Indian Institute of Science Bangalore
etd@IISc Electronic Theses and Dissertations at Indian Institute of ScienceNot a member yet
6204 research outputs found
Sort by
Opacity and its Trade-offs with Security in Linear Dynamical Systems
No full textOpacity is notion of privacy that is well-studied in computer science and discrete-event systems. In our work, we extend the opacity notion to linear dynamical systems. Opacity describes an eavesdropper’s inability to estimate a system’s “secret” states by observing the system’s outputs. We consider four opacity classes - initial-state, current-state, K-step and infinite-step opacity, and show that they are fundamentally connected with two subspaces of the linear system - the weakly unobservable subspace and the weakly unconstructible subspace. Further, we establish that a trade-off exists between opacity and security in the system. We show this in two ways – (i) we prove that an opaque system always permits undetectable attacks, (ii) we show that expanding the set of opaque states in the system always expands the set of undetectable attacks. We also propose optimization algorithms to minimally perturb a non-opaque system to make it opaque. We demonstrate our results on a smart grid system. Our work is the first to study opacity in such generality for linear dynamical systems, and provides necessary mathematical foundation for system designers to develop and build opaque systems, while ensuring adequate security
Correlations in multispecies asymmetric exclusion processes
No full textThe main aim of this thesis is to study the correlations in multispecies exclusion processes inspired by the research of Ayyer and Linusson (Trans. AMS., 2017) where they studied correlations in the multispecies TASEP on a ring with one particle of each species. The focus is on studying various models, such as multispecies TASEP on a continuous ring, multispecies PASEP on a ring, multispecies B-TASEP and multispecies TASEP on a ring with multiple copies of each particle. The primary goal is to investigate the two-point correlations of adjacent particles in these models. The details of these models are given below:
We study the multispecies TASEP on a continuous ring and prove a conjecture by Aas and Linusson (AIHPD, 2018) regarding the two-point correlations of adjacent particles. We use the theory of multiline queues developed by Ferrari and Martin (Ann. Probab., 2007) to interpret the conjecture in terms of the placements of numbers in triangular arrays. Additionally, we use projections to calculate correlations in the continuous multispecies TASEP using a distribution on these placements.
Next, we study the correlations of adjacent particles on the first two sites in the multispecies PASEP on a finite ring. To prove the results, we use the multiline process defined by Martin (Electron. J. Probab., 2020), which is a generalisation of the multiline process defined earlier by Ferrari and Martin.
We then study the multispecies B-TASEP with open boundaries. Aas, Ayyer, Linusson and Potka (J. Physics A, 2019) conjectured a formula for the correlations between adjacent particles on the last two sites in the multispecies B-TASEP. To approach this problem, we use a Markov chain that is a 3-species TASEP defined on the Weyl group of type B. This allows us to make conjectures and prove some results towards the above conjecture.
Finally, we study a more general multispecies TASEP with multiple particles for each species. We extend the results of Ayyer and Linusson to this case and prove formulas for two-point correlations and the TASEP speed process.Support in part by SERB grant CRG/2021/00159
Experimental and Numerical Studies on an Automobile Air Conditioning System with Refrigerants R134a, R1234yf and R1234ze(E)
No full textThe HydroFluoroCarbons (HFCs) synthesized as alternatives to ChloroFluoroCarbons (CFCs), though friendly to tratospheric ozone, have high Global Warming Potential (GWP). Despite this, numerous applications currently employ HFCs for refrigeration and air conditioning. The Kyoto protocol, negotiated in 1997 and came into force in 2005, put the HFCs in the green house basket and stated that the emissions of these gases need to be checked and controlled. The 2015 EU regulation and the 2016 Kigali amendment to the Montreal protocol suggested phase-out of the HFCs and this process will go on until 2036 in industrialized nations and until 2047 in non-industrial nations to accomplish a condition of 85% decrease of HFCs. The HFC134a refrigerant used in vehicle air conditioning has a 1300 Global Warming Potential (GWP), which prompted researchers to look for new low-GWP refrigerants. Recent research has revealed that the HydroFluoroOlefin (HFO) refrigerants HFO1234yf and HFO1234ze(E), with a GWP of 4 or less, show promise for
Application in Automobile Air Conditioning (AAC) field. The AAC requires special attention due to frequent leakages of HFC caused by pipe failures due to vibration. In this research, the low-GWP refrigerants R1234yf and R1234ze(E) are
used to explore AAC system performance, and comparisons with the currently used refrigerant HFC134a is made. An experimental setup is developed to simulate an AAC system containing evaporator, condenser, swash plate
reciprocating compressor and expansion valve (thermostatic type) as the main components with interconnecting copper pipelines and necessary controls and instrumentation. The setup also accommodates an Internal (liquid-to-suction)
Heat Exchanger as an optional component. The complete experimental setup was mounted on a mild steel frame in the
laboratory. The experiments were carried out to find out how various quantities of interest, such as the condenser heat rejection rate, Coefficient of Performance (COP), pressure ratio, cooling capacity & mass flow rate of the refrigerant, were affected by the evaporator face velocity, temperature at condenser inlet, compressor speed, temperature at evaporator inlet and air temperature, condenser face velocity. For all the three refrigerants the transport and thermodynamic properties are numerically generated using Helmholtz type equations of state or standard correlations and are validated against the Refprop 9.0 version, Consequently, the option to calculate the transport and thermodynamic properties using either customised algorithms or the Refprop software is provided. From the
different models, an integrated model for the entire system is created using the formulations for various components. The component and system models are validated against the results of published literature. For the 3 refrigerants
considered, the integrated model can simulate numerically the AAC system performance with and without the IHX. The results show that the higher evaporator inlet air temperature, higher compressor speed, higher condenser inlet air velocity, higher velocity of air at the inlet of evaporator and lower temperature of air at the inlet of condenser better
performance. The deviation in results between the numerical and experimental investigations are less than 8% for a system without IHX and less than 15% for a system with IHX. The performance of R134a is better than the alternatives
considered. The difference between R134a and R1234yf results are less than 15% without IHX and less than 10% with IHX. The difference between R134a and R1234ze(E) results are less than 33% without IHX and 25% with IHX. A
noticeable decrease in both the power required for compression and the cooling capacity is observed in case of R1234ze(E). This indicates that in order increase the refrigerating capacity of R1234ze(E) a compressor with enhanced volumetric displacement should be used. According to the results of the current study, use of IHX is advantageous
for both R1234ze(E) & R1234yf, and R1234yf performs better than R1234ze (E). The COP of R1234yf with and without IHX is on an average 9% and 5% lower respectively compared to R134a without IHX, and the COP of R1234ze(E) with
and without IHX is on an average 5% and 3% lower respectively compared to R134a without IHX. The cooling capacity of R1234yf is on an average 8% less than the cooling capacity of R134a without IHX. When an IHX is interposed in
the circuit, this difference is reduced to an average of 4%. The cooling capacity of R1234ze(E) is on an average 28% less than the cooling capacity of R134a without IHX. This difference is reduced to an average of 23% with the use of IHX
for R1234ze(E). Even though R134a performed better, R1234yf with IHX is a better alternative in the current AAC system working with R134a without IHX, with only a slight compromise in the system's performance. Thus, if the AAC systems change to R1234yf with an IHX, the directives set out in the Kigali amendment of 2016 to Montreal Protocol (namely the discontinuation of HFCs for refrigeration) will be satisfied without any significant loss in the performance
On the creep behaviour of Ni based solid solution alloys from binary to quaternary
No full textOn the creep behavior of Ni based solid solution alloys from binary to quaternary
A power-law relationship between the steady state strain rate (̇) and imposed stress (σ) is well established so that the ̇∝^, at a fixed temperature, where n is termed the stress exponent. Additional microstructural terms influencing creep such as the grain size and stacking fault energy (γ) have been incorporated into the creep equation in a power-law form, such as ̇∝^. In the case of an intragranular dislocation climb controlled creep in a solid solution alloy, the creep rate has been expressed as ̇∝^5^3, where D is the appropriate diffusion coefficient, n~5 and q~3. However, an evaluation of the earlier creep data suggests that while q~3 is reasonable for pure metals, there is considerable uncertainty in the value of q for solid solution alloys.
The current study focuses on characterizing the creep behavior and the value of q for Ni – xCo alloys, with x = 10, 33, and 60, where the addition of Co reduces the stacking fault energy. Following creep in Ni-Co alloys, the creep behaviour of CSSAs NiCoCr and NiCoCrFe was also investigated to probe potential changes in the creep mechanisms with the addition of alloying element. All alloys are single phase solid solutions with a face-centered cubic (f.c.c.) crystal structure. Prior to the creep, all alloys had a grain size d~100 μm. The observed stacking fault energies through weak beam dark field technique are 10 – 36 mJ m^-2 for Ni – 60 Co, 14 – 27 mJ m^-2 for NiCoCr, and 11 – 26 mJ m^-2 for NiCoCrFe alloys. At a creep testing temperature of 1015 K, the stress exponent n ~ 5 for binary Ni – Co alloys, suggesting that dislocation climb is the creep rate controlling mechanism. The cell structure observed through electron channel contrast imaging in the crept samples of the alloys is consistent with a dislocation climb mechanism. The stacking fault energy exponent q ~ 2 in Ni– (x) Co binary solid solution alloys.
At 990 K, the creep rates of the NiCoCr and NiCoCrFe alloys were observed to be similar with a stress exponent n~5. Both the ternary and quaternary alloys showed significantly lower creep rates compared to the Ni – (x) Co binary alloys. Creep deformation did not cause any phase change in the NiCoCrFe alloy. Although the Ni – (x) Co alloys, NiCoCr, and NiCoCrFe alloys exhibited a similar stress exponent of n~5, the crept substructure in the NiCoCr and NiCoCrFe alloy showed planar band features. The possible causes for the observed differences in creep behavior between the Ni-Co binary alloys and the other concentrated solid solution alloys will be discussed
3D-Printing of Lunar Soil Simulant by Direct-Extrusion method.
No full textThe extrusion-based additive manufacturing (EAM) technique is recently being widely employed for the 3D printing of complex-shaped components made of ceramic powder (containing irregular-shaped particles) when it is cast in the form of a slurry/ink. In this work, we utilize a direct extrusion method for printing structures from extra-terrestrial soil simulants using a piston-based extruder. Printing is demonstrated using a slurry composed of lunar soil simulant (LSS) variant ISAC-1 (avg. particle size ~ 90µm) mixed with biopolymer guar gum as a sustainable binding agent and DI water as a solvent. Parts were printed using a 2 mm diameter nozzle by optimizing print speed, nozzle height, inter-layer drying time, and build temperature, to ensure shape retention post-printing. The final green parts were dried in a hot air oven (50°C) for 48hrs, followed by sandpaper polishing. The strengths of the printed specimens were evaluated using compression and flexure tests and were found to be comparable to that of bio-consolidated structures. Unlike solid geometries, the well-known shell-infill type area-filling strategy generated several travels and re-tracings in the toolpath for cellular geometries. Owing to the yield stress of slurry, the travels and re-tracings resulted in discontinuous print and poor dimensional accuracy respectively. This necessitated a toolpath with increased continuity in the extrusion path. The customized toolpath is generated by defining a continuous nodal path over a lattice structure corresponding to the cellular frame. The extrusion flow rate is tuned according to the nodal path and the requirement of material deposition. Qualitatively the increased extrusion continuity in the customized toolpath resulted in continuous print with improved dimensional accuracy, whereas quantitatively a significant (~ 60%) reduction in print time is observed. These results show the potential for using the direct extrusion 3D printing method in remote extra-terrestrial environments to obtain lightweight load-bearing structures like cellular frames
Constraining the Lithium Seawater Mass and Isotope Budget: Diagenetic Processes Through Marine Pore Waters
No full textSilicate weathering consumes CO2 and controls cation fluxes to the ocean, thus playing a critical role in modulating long-term seawater chemistry and climate. There are very few markers of seawater chemistry whose value has changed over time as a function of the uplift – weathering – subduction cycle. Lithium, being one such proxy has been extensively utilized as a geochemical tracer whose long-term evolution in seawater is a function of Urey’s tectonic cycle. Marine pore-waters are an excellent archive to study the sedimentary processes affecting seawater chemistry. Thus, marine pore water Li concentration and isotopes are utilized in elucidating numerous under-constrained diagenetic processes such as marine clay formation, clay transformation, carbonate diagenesis, subduction of slab, and clay dewatering.
To constrain the above processes through Li isotopes, we have developed a method for separation of Li from matrix elements using a single-step column chromatography technique and precise measurement of Li isotope ratios using inhouse ICP-QQQ. Some of the features of the developed method are high column recovery 101.0 ± 1.2 % (n = 20), low cumulative Li blank (<0.6 pg) and crustal element blanks (<1.5 ng), high Na tolerance (up to 100:1 of Na:Li), low mass requirement (<0.15 ng per analysis), and a sub-permil precision (±0.6 ‰, 2s). Utilizing the above method, we analysed pore water samples from IODP Leg 339 (Mediterranean outflow) and Leg 379 (Amundsen Sea, Southern Ocean) for Li isotopes to deduce the processes occurring at these sites and its implications on Li seawater budget.
At the pore water-sediment interface of continental margins, authigenic alumino-silicate clay (smectite) formation also termed as Reverse Weathering, removes Li from seawater/pore water. The reverse weathering process preferentially uptakes 6Li over 7Li. Thus, pore waters are depleted in Li compared to seawater ([Li]PW δ7LiSW). This process occurs in shallow sediment depths where smectite is the dominant clay and illitization is not commenced. However, most pore water profiles exhibit higher Li concentrations and lighter isotopic compositions indicating clay transformation process. During clay transformation i.e., smectite to illite transformation, owing at high pressure and temperature (150-200 Celsius) during burial, isotopically light Li is released from the clays. This release of isotopically light Li increases the pore water Li concentration while driving it isotopically light. During sediment subduction, a significant fraction of this clay bound isotopically light Li is released as a part of clay dewatering. This thesis investigates IODP Leg 339 pore water samples with clay dewatering evidence (recorded by δ18O of pore waters) and IODP Leg 379 pore water samples in silica-rich environment with high Li concentration and lighter isotopic composition relative to seawater helps us to constrain Li seawater mass budget further.
A preliminary set of equations that govern flux of an element between the marine sediments and seawater following the general diagenetic equation (GDE) is also developed in the present work. These equations incorporate the effects of diffusion, advection and reaction kinetics in the sediments and thus, govern the transfer of Li within the sediment column. Flux calculations for 5 IODP sites and 12 from the literature have been included in the work, and the implications of these flux calculations have been discussed in detail. A thorough development of this model will lead to establishing a mass and isotope budget in seawater which will be applicable across elements and processes. This fundamental study of Li seawater chemistry brings us a step closer in understanding the complex dynamics of ocean systems
Inverse Material Design using Deep Convolutional Neural Networks and Ontology
No full textMaterials design is the process of tailoring materials with specific properties and performance characteristics to meet the particular needs of an application. It involves going back on the process-structure-property (PSP) relation from properties to manufacturing processes. Consequently, much research has dived into PSP relation modeling, and recently, data-driven methodologies for modeling PSP relations are becoming a way of research in the field. However, most of the efforts investigate only the structure-property relation, leaving the other part relatively unexplored. Moreover, they tend to focus only on specific material systems and lack the consideration for microstructural metadata (The peripheral and metallurgical data related to the microstructure such as material composition, magnification, primary microconstituents and their phase volumes, grain size and morphology, etc.); These limitations are perhaps in part due to the absence of a standard format or protocol for collecting, storing, and sharing the PSP relationship datasets. Also, much of the knowledge in the field of materials design and materials science is primarily empirical and experiential. Consequently, capturing this prior knowledge in a structured, human-readable, and computable format is necessary. The handling of the sequential data regarding the manufacturing processes undergone by materials is yet another open area of research.
In this work, we develop a scalable convolutional neural network (CNN) based approach for modeling the inverse process-structure relation using Ultrahigh carbon steel and Ni-Co alloy materials datasets. Here, the scalability of the models was achieved by implementing multi-input multi-output CNN models on various learnable relations. Learnable relations included a variety of direct approaches and a feature augmentation-based approach. The assortment of models concluded that the direct approach models having microstructure and its metadata both as input produced the best results for predicting the process parameters. Thus, corroborating the intuitive importance of microstructure metadata. In addition, typical problems with the PSP relationship datasets were identified, and potential solutions such as data and test-time augmentation techniques and an ontology-based standardization were suggested and evaluated.
To implement the above-mentioned solution and capture some of the essential knowledge in the materials design field in a structured and computable format, an application ontology called PSPOnto was developed. The devised ontology could readily handle data regarding the complete manufacturing history of a material, along with its microstructure and properties, in a generic way. Provisions for capturing the details of microstructure preparation methods and material property measurement procedures were also made. PSPOnto was designed to preserve and exploit the capabilities of an ontology in handling the many-to-many relations among the entities stored in it. Consequently, the novel structure of the ontology enabled a conventional SPARQL query engine to search for the materials with specified property requirements and the associated manufacturing process sequences without any ab-initio manufacturing process sequences being input into the knowledge base. Thus, PSPOnto is regarded as laying the foundation for the knowledge-based establishment of PSP relations and materials design. The proposed ontology was then validated for its efficacy by creating a knowledge base on it using a Ni-Co alloy dataset and querying it using the SPARQL query language. The exhaustive enumeration of the manufacturing processes, corresponding process parameters, and the compilation of reference articles used in its creation serve as an independent resource for other researchers working on PSP relations and accelerated materials design
Homogenization of PDEs on oscillating boundary domains with L1 data and optimal control problems
No full textThis thesis comprehensively studies the homogenization of partial differential equations (PDEs) and optimal control problems with oscillating coefficients in oscillating domains. After the introduction, the thesis is divided into two parts. In part I, we consider problems in oscillating circular domains with three chapters(chapters 1-3), whereas in part II(chapters 4-5), we deal with domains having oscillations in low dimensions.
The first chapter investigates the homogenization of a second-order elliptic PDE with oscillating coefficients in a circular oscillating boundary domain. By using the polar form of the differential equations and a polar unfolding operator, we consider the general type of oscillations and study the asymptotic behavior of the renormalized solution of the PDE with a source term in .
The second chapter examines the homogenization of an elliptic variational form with oscillating coefficients in a circular domain that is highly oscillating itself. The source term is in , and we take into account the non-uniform ellipticity that arises due to the highly oscillating boundary, rapidly oscillating coefficient, and the oscillating part made up of highly contrasting materials.
The third chapter focuses on the homogenization of optimal control problems governed by second-order semi-linear elliptic PDEs with matrix coefficients in a circular domain. The cost functionals considered are of general energy type and may have different oscillating matrix coefficients than the constrained PDEs. We prove the existence of well-defined limit problems and derive explicit expressions for the limiting coefficient matrices.
The fourth chapter extends the study to an -dimensional domain with an oscillating boundary that oscillates in directions, where . This is a relatively unexplored area in the literature, and we show that in this case, the limit problem has derivatives in all non-oscillating directions, or . Specifically, we study the homogenization of an elliptic PDE in such a domain with data. Our work expands upon previous research and could have potential applications in various fields.
The fifth chapter investigates the homogenization of optimal control problems governed by second-order semi-linear elliptic PDEs with matrix coefficients in a low-dimensional oscillating domain. The cost functionals considered are of general energy type and may have different oscillating matrix coefficients than the constrained PDEs. We prove the existence of well-defined limit problems and derive explicit expressions for the limiting coefficient matrices. We show that, in this case, the limit problem has derivatives in all non-oscillating directions. Our results show that the limiting cost functional's coefficient matrix is a combination of the original cost functional's and constrained PDE's coefficient matrices
Assessing protein contribution to phenotypic change using short, coarse grained molecular dynamics simulations
No full textUnderstanding the functional mapping between genotype and phenotype is an important problem that has ramifications for various diseases. Various existing computational methods can infer these disease-related functional mappings. Molecular dynamics (MD) is one such advantageous method that does not rely on prior information or learning, as they use the first principles (Newton's laws of motion) to determine protein movement. Thus, they are suited for understanding and rationally evaluating phenotype alteration with minimal bias. However, MD simulations are computationally expensive and require a lot of resources and time. Therefore, using lengthy all-atom MD simulations to reproduce microsecond to millisecond scale biological phenomena is prohibitive. A previous study assessing phenotype alteration recorded the structure's root-mean-square fluctuation (RMSF) from a coarse-grained MD simulation of 1 microsecond. Our study uses a short coarse-grained MD simulation (<10 nanoseconds) to generate the RMSF data in combination with a new scoring function for prediction. The designed scoring function captures the changes in the RMSF between the wild type and the variant, normalized for comparison.
The shortened simulation time allows us to evaluate more variants in a reasonable time. We predicted phenotype change scores for 14,691 variants of Calmodulin, SUMO-conjugating enzyme UBC9 (UBE2I), Small ubiquitin-related modifier 1 (SUMO1), and Methylenetetrahydrofolate reductase (MTHFR) catalytic and regulatory domains, for which quantitative experimental data as a variant phenotype score was available. We found a high Pearson correlation coefficient when calculating the values at various minor levels of outlier exclusion. We obtained a consistently superior performance for all proteins except for the catalytic domain of MTHFR when compared against the state-of-the-art machine learning-based method Polyphen2. The performance of the catalytic domain of MTHFR was comparable to that of Polyphen2. We analyzed our results across all proteins to understand why the prediction erred on a subset of variants. We believe that the insights gained from this work will help strengthen the rational interpretation of single nucleotide polymorphism of the genome in the context of observed phenotypic change
Fragile Interpretations and Interpretable models in NLP
No full textDeploying deep learning models in critical areas where the cost of making a wrong decision
leads to a substantial financial loss, like in the banking domain, or even loss of life, like in the
medical field, is significantly less. We cannot entirely rely on deep learning models as they
act as black boxes for us. This problem can be resolved by Explainable AI, which aims to
explain these black boxes. There are two approaches to explaining these black boxes, one being
via posthoc explainability techniques and the other being by designing inherently interpretable
models. These two are the basis of our work.
In the first part, we talk about the instability of posthoc explanations, leading to fragile in-
terpretations. This work focuses on the robustness of NLP models along with the robustness
of interpretations. We have proposed an algorithm that perturbs the input text such that the
generated text is semantically, conceptually, and grammatically similar to the input text, yet
the interpretations produced are fragile. Through our experiments, we have shown how the
interpretations of two very similar sentences vary significantly. We have shown that posthoc
explanations can be unstable, inconsistent, unfaithful, and fragile; and, therefore, cannot be
trusted. Finally, we have concluded whether to trust the robust NLP models or the posthoc
explanations.
In the second part, we have designed two inherently interpretable models, one for offensive lan-
guage detection tasks in the case of multi-task learning for three subtasks, sharing a hierarchical
relationship between them and the other for the question pair similarity task. Our offensive
language detection model achieved an F1 score of 0.78 on the OLID dataset and 0.85 on the
SOLID dataset. Our question pair similarity model achieved an F1 score of 0.83. We also
provide a detailed analysis of the model interpretability as well as prediction interpretability