arXiv subjects

Alankar Alankar

Publications and source records attributed to Alankar Alankar.

10 recordsLinked to original sources

arXiv2026

Generative artificial intelligence for reliable mechanistic reasoning for corrosion

Corrosion accounts for approximately 4% of global GDP, and reliable prediction is essential for timely mitigation. Machine learning effectively predicts corrosion rates from composition, microstructure, and environmental variables, but cannot explain the underlying mechanisms. A reliable approach in safety-critical materials engineering requires not only accurate retrieval but also mechanistically defensible reasoning, a capability that existing factuality metrics cannot assess. This work presents a domain-adapted retrieval-augmented generation framework for corrosion knowledge synthesis, demonstrated on magnesium alloy corrosion. Three open-weight language models (Llama-3.1-8B, Qwen-2.5-7B, Mistral-7B) are fine-tuned on 3,309 expert-verified question-answer pairs from 840 peer-reviewed papers and integrated with a hybrid dense-lexical retrieval pipeline. Retrieval augmentation produces Token F1 gains of 143-194%, with system faithfulness of 0.964 and context recall of 0.988. Blind external validation on newly published literature and in-house electrochemical data confirms trend-level generalisation. Reason Map, a proposition-graph framework, is further introduced; it independently constructs directed evidence graphs from generated answers and retrieved literature, enabling systematic detection of causal direction inversions and unsupported inferential leaps that flat factuality metrics cannot expose. The modular architecture can be applied across domains, offering a generalizable blueprint for trustworthy AI-assisted knowledge synthesis to circumvent corrosion, which can also be applied to other engineering domains.

cs.LG↗

arXiv2026

Studying Creep-Fatigue interaction of Nickel-Based Superalloys using Crystal Plasticity and Entropy-Based life prediction model

Creep-fatigue interaction in single-crystal nickel superalloys is difficult to predict because the response depends on the combined effects of loading parameters, hold time, temperature, and the underlying deformation mechanisms. This is important for turbine blade applications, where components experience both fatigue and creep during service. In the present work, a crystal plasticity finite element (CPFE) framework is used to study the creep-fatigue response of a single-crystal nickel superalloy under a range of practically relevant thermo-mechanical loading conditions. In particular, the effects of strain amplitude, R-ratio, hold duration, and temperature on cyclic deformation, stress relaxation, damage evolution, and creep-fatigue life are examined. Particular attention is given to separate the roles of fatigue and creep damage, understanding their interaction, and identify the creep-dominated and fatigue-dominated regimes as a function of strain amplitude and hold time. The study brings together these effects within a single framework and shows that the predicted trends in cyclic response and life are in good agreement with experimental observations reported in the literature.

cond-mat.mtrl-sci↗

arXiv2025

Generative diffusion modeling protocols for improving the Kikuchi pattern indexing in electron back-scatter diffraction

Electron back-scatter diffraction (EBSD) has traditionally relied upon methods such as the Hough transform and dictionary Indexing to interpret diffraction patterns and extract crystallographic orientation. However, these methods encounter significant limitations, particularly when operating at high scanning speeds, where the exposure time per pattern is decreased beyond the operating sensitivity of CCD camera. Hence the signal to noise ratio decreases for the observed pattern which makes the pattern noisy, leading to reduced indexing accuracy. This research work aims to develop generative machine learning models for the post-processing or on-the-fly processing of Kikuchi patterns which are capable of restoring noisy EBSD patterns obtained at high scan speeds. These restored patterns can be used for the determination of crystal orientations to provide reliable indexing results. We compare the performance of such generative models in enhancing the quality of patterns captured at short exposure times (high scan speeds). An interesting observation is that the methodology is not data-hungry as typical machine learning methods.

cond-mat.mtrl-sci↗

arXiv2025

SHA-256 Infused Embedding-Driven Generative Modeling of High-Energy Molecules in Low-Data Regimes

High-energy materials (HEMs) are critical for propulsion and defense domains, yet their discovery remains constrained by experimental data and restricted access to testing facilities. This work presents a novel approach toward high-energy molecules by combining Long Short-Term Memory (LSTM) networks for molecular generation and Attentive Graph Neural Networks (GNN) for property predictions. We propose a transformative embedding space construction strategy that integrates fixed SHA-256 embeddings with partially trainable representations. Unlike conventional regularization techniques, this changes the representational basis itself, reshaping the molecular input space before learning begins. Without recourse to pretraining, the generator achieves 67.5% validity and 37.5% novelty. The generated library exhibits a mean Tanimoto coefficient of 0.214 relative to training set signifying the ability of framework to generate a diverse chemical space. We identified 37 new super explosives higher than 9 km/s predicted detonation velocity.

cs.LG↗

arXiv2024

Graph Residual based Method for Molecular Property Prediction

Machine learning-driven methods for property prediction have been of deep interest. However, much work remains to be done to improve the generalization ability, accuracy, and inference time for critical applications. The traditional machine learning models predict properties based on the features extracted from the molecules, which are often not easily available. In this work, a novel Deep Learning method, the Edge Conditioned Residual Graph Neural Network (ECRGNN), has been applied, allowing us to predict properties directly only the Graph-based structures of the molecules. SMILES (Simplified Molecular Input Line Entry System) representation of the molecules has been used in the present study as input data format, which has been further converted into a graph database, which constitutes the training data. This manuscript highlights a detailed description of the novel GRU-based methodology, ECRGNN, to map the inputs that have been used. Emphasis is placed on highlighting both the regressive property and the classification efficacy of the same. A detailed description of the Variational Autoencoder (VAE) and the end-to-end learning method used for multi-class multi-label property prediction has been provided as well. The results have been compared with standard benchmark datasets as well as some newly developed datasets. All performance metrics that have been used have been clearly defined, and their reason for choice.

q-bio.QM↗

arXiv2024

Towards Development of Automated Knowledge Maps and Databases for Materials Engineering using Large Language Models

In this work a Large Language Model (LLM) based workflow is presented that utilizes OpenAI ChatGPT model GPT-3.5-turbo-1106 and Google Gemini Pro model to create summary of text, data and images from research articles. It is demonstrated that by using a series of processing, the key information can be arranged in tabular form and knowledge graphs to capture underlying concepts. Our method offers efficiency and comprehension, enabling researchers to extract insights more effectively. Evaluation based on a diverse Scientific Paper Collection demonstrates our approach in facilitating discovery of knowledge. This work contributes to accelerated material design by smart literature review. The method has been tested based on various qualitative and quantitative measures of gathered information. The ChatGPT model achieved an F1 score of 0.40 for an exact match (ROUGE-1, ROUGE-2) but an impressive 0.479 for a relaxed match (ROUGE-L, ROUGE-Lsum) structural data format in performance evaluation. The Google Gemini Pro outperforms ChatGPT with an F1 score of 0.50 for an exact match and 0.63 for a relaxed match. This method facilitates high-throughput development of a database relevant to materials informatics. For demonstration, an example of data extraction and knowledge graph formation based on a manuscript about a titanium alloy is discussed.

cs.DL↗

arXiv2022

Predicting the formation and stability of oxide perovskites by extracting underlying mechanisms using machine learning

The optimization of properties of perovskite oxides has drawn interest on account of their diverse areas of application. In this work, the hierarchical clustering technique is used to reduce the multi-collinearity among selected features from literature that are reported to have an effect on perovskite formation and stability. Operating on the vast composition space of double oxide perovskite compositions available in literature and online repositories, in this manuscript, an attempt has been made to extract the relationship between the composition and structure to predict their formability and stability. Machine learning (ML) classifiers are trained on these datasets to predict novel stable perovskite compositions. The study uses a vast feature space to narrow down the most important factors affecting the formability and stability in perovskite compounds. It also identifies stable compositions that have band gaps suitable for photovoltaic and photocatalytic applications. The developed random forest (RF)-based models may be extended to include the implications beyond photosensitive applications by focusing on the physico-chemical mechanisms driving the phenomena behind each application.

cond-mat.mtrl-sci↗

arXiv2022

Physics-Informed Machine Learning and Uncertainty Quantification for Mechanics of Heterogeneous Materials

In this work, a model based on the Physics - Informed Neural Networks (PINNs) for solving elastic deformation of heterogeneous solids and associated Uncertainty Quantification (UQ) is presented. For the present study, the PINNs framework - Modulus developed by Nvidia is utilized, wherein we implement a module for mechanics of heterogeneous solids. We use PINNs to approximate momentum balance by assuming isotropic linear elastic constitutive behavior against a loss function. Along with governing equations, the associated initial / boundary conditions also softly participate in the loss function. Solids where the heterogeneity manifests as voids (low elastic modulus regions) and fibers (high elastic modulus regions) in a matrix are analyzed, and the results are validated against solutions obtained from a commercial Finite Element (FE) analysis package. The present study also reveals that PINNs can capture the stress jumps precisely at the material interfaces. Additionally, the present study explores the advantages associated with the surrogate features in PINNs via the variation in geometry and material properties. The presented UQ studies suggest that the mean and standard deviation of the PINNs solution are in good agreement with Monte Carlo FE results. The effective Young's modulus predicted by PINNs for single representative void and single fiber composites compare very well against the ones predicted by FE, which establishes the PINNs formulation as an efficient homogenization tool.

cond-mat.mtrl-sci↗

arXiv2022

An Experimentally Informed Continuum Grain Boundary Model

A continuum grain boundary model is developed that uses experimentally measured grain boundary energy data as a function of misorientation to simulate idealized grain boundary evolution in a 1-D grain array. The model uses a continuum representation of the misorientation in terms of spatial gradients of the orientation as a fundamental field. The grain boundary energy density employed is non-convex in this orientation gradient, based on physical grounds. Simple gradient descent dynamics of the energy are utilized for idealized microstructure evolution, which requires higher-order regularization of the energy density for the model to be well-set; the regularization is physically justified. Microstructure evolution is presented using two plausible energy density functions, both defined from the same experimental data: a 'smooth' and a 'cusp' energy density. Results of grain boundary equilibria and microstructure evolution representing grain reorientation in one space dimension are presented. The different shapes of the energy density functions representing a common data set are shown to result in different overall microstructural evolution of the system. Mathematically, the constructed energy functional formally is of the Aviles-Giga/Cross-Newell type but with unequal well-depths, resulting in a difference in the structural feature of solutions that can be identified with grain boundaries, as well as in the approach to equilibria from identical initial conditions. This study also investigates the metastability of grain boundaries. It supports the general thermodynamics belief that they persist for extended periods before eventually vanishing due to the lowest energy configuration favored by fluctuations over infinite time.

cond-mat.mtrl-sci↗

arXiv2020

Relative relevance of mobility and driving force on edge dislocation climb by the vacancy mechanism

In this work we examine the driving force for edge dislocations to climb in $α$-Fe from atomistic and mesoscale viewpoints. We study the bias for the climb process depending on the dislocation orientation and the applied stress as coming from both the gradient of the chemical potential and the transport coefficients. Both terms are modified by the applied stress and therefore contribute to climb. Surprisingly, even though the vacancy migration barrier distribution is modified by the external stress as obtained by nudged-elastic band calculations, the mobilities resulting from a kinetic Monte Carlo model applied on the obtained energy landscape are indistinguishable, independently of the stress. Moreover, an object kinetic Monte Carlo (OKMC) model including the effect of the dislocation strain field to first order shows indeed a slight anisotropic component in the diffusion in more complex dislocation configurations. However, the OKMC results highlight the fact that the thermodynamic component is the dominant driving force. We conclude that in $α$-Fe under thermal conditions, the main source of bias is given by the difference in vacancy chemical potentials, which is small enough to hinder the process for dynamic atomistic simulations.

cond-mat.mtrl-sci↗