Lakshminath Bezbaroa Central Library Digital Repository
Welcome to the Institutional Digital Repository of Lakshminath Bezbaroa Central Library.
- This digital archive comprised of the Institutes' intellectual output.
- It manages, preserves & makes available the academic works of faculty and research scholars.
- It is established to facilitate deposit of digital content of scholarly or heritage nature.
- Allowing academics & their departments to share & preserve contents in a managed environment.
Communities in LBCL Digital Repository
Select a community to browse its collections.
Recent Submissions
Isolation Forest Based Efficient Unsupervised Machine Learning Algorithms
(2025) Ahlawat, Nidhi
Many ML algorithms have common redundancies that make them impractical for large datasets. The overarching goal of this thesis is to prune the redundant computations with minimal loss in the quality of the downstream tasks. This dissertation focuses on three unsupervised machine-learning tasks: clustering, anomaly detection, and model update. We utilize the Isolation Forest data structure as a tool to improve efficiency for all three tasks. This data structure was initially developed to perform anomaly detection task in an unsupervised manner. Specifically, we focus on the following three scenarios: 1. When an application needs all-pair distances: How to compute all-pair distances faster by optimizing the order of distance computation? 2. When an application needs only a subset of all-pair distances: How do we quickly identify the required subset of all pairs? 3. When new data causes concept drift: How to update the model quickly? For the first scenario, we develop an algorithm: fast MBD (fMBD) that computes all-pair distances with up to 5X speed-up. Our fMBD algorithm has no approximation or heuristic, and it computes the exact distance for each data point pair. We demonstrate the effectiveness of the fMBD algorithm with clustering and anomaly detection applications. For the second scenario, we develop a scalable MBScan (sMBScan) clustering algorithm that selectively computes distances between data point pairs. Our algorithm achieves up to 53X speed up with up to 96% reduction in the memory footprint and no loss in the clustering quality. For the third scenario, we develop an Incremental Isolation Forest (I2Forest) that quickly updates the Isolation Forest data structure in response to the arrival of new data. I2Forest is particularly effective when the new data causes concept drift. I2Forest has significantly lower training time than retraining the model from scratch. I2Forest also performs better than other incremental approaches for model update.
Thermoeconomic optimization studies of the refrigeration systems using metaheuristic techniques
(2023) Nagraj, Makkitaya Swarna
Refrigeration systems play a crucial role in maintaining thermal comfort by removing excess heat from enclosed spaces. These systems are vital for various applications, from cooling buildings to preserving perishable goods. The primary goal is to provide a cooling load to the evaporator through chilled water to maintain a specific temperature within the enclosed space. Among the various refrigeration systems, the vapor compression refrigeration system (VCRS) stands as the most widespread choice due to its high coefficient of performance (COP). However, VCRS consumes a significant amount of electricity because of its compressor, contributing indirectly to substantial CO2 emissions. To address the issues of energy consumption and CO2 emissions, researchers have turned their attention to refrigeration systems driven by alternative low-grade energy sources, such as vapor absorption refrigeration systems (VARS) and vapor adsorption refrigeration systems. Unfortunately, these sorption technologies tend to suffer from lower efficiency compared to conventional VCRS.
Development of DNA Aptamers against selected Celiac disease epitopes in Glutenin
(2023) Kalita, Jon Jyoti
Celiac disease, an autoimmune enteropathy of small intestine is caused by the indigestible immunostimulatory peptide sequences present in gluten protein. Gluten is present in the cereals like wheat, rye and barley. Consumption of gluten free diet is the only option available for celiac disease patient. The immunostimulatory peptide sequences are present in both the gliadin and glutenin fractions of gluten. Prior research were emphasized on detection of gliadin protein mostly, by different methods including antibodies and aptamers. This thesis focuses on the in-vitro selection and characterization of aptamers against two 14 mer peptide sequences, one from the high molecular weight glutenin of wheat (Triticum aestivum), GQGQQGYYPTSPQQ (GQ-14), which contains a celiac disease epitope QGYYPTSPQ and the other from low molecular weight glutenin, SQQQPPFSQQQQPV (SV-14), which contains epitope PFSQQQQPV. The aptamer sequences apt_J91P (56 bp) and apt_M09P (76 bp) were selected through conventional SELEX method against peptide targets GQ-14 and SV-14 respectively. The binding characterization between apt_J91P and GQ-14 by ITC reveals the dissociation constant of 2.26 μM and 4.385 mM for the primary and the secondary site of binding respectively. The dissociation constants are 17.6 μM and 8.33 mM for the primary and the secondary site of binding between apt_M09P and SV-14 respectively. The binding was also characterized by circular dichroism spectroscopy. The limit of detections of the aptamers evaluated by direct-ELAA method are found to be 16.0875 μM for apt_J91P and 20.00 μM for apt_M09P against its respective targets. The preliminary investigations carried out to check the applicability of aptamers apt_J91P and apt_M09P in development of bioassays, gold nanoparticle based aptamer assay and aptamer mediated magnetic bead based extraction assay demonstrated target binding and specificity. The future scope include improvement the binding affinity of the aptamers for detection of the target peptides in real wheat samples.
Studies on Oil-Water Separation using Functionalized Magnetic Nanoparticle Dispersed Polymeric Membranes
(2023) Boruah, Pankaj
The motivation of the current research work is to develop a suitable method for the separation of oil from oily wastewater. A bio dispersant and a polymeric membrane were developed for the treatment of oily wastewater. The dispersant was developed by using a biopolymer as a bio-dispersant to stabilize crude oil. The polymeric membrane was developed using polyvinylidene fluoride (PVDF) and a lab-developed bioderived nanofiller, cellulose nanocrystal (CNC). Oligo lactic acid conjugate chitosan (OCH) were used as a bio-dispersant, an alternative to the synthetic chemical dispersant, which reduces the toxicity associated with chemical dispersants used in oil spill remediation. Results from the research indicate that byapplying Xanthan gum (XG) and OCH, a considerably stable crude oil-in-water emulsion was obtained, which remains stable for more than 3 weeks, as confirmed by the optical microscopic image. XG was used to increase the viscosity of water, which helps in further stabilizing the oil droplets. Degradation of these stabilized emulsion droplets was also carried out by isolated bacteria Pseudomonas aeruginosa CoE-SusPol3. The degradation percentage obtained from gas chromatography (G.C.) analysis for the stabilized emulsion is 74.31%, whereas the unstabilized emulsion is 34.06%. The in-situ production of biosurfactants from the isolated bacteria also assisted the degradation process by decreasing the surface tension of the crude oil. This research work addressed in this thesis also discussed the use of CNC as a bionanofiller, an alternative to the inorganic filler in the membrane. CNC was modified to form magnetic CNC (FeCNC) by single-step co-precipitation method, and a thin film membrane was prepared by adding FeCNC into the PVDF matrix. Detailed characterization for morphology, wettability, porosity and performance analysis was carried out for the composite membranes. For comparison, PVDF membranes with CNC were also prepared, and its detailed characterization and performance analysis were carried out. Further, the application of a magnetic field during membrane formation gives a new dimension to understanding the magnetoresponsive behaviour of FeCNCs for orientation at the surface of the membrane. More prominent finger-like structures were obtained for the PVDF/FeCNC membrane. This migration of FeCNCs to the surface improves the hydrophilic property of the membrane and thus results in superior antifouling properties. This current research helps in understanding the application of biomaterials for treating oily wastewater and how a bio-based material could be a potential and sustainable solution for treating oil spills instead of using harmful synthetic chemical dispersants. The reported findings also give an idea for the improvement of properties of the membrane by application of surface modification of the CNCs and how CNCs can be oriented at the surface of the membrane to improve the performance.
Heat Transfer Analysis of a Fin-tube Heat Exchanger using Shearthinning Fluid and Winglet Type Vortex Generators
(2024) Kumar, Dheeraj
The present thesis work investigated the effect of various shapes of winglets, such as delta winglet pairs (DWP), rectangular winglet pairs (RWP), and curved rectangular winglet pairs (CRWP), on the flow behavior of an aqueous solution of carboxymethylcellulose (CMC) through a rectangular channel with and without the inbuilt cylinders. Threedimensional numerical simulations were carried out for a range of Reynolds numbers (50-200) using aqueous carboxymethylcellulose, commonly known as non-Newtonian shear-thinning fluid as a working fluid. Nusselt number (Nu), friction factor (f), and the combined effect of Nu and f, commonly known as quality factor (Qf), were calculated and compared for each case. Additionally, the comparison is also made with the base channel considering a dimensionless number (η) defined as (Num/Nuo)*(fo/f)0.33. As well, an investigation was conducted into the impact of attack angle, winglet placement, and tube placement on the thermohydraulic performance of the channel. The fintube heat exchanger (FTHE) performance parameters were effectively compared and evaluated using an aqueous
solution of CMC and water. It was discovered that incorporating DWP in FTHE improves overall performance compared to RWP. Still, when only heat transfer applications are considered, RWP outperforms DWP. When comparing the thermohydraulic performance, the use of CRWP is unequivocally superior to RWP. The former option provides a more efficient and effective means of achieving the desired outcomes. Reducing the attack angle improves performance up to a limit, but further decrease reduces performance. Utilizing an aqueous solution of CMC as a working fluid has been established as a practical approach to augment heat transfer while simultaneously reducing pressure drop. Based on the thorough evaluation of the overall performance, it is evident that FTHE with winglets significantly outperforms the base channel. Staggered tube configurations of FTHE show better thermal performance than inline tube configurations. They generate higher values of critical parameters such as Nu, f, and Qf, which enhances the efficiency and effectiveness of relevant applications.