Indian Institute of Technology Patna
Recent publications
The construction of any infrastructure on highly compressible soil (HCS) is considered as problematic due to its low bearing capacity, and high swelling/shrinkage behavior. Earlier, the stabilization of problematic clay soils is achieved commonly by using various chemical additives alone/or in combination with other pozzolanic materials such as bentonite, fly ash, and slag. Recently, stabilization of soil by using diminutive amount of nano-additives alone or with conventional stabilizing materials is an emerging trend. Therefore, the present study investigates the effect of nano-clay (NC) on the engineering and leachability behavior of both HCS and lime-treatment HCS. A series of unconfined compressive strength (UCS), consolidation, permeability, leaching potential, physicochemical, microstructural, and mineralogical analysis have been carried out for its possible use in various civil/geotechnical engineering projects. It is revealed that nano-clay, irrespective of lime treatment and curing period, has caused an adverse effect to the UCS of HCS. Interestingly, one-dimensional swelling percentage of untreated and treated HCS has reduced, whereas compressibility of samples has increased with nano-clay percentage. Further, application of nano-clay has reduced the hydraulic conductivity significantly of flocculated lime-treated HCS. Leaching potential of heavy metal ions, namely, Cd, Mg, and Zn through the HCS-NC and lime-treated HCS-NC mixed is found to be lower than the allowable limit. Finally, mechanism of alteration in physical and engineering properties of both untreated and treated samples with nano-clay percentage is elucidated by performing physicochemical, mineralogical, and microstructural changes.
The cold-formed steel (CFS) members are highly susceptible to the fire and the presence of full or partial insulation can significantly influence the overall thermal profile of the member. Unlike uniformly heated members, a gradient thermal exposure can reduce the member’s mechanical properties unevenly. This study investigated the effects of uniform and gradient thermal exposure on the structural behaviour of flexural members. The non-linear finite element (FE) model is developed and validated with the experimental and numerical results available in the existing literature. A series of numerical FE parametric studies on 1425 members is performed considering several member geometry and spans ranging for different beams covering non-dimensional slenderness ranging from 0.28 to 1.81. Two common loading patterns (4-point loading and uniform moment) and five thermal distribution patterns are considered, covering thermal bowing in the direction of loading as well as in opposite to that. Results of the extensive parametric study indicate the fact that the failure temperature of the member is largely dependent on the applied thermal profile of the member. Parameters like depth of the member cross-section, non-dimensional slenderness and thermal bowing of the member, which largely influenced the critical temperature of the CFS flexural member, are studied in detail. In several cases, the failure temperature of a partially heated member can be lower than that of the member fully exposed to fire. Parametric study results also highlighted the fact that the existing limiting temperature of 350° of the European design rules (Eurocode 3, Part 1.2) for CFS members is highly over-conservative.
Cervical cancer is one of the most fatal and prevalent illnesses affecting women globally. Early detection of cervical cancer is crucial for effective treatment. Pap smear tests are commonly used, but population-based screening is time-consuming, expensive, and requires expert physicians. Computer-Aided Diagnosis (CAD) has shown promise in addressing this challenge. However, accurately predicting the disease using a single model can be difficult due to the complex data patterns involved. This research proposes a multi-stage architecture to improve cervical cancer screening. Initially, three pre-trained models are employed for image classification, after which the proposed advanced fusion technique is applied to combine the predictions. Additionally, we introduce a filtering approach in the third stage to refine the predictions. Unlike traditional fusion methods, the proposed architecture considers the confidence score of the base classifiers in making the final predictions on test samples. To enhance the performance of the models, we incorporate advanced augmentation techniques, including CutMix, CutOut, and MixUp. We assessed the performance of the proposed framework using a 5-fold cross-validation technique on two benchmark datasets. We evaluated the performance of the proposed framework through 5-fold cross-validation on two benchmark datasets. Remarkably, our framework achieved a classification accuracy of 97.62% and an F1-score of 97.64% on the SIPaKMeD dataset, demonstrating its effectiveness in accurately categorizing various cell types in the dataset.
Increasing the Te content in stoichiometric Bi 0.5 Sb 1.5 Te 3 facilitates effective control over the anti-site defects and nanostructure; however, arresting excess Te in the host matrix is challenging. Herein, we report the success of a saturation-annealing treatment in a vacuum, followed by air-quenching as a promising approach for synthesizing high figure-of-merit ( zT ) Bi 0.5 Sb 1.5 Te 3 +xTe (x = 0, 2, 5 and 10 wt%) materials. A remarkably high-power factor ( α 2 σ ~ 6 mW at 300 K) is achieved in p -type Bi 0.5 Sb 1.5 Te 3 + 5 wt% Te composition due to high carrier concentration ( n ) and good carrier mobility ( µ ). Microstructural analysis revealed the formation of densely interconnected polycrystalline grains featuring fine grain boundaries, planar/point defects, and strain field domains, contributing towards wide-length scale phonon scattering. The cumulative effect of drastically reduced thermal conductivity (κ ~ 0.8 W/m-K at 300 K), and enhanced power factor resulted in a record zT value ~ 2.2 at 300 K in Bi 0.5 Sb 1.5 Te 3 + 5 wt% Te, with an average zT value up to 1.35 in temperatures ranging from 303 to 573 K. The COMSOL simulations predict a maximum conversion efficiency ( η max ) of ~ 15%, at a temperature gradient ( ∆T ) of 270 K, for a single-leg thermoelectric generator (TEG) developed using this material.
In this article, we introduce the nonlinear fractional Lane–Emden equations utilizing the Caputo fractional derivatives of orders αα \alpha and ββ \beta . The fractional Lane–Emden equation adds fractional derivatives to the classical Lane–Emden framework to make it easier to model complicated physical phenomena. We have developed a collocation method, namely, the uniform fractional Haar wavelet collocation method, and used it to compute solutions. The proposed method combines the quasilinearization method with the uniform Haar wavelet collocation method. This approach uses fractional Haar integrations to identify the linear system, which when solved yields the desired solution. Our findings suggest that as the values of (α,β)(α,β) \left(\alpha, \beta \right) approach (2,1)(2,1) \left(2,1\right) , the solutions of the fractional and classical Lane–Emden become identical.
Magnetorheological (MR) fluids are smart composites that can exhibit reversible behavior changes and quickly transit from a liquid state to a nearly solid state when exposed to magnetic fields. Existing theories, such as the Herschel–Bulkley (H-B) model and artificial neural network, are frequently employed to describe the deformation of MR fluids; however, these are limited to certain modes of deformation and have lagged in the physical interpretation of magneto–mechanical interaction. To address the limitations, this study provides an in-depth investigation of MR fluid behavior, highlighting its non-Newtonian characteristics and the occurrence of a yielding phenomenon under finite deformation. A unified framework is developed to describe solid and fluid characteristics using classical continuum mechanics and Ericksen's seminal work consistent with the second law of thermodynamics. To highlight the distinctive characteristics of MR fluids, the investigation explores different deformation modes, including shear, flow, and squeezing. This study comprehensively addresses essential phenomena in MR fluids, including yielding, rate-dependent viscosity, and the Weissenberg effect, by leveraging the physical insights from their interaction with magnetic fields. We performed experiments with a rheometer to validate these conclusions and contrasted the analytical findings with the experimental data. Additionally, we confirmed the theoretical predictions for flow mode deformation by comparing them with existing experiments, offering a thorough explanation of field-dependent flow properties. The presented theory is also used to analyze the breaking for the MR brake system analytically, which is then validated by comparing it with the experiment. The proposed framework serves as a valuable tool for understanding and predicting the behavior of MR fluids, enabling the realization of their full potential in practical applications, such as MR fluid-based clutches and vibration dampers.
This article deals with a class of nonsmooth multiobjective geodesic pseudolinear programming problems (in short, (NMGPP)) in terms of bifunctions in the framework of Hadamard manifolds. Employing Slater-type constraint qualification, we establish the necessary criteria of Pareto efficiency for (NMGPP). Moreover, we establish sufficient criteria of Pareto efficiency for the considered problem (NMGPP). The Mond-Weir type dual model (in short, (MWDP)) related to (NMGPP) is formulated. Weak and strong duality results that relate (NMGPP) and (MWDP) are derived. Moreover, we employ the concept of proper Pareto efficiency to relate (MWDP) and (NMGPP). Non-trivial numerical examples are furnished to validate the significance of the results deduced in this paper. As far as the best of our knowledge is concerned, this is the first time that optimality conditions and duality results for (NMGPP) have been derived in terms of bifunctions in the Hadamard manifold framework.
Background Epileptic seizure detection using electroencephalogram (EEG) signals is crucial for automated diagnosis and monitoring of neurological disorders. Traditional machine learning and deep learning models often face challenges in capturing subtle seizure patterns and maintaining high generalization across datasets. To address these limitations, this study proposes ACE-SeizNet, a CNN-based attention-enhanced architecture for seizure detection. Method ACE-SeizNet integrates hierarchical feature extraction using convolutional layers, dimensionality reduction via pooling layers, and attention mechanisms for feature refinement. The model is trained and evaluated on EEG datasets using performance metrics such as accuracy, precision, recall, F1-score, Jaccard similarity, sensitivity, specificity, and AUC-ROC. Result ACE-SeizNet achieves an accuracy of 98.89%, sensitivity of 98.77%, specificity of 98.59%, and an F1-score of 99.67%, surpassing state-of-the-art models in seizure classification. The high AUC-ROC (0.999885 on the training set) validates its robustness in distinguishing seizure from non-seizure states. The model also demonstrates consistent performance across training, validation, and test datasets, confirming its reliability for clinical applications. Conclusion ACE-SeizNet provides an efficient and accurate solution for EEG-based seizure detection, offering superior performance over existing methods. Its ability to generalize across datasets and effectively detect seizures across multiple EEG channels makes it a promising tool for real-time clinical deployment. Future work will focus on further optimizing computational efficiency and enhancing interpretability for clinical usability.
A rapid development strategy was successfully implemented to create a suEPSR111459pramolecular metallogel comprising Mn(II) (referred to as MnA-TA) and Zn(II) (referred to as ZnA-TA) ions. These gels were formed using L-(+)-tartaric acid as a low molecular weight gelator in DMF medium at ambient conditions. Rheological analysis was employed to assess the mechanical stability of the synthesized MnA-TA and ZnA-TA metallogel. The results of the analysis revealed the gel’s impressive resilience when subjected to various angular frequencies and levels of oscillator stress. The exploration of MnA-TA and ZnA-TA metallogel’s distinct morphological patterns was conducted using FESEM images. FESEM analysis revealed that MnA-TA metallogel exhibited a flake-like hierarchical network structure, while ZnA-TA metallogel demonstrated a diamond-shaped architecture. EDX analysis was utilized for elemental mapping, confirming the presence of primary chemical constituents in the metallogels. The formation strategy and nature of the gel materials were examined through FT-IR spectroscopy and PXRD analysis. The synthesized metallogels exhibited semiconducting properties, as confirmed by optical band-gap measurements. Furthermore, a metal-semiconductor junction-based device was successfully fabricated by combining Al metal with Mn(II)- and Zn(II)-metallogels. The device displayed nonlinear charge transport behavior, resembling that of a Schottky diode, as evidenced by its I-V characteristic. This indicates the potential use of the sandwich-like configuration of ITO/MnA-TA metallogel/Al and ITO/ZnA-TA metallogel/Al in the development of advanced electronic devices based on supramolecular Mn(II)- and Zn(II)-metallogels. Notably, the direct utilization of tartaric acid and Mn(II)/Zn(II) sources in the MnA-TA and ZnA-TA metallogels presents an innovative approach, highlighting their suitability as semiconducting materials for device fabrication. This study delves into the multifunctional applications of MnA-TA and ZnA-TA metallogels, providing valuable insights for researchers in the field of material science. Graphical Abstract Derived from a low molecular weight gelator tartaric acid, supramolecular metallogels composed of Mn(II)- and Zn(II)-ions demonstrate remarkable stability at room temperature offer promising prospects for integration into electronic devices, specifically Schottky barrier diodes, operating effectively at room temperature.
In medical dialogue systems, recent advancements underscore the critical role of incorporating relevant medical knowledge to enhance performance. However, existing knowledge bases often lack completeness, posing a challenge in sourcing pertinent information. We present MedProm, a novel generative model tailored for medical dialogue generation to address this gap. Motivated by the need for comprehensive and contextually relevant responses, MedProm leverages state-of-the-art language models such as BioGPT. Our model is designed to integrate extensive medical knowledge into conversations, facilitating effective communication between patients and healthcare providers. At the core of MedProm lies the MediConnect Graph, a meticulously constructed knowledge graph capturing intricate relationships among medical entities extracted from dialogue contexts. By employing a KnowFusion encoder with a pretraining objective and masked multi-head self-attention, MedProm effectively processes the MediConnect graph, enabling precise control over information flow to capture its underlying structure. Furthermore, MedProm incorporates a sophisticated Curriculum Knowledge Decoder, leveraging transformer-based decoding to generate response utterances conditioned on input representations from the KnowFusion Encoder. The training process is guided through curriculum learning, gradually increasing optimization difficulty based on a coherence-based criterion. Experimental results on two datasets demonstrate the efficacy of MedProm in generating accurate and contextually relevant responses compared to state-of-the-art models.
Covalent organic frameworks (COFs) are crystalline porous materials bearing well-ordered two- or three-dimensional molecular tectons in their polymeric skeletal framework. COFs are structurally robust as well as physiochemically stable. Recently,...
NiCrBSi coatings have proven to be highly effective in delivering superior resistance to structures and components exposed to wear conditions. However, in harsh conditions, NiCrBSi faces high degradation, owing to its inferior mechanical properties. Further microstructural refinement and mechanical property enhancement of base NiCrBSi coatings can be achieved through secondary phase reinforcement. We deposited NiCrBSi composite coatings with 1wt% and 2wt% Nanodiamond (ND) using process optimization in Atmospheric Plasma Spraying (APS). The base and ND-reinforced NiCrBSi coatings were examined for microstructural, microhardness, and adhesion strength changes using SEM, XRD, ImageJ, and mechanical measurements. We have observed that ND reinforcing resulted in a smoother coating surface with reduced porosity (72%), increased microhardness (28.5%), and enhanced adhesion strength (65%) compared to the base NiCrBSi coatings. Hence, ND reinforcement to the plasma-sprayed NiCrBSi coating can be recommended as a potential method to further enhance the microstructural and mechanical properties of the plasma-sprayed NiCrBSi coatings.
Deep learning networks have been trained using first-order-based methods. These methods often converge more quickly when combined with an adaptive step size, but they tend to settle at suboptimal points, especially when learning occurs in a large output space. When first-order-based methods are used with a constant step size, they oscillate near the zero-gradient region, which leads to slow convergence. However, these issues are exacerbated under nonconvexity, which can significantly diminish the performance of first-order methods. In this work, we propose a novel Boltzmann Probability Weighted Sine with a Cosine distance-based Adaptive Gradient (BSCAGrad) method. The step size in this method is carefully designed to mitigate the issue of slow convergence. Furthermore, it facilitates escape from suboptimal points, enabling the optimization process to progress more efficiently toward local minima. This is achieved by combining a Boltzmann probability-weighted sine function and cosine distance to calculate the step size. The Boltzmann probability-weighted sine function acts when the gradient vanishes and the cooling parameter remains moderate, a condition typically observed near suboptimal points. Moreover, using the sine function on the exponential moving average of the weight parameters leverages geometric information from the data. The cosine distance prevents zero in the step size. Together, these components accelerate convergence, improve stability, and guide the algorithm toward a better optimal solution. A theoretical analysis of the convergence rate under both convexity and nonconvexity is provided to substantiate the findings. The experimental results from language modeling, object detection, machine translation, and image classification tasks on a real-world benchmark dataset, including CIFAR10, CIFAR100, PennTreeBank, PASCALVOC and WMT2014, demonstrate that the proposed step size outperforms traditional baseline methods. Graphical abstract
This study analyses trends in landslide publication in India, between 2010 and 2020, focusing on 79 studies, sourced from platforms like Google Scholar, ResearchGate, Web of Science, and ScienceDirect. The analysis reveals that approximately 65% of the publications were featured in Q1 and Q2 journals, with significant contributions from premium institutions like IIT-B (Mumbai), IIT-ISM (Dhanbad), and WIGH (Dehradun). The studies primarily focus on regions such as Uttarakhand, Himachal Pradesh, Tamil Nadu, and Kerala. Remote sensing and GIS emerged as the most frequently used approaches, appearing in 31 studies, followed by empirical methods (25 studies) and numerical techniques (20 studies). Researchers duly examined susceptibility, vulnerability, and risk using statistical techniques, with kinematic analysis and Slope Mass Rating (SMR) being notable empirical methods. The finite-element method was preferred for numerical slope failure analysis. The study found that 57% of the publications relied on existing data without conducting laboratory tests, while the remaining 43% conducted tests focusing on rock strength, shear strength, and physico-mechanical parameters. Direct shear tests and unconfined compressive strength (UCS) were the most commonly performed laboratory tests. Lithological analysis identified 20 different rock/soil types, with gneiss, quartzite, sandstone, and others frequently documented. Mitigation strategies such as informed decision-making, mechanical remediation, drainage system installation, and slope re-scaling were commonly suggested. Among the 31 computer-based programs used, ArcGIS, DIPS, and RS2 were the top tools for assessing mass-movement issues. The most common failure types reported were rock falls, debris flows, and rock slides, with rainfall identified as the primary landslide-triggering factor in 52 instances. The study highlights geology, geomorphology, and hydrology as key aggravating factors for landslide instances.
This paper explores integrating federated learning methodologies to optimize and generalize sensor design for plasmonic-based fiber optic sensors (FOS) applicable in biosensing, removing reliance on specific experimental datasets. By employing machine learning (ML) models, the enhancement of FOS design’s figure of merit (FOM) becomes achievable through training on localized data. However, to establish a more universally applicable ML model tailored to distinct applications, amalgamating data and training from diverse sources becomes imperative. FOS finds extensive utility in medical contexts where data privacy stands as a paramount concern, necessitating stringent consent and regulatory adherence for data sharing. Given the challenges posed by decentralized datasets and the criticality of data privacy, federated learning emerges as an indispensable framework, enabling the refinement of generalized ML models while upholding the sanctity of individual data privacy. Through the utilization of the Gaussian Process Regressor (GPR) model for localized training within discrete datasets, federated learning facilitates collaborative model refinement without compromising data privacy. This collaborative approach harnesses collective insights to bolster sensor performance, preserving data privacy boundaries and demonstrating potential enhancements without centralizing data aggregation. The research underscores the potential of federated learning in optimizing sensor design, accentuating its pivotal role in elevating sensing proficiency while safeguarding individual data privacy constraints, thereby paving the way for forthcoming practical implementations in biosensing applications.
Institution pages aggregate content on ResearchGate related to an institution. The members listed on this page have self-identified as being affiliated with this institution. Publications listed on this page were identified by our algorithms as relating to this institution. This page was not created or approved by the institution. If you represent an institution and have questions about these pages or wish to report inaccurate content, you can contact us here.
2,855 members
Md. Lokman Hakim Choudhury
  • Department of Chemistry
Sahid Hussain
  • Department of Chemistry
Jawar Singh
  • Department of Electrical Engineering
Pramod Kumar Tiwari
  • Department of Electrical Engineering
Information
Address
Patna, India
Head of institution
Prof. Pushpak Bhattacharya