yet they often suffer from model dependency and stability issues. We present a novel method, SwdFold, which utilizes the principles of optimal transport to provide a robust, model-independent framework to estimate the probability density ratio for data unfolding. It not only unfold the toy experimental event by reweighted simulated data distributions closely with true distributions but also maintains the integrity of physical features across various observables. We can expect it can enable more reliable predictions and comprehensive analyses as a high precision reweighting and unfolding tool in high-energy physics.
{"title":"SwdFold:A Reweighting and Unfolding method based on Optimal Transport Theory","authors":"Chu-Cheng Pan, Xiang Dong, Yu-Chang Sun, Ao-Yan Cheng, Ao-Bo Wang, Yu-Xuan Hu, Hao Cai","doi":"arxiv-2406.01635","DOIUrl":"https://doi.org/arxiv-2406.01635","url":null,"abstract":"High-energy physics experiments rely heavily on precise measurements of\u0000energy and momentum, yet face significant challenges due to detector\u0000limitations, calibration errors, and the intrinsic nature of particle\u0000interactions. Traditional unfolding techniques have been employed to correct\u0000for these distortions, yet they often suffer from model dependency and\u0000stability issues. We present a novel method, SwdFold, which utilizes the\u0000principles of optimal transport to provide a robust, model-independent\u0000framework to estimate the probability density ratio for data unfolding. It not\u0000only unfold the toy experimental event by reweighted simulated data\u0000distributions closely with true distributions but also maintains the integrity\u0000of physical features across various observables. We can expect it can enable\u0000more reliable predictions and comprehensive analyses as a high precision\u0000reweighting and unfolding tool in high-energy physics.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"4 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252728","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this research, we propose the first approach for integrating the Kolmogorov-Arnold Network (KAN) with various pre-trained Convolutional Neural Network (CNN) models for remote sensing (RS) scene classification tasks using the EuroSAT dataset. Our novel methodology, named KCN, aims to replace traditional Multi-Layer Perceptrons (MLPs) with KAN to enhance classification performance. We employed multiple CNN-based models, including VGG16, MobileNetV2, EfficientNet, ConvNeXt, ResNet101, and Vision Transformer (ViT), and evaluated their performance when paired with KAN. Our experiments demonstrated that KAN achieved high accuracy with fewer training epochs and parameters. Specifically, ConvNeXt paired with KAN showed the best performance, achieving 94% accuracy in the first epoch, which increased to 96% and remained consistent across subsequent epochs. The results indicated that KAN and MLP both achieved similar accuracy, with KAN performing slightly better in later epochs. By utilizing the EuroSAT dataset, we provided a robust testbed to investigate whether KAN is suitable for remote sensing classification tasks. Given that KAN is a novel algorithm, there is substantial capacity for further development and optimization, suggesting that KCN offers a promising alternative for efficient image analysis in the RS field.
在这项研究中，我们利用 EuroSAT 数据集，首次提出了将 Kolmogorov-Arnold 网络（KAN）与各种预先训练好的卷积神经网络（CNN）模型相结合的方法，用于遥感（RS）场景分类任务。我们的新方法被命名为 KCN，旨在用 KAN 替代传统的多层感知器（MLP），以提高分类性能。我们采用了多种基于 CNN 的模型，包括 VGG16、MobileNetV2、EfficientNet、ConvNeXt、ResNet101 和 Vision Transformer (ViT)，并评估了它们与 KAN 配对后的性能。我们的实验证明，KAN 可以用较少的训练历时和参数实现较高的准确率。具体来说，与 KAN 配对的 ConvNeXt 表现最佳，在第一个训练周期中达到 94% 的准确率，在随后的训练周期中准确率提高到 96%，并且保持一致。结果表明，KAN 和 MLP 的准确率相近，KAN 在后期的表现略好。通过利用 EuroSAT 数据集，我们为研究 KAN 是否适用于遥感分类任务提供了一个稳健的测试平台。鉴于 KAN 是一种新型算法，有很大的进一步开发和优化空间，这表明 KCN 为遥感领域的高效图像分析提供了一种很有前途的替代方法。
{"title":"Kolmogorov-Arnold Network for Satellite Image Classification in Remote Sensing","authors":"Minjong Cheon","doi":"arxiv-2406.00600","DOIUrl":"https://doi.org/arxiv-2406.00600","url":null,"abstract":"In this research, we propose the first approach for integrating the\u0000Kolmogorov-Arnold Network (KAN) with various pre-trained Convolutional Neural\u0000Network (CNN) models for remote sensing (RS) scene classification tasks using\u0000the EuroSAT dataset. Our novel methodology, named KCN, aims to replace\u0000traditional Multi-Layer Perceptrons (MLPs) with KAN to enhance classification\u0000performance. We employed multiple CNN-based models, including VGG16,\u0000MobileNetV2, EfficientNet, ConvNeXt, ResNet101, and Vision Transformer (ViT),\u0000and evaluated their performance when paired with KAN. Our experiments\u0000demonstrated that KAN achieved high accuracy with fewer training epochs and\u0000parameters. Specifically, ConvNeXt paired with KAN showed the best performance,\u0000achieving 94% accuracy in the first epoch, which increased to 96% and remained\u0000consistent across subsequent epochs. The results indicated that KAN and MLP\u0000both achieved similar accuracy, with KAN performing slightly better in later\u0000epochs. By utilizing the EuroSAT dataset, we provided a robust testbed to\u0000investigate whether KAN is suitable for remote sensing classification tasks.\u0000Given that KAN is a novel algorithm, there is substantial capacity for further\u0000development and optimization, suggesting that KCN offers a promising\u0000alternative for efficient image analysis in the RS field.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"181 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Shengyu Tao, Mengtian Zhang, Zixi Zhao, Haoyang Li, Ruifei Ma, Yunhong Che, Xin Sun, Lin Su, Xiangyu Chen, Zihao Zhou, Heng Chang, Tingwei Cao, Xiao Xiao, Yaojun Liu, Wenjun Yu, Zhongling Xu, Yang Li, Han Hao, Xuan Zhang, Xiaosong Hu, Guangmin ZHou
Manufacturing complexities and uncertainties have impeded the transition from material prototypes to commercial batteries, making prototype verification critical to quality assessment. A fundamental challenge involves deciphering intertwined chemical processes to characterize degradation patterns and their quantitative relationship with battery performance. Here we show that a physics-informed machine learning approach can quantify and visualize temporally resolved losses concerning thermodynamics and kinetics only using electric signals. Our method enables non-destructive degradation pattern characterization, expediting temperature-adaptable predictions of entire lifetime trajectories, rather than end-of-life points. The verification speed is 25 times faster yet maintaining 95.1% accuracy across temperatures. Such advances facilitate more sustainable management of defective prototypes before massive production, establishing a 19.76 billion USD scrap material recycling market by 2060 in China. By incorporating stepwise charge acceptance as a measure of the initial manufacturing variability of normally identical batteries, we can immediately identify long-term degradation variations. We attribute the predictive power to interpreting machine learning insights using material-agnostic featurization taxonomy for degradation pattern decoupling. Our findings offer new possibilities for dynamic system analysis, such as battery prototype degradation, demonstrating that complex pattern evolutions can be accurately predicted in a non-destructive and data-driven fashion by integrating physics-informed machine learning.
{"title":"Non-destructive Degradation Pattern Decoupling for Ultra-early Battery Prototype Verification Using Physics-informed Machine Learning","authors":"Shengyu Tao, Mengtian Zhang, Zixi Zhao, Haoyang Li, Ruifei Ma, Yunhong Che, Xin Sun, Lin Su, Xiangyu Chen, Zihao Zhou, Heng Chang, Tingwei Cao, Xiao Xiao, Yaojun Liu, Wenjun Yu, Zhongling Xu, Yang Li, Han Hao, Xuan Zhang, Xiaosong Hu, Guangmin ZHou","doi":"arxiv-2406.00276","DOIUrl":"https://doi.org/arxiv-2406.00276","url":null,"abstract":"Manufacturing complexities and uncertainties have impeded the transition from\u0000material prototypes to commercial batteries, making prototype verification\u0000critical to quality assessment. A fundamental challenge involves deciphering\u0000intertwined chemical processes to characterize degradation patterns and their\u0000quantitative relationship with battery performance. Here we show that a\u0000physics-informed machine learning approach can quantify and visualize\u0000temporally resolved losses concerning thermodynamics and kinetics only using\u0000electric signals. Our method enables non-destructive degradation pattern\u0000characterization, expediting temperature-adaptable predictions of entire\u0000lifetime trajectories, rather than end-of-life points. The verification speed\u0000is 25 times faster yet maintaining 95.1% accuracy across temperatures. Such\u0000advances facilitate more sustainable management of defective prototypes before\u0000massive production, establishing a 19.76 billion USD scrap material recycling\u0000market by 2060 in China. By incorporating stepwise charge acceptance as a\u0000measure of the initial manufacturing variability of normally identical\u0000batteries, we can immediately identify long-term degradation variations. We\u0000attribute the predictive power to interpreting machine learning insights using\u0000material-agnostic featurization taxonomy for degradation pattern decoupling.\u0000Our findings offer new possibilities for dynamic system analysis, such as\u0000battery prototype degradation, demonstrating that complex pattern evolutions\u0000can be accurately predicted in a non-destructive and data-driven fashion by\u0000integrating physics-informed machine learning.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"22 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141253123","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Contrast variation small-angle neutron scattering (CV-SANS) is a powerful tool to evaluate the structure of multi-component systems by decomposing scattering intensities $I$ measured with different scattering contrasts into partial scattering functions $S$ of self- and cross-correlations between components. The measured $I$ contains a measurement error, $Delta I$, and $Delta I$ results in an uncertainty of partial scattering functions, $Delta S$. However, the error propagation from $Delta I$ to $Delta S$ has not been quantitatively clarified. In this work, we have established deterministic and statistical approaches to determine $Delta S$ from $Delta I$. We have applied the two methods to experimental SANS data of polyrotaxane solutions with different contrasts, and have successfully estimated the errors of $S$. The quantitative error estimation of $S$ offers us a strategy to optimize the combination of scattering contrasts to minimize error propagation.
{"title":"Error evaluation of partial scattering functions obtained from contrast variation small-angle neutron scattering","authors":"Koichi Mayumi, Shinya Miyajima, Ippei Obayashi, Kazuaki Tanaka","doi":"arxiv-2406.00311","DOIUrl":"https://doi.org/arxiv-2406.00311","url":null,"abstract":"Contrast variation small-angle neutron scattering (CV-SANS) is a powerful\u0000tool to evaluate the structure of multi-component systems by decomposing\u0000scattering intensities $I$ measured with different scattering contrasts into\u0000partial scattering functions $S$ of self- and cross-correlations between\u0000components. The measured $I$ contains a measurement error, $Delta I$, and\u0000$Delta I$ results in an uncertainty of partial scattering functions, $Delta\u0000S$. However, the error propagation from $Delta I$ to $Delta S$ has not been\u0000quantitatively clarified. In this work, we have established deterministic and\u0000statistical approaches to determine $Delta S$ from $Delta I$. We have applied\u0000the two methods to experimental SANS data of polyrotaxane solutions with\u0000different contrasts, and have successfully estimated the errors of $S$. The\u0000quantitative error estimation of $S$ offers us a strategy to optimize the\u0000combination of scattering contrasts to minimize error propagation.","PeriodicalId":501065,"journal":{"name":"arXiv - PHYS - Data Analysis, Statistics and Probability","volume":"43 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141252761","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：