Dhivyaraja was awarded Prime Minister’s fellowship to pursue doctoral study on “Advanced micro spray cooling technologies for high power density hydraulics “in April 2015. He has worked on developing the miniaturization of pressure swirl atomizer using MicroElectroMechanicalSystem (MEMS) technology with an active collaboration with Prof. Srinivas Tadigadapa, Penn State University, USA. His research interest is on understanding the liquid atomization, spray formation in the micro injector and the dynamics involved in mutual intersection of multiple micro scale sprays, which has a potential application in electronic spray cooling and tablet coating.

He experimentally examined the characteristics of intersecting sprays using three types of atomizer configurations (i) single spray (ii) double spray and (iii) triple spray. His study demonstrates that the dynamics of wave propagation on the liquid sheets in intersecting sprays is markedly different from the non-intersecting spray. He has also experimentally investigated the prospect of dynamical similarity and universality of drop size and velocity spectra in sprays by fabricating a cohort of geometrically similar pressure swirl atomizers. His experimental study includes the non-intrusive spray diagnostic techniques such as (i) High Speed Imaging (HSI), (ii) Phase Doppler Interferometry (PDI) and (iii) Optical Spray Patternation.

K.S. Siddharth has worked on Feature Correlation Velocimetry for Measuring Instantaneous Liquid Sheet Velocity.When the liquid sheet interacts with ambient air,short wavelength corrugations are naturally formed on its surface. Feature correlation velocimetry (FCV), is the method to extract velocity information directly from successive high speed images of a liquid sheet by relying on the advection of the short wavelength corrugations.

With regard to GCSC injectors used in rocket propulsion applications, Siddharth has examined the effect of gas swirl on the performance of a gas-centered swirl co-axial injector. Introduction of swirl in the gas jet is likely to improve atomization. Also, the effect of a high pressure environment on the near injector spray structure of a Gas-centered swirl co-axial (GCSC) injector was studied. Ambient pressure was shown to have a strong influence on the morphology of the annular sheet.

Near nozzle structure of annular sheet exiting a GCSC injector. The highlighted box represents the area of interest where method of FCV is applied to retrieve velocity information.

Cross flow jet is a configuration where a slow moving liquid jet interacts with fast moving air stream at a certain angle (usually perpendicular). The shearing action caused due to their interaction is responsible for causing fluctuations in the liquid stream and ultimately breaks up the bulk liquid into droplets. Cross flow configuration of jets and sprays finds application in natural an industrial processes such as fuel injection in gas turbine combustors where it is necessary to determine the velocity and size of the particle formed from shearing of jet since inhomogeneous velocity and size distribution of the droplets may reduce the performance of the combustor by creating local hot spots. For coating purpose also inhomogeneous velocity and size distribution will cause uneven coating on the material. High speed imaging technique is used to study the modes of break up and Phase Doppler Particle Analyser is used to estimate drop size and velocity of the droplets formed by breakup of the jet.

Harinadha Gidituri has worked on problems in the area of soft and condensed matter physics using particle based mesoscopic method Dissipative Particle Dynamics (DPD). One of them is the collective dynamics in active and passive binary mixtures. Three different states were observed namely, Meso-Turbulent state (disordered), Polar Flock state (ordered), Vortical state (ordered) for different values of Peclet number (Pe--Ratio of self propulsion to thermal noise) and area fraction of the active particles. The other is the phase separation dynamics of the binary fluids in periodic and confined domains. Growth dynamics in confined geometries were observed to be slow in comparison with periodic domains. Furthermore, under external imposed shear the growth dynamics are suppressed.

Trajectories of drops executing a random walk. Different colors indicate different trials.

Trajectories of drops executing a Levy walk

Temporal evolution of a pure fluid is compared against the morphological evolution of binary phase mixtures under deep and shallow quenches (a) Pure fluid (b) shallow quench and (c) deep quench

Dr. Sathish Akella’s primary interest is soft condensed matter physics which deals with the classical (physics) understanding of liquid crystals, surfactants, active matter, glasses, gels etc. Currently he is working on two different problems in active matter

1. Self-propelling drops

Octanoic Acid (OA) drops on water surface are spontaneously set in motion by the surface tension forces. These forces are known as Marangoni forces. While in motion these drop deform because of the interplay between the inertial, viscous and surface tension forces. However, when inertial and viscous forces dominate the surface tension force, the drop motion exhibits both ballistic and diffusive transport and is characterized by the non-linear scaling of the mean square displacement with time. The non-linear scaling has also been observed in physics, astronomy and biology. This type of transport is classified as Lévy walk named after the French mathematician Paul Pierre Lévy (1886 – 1971). The resultant transport in a Lévy walk is a superposition of diffusive displacements interspersed with sporadic ballistic displacements. In future, I would like to experimentally study and understand the physics of a collection of these Lévy walkers.

Trajectories of drops executing a random walk.Different colors indicate different trials.

2. Bacterial suspensions

The emerging physics when a collection of bacteria swim together is completely different from a single bacterium. In a bacterial suspension, each bacterium is couple to the rest via the hydrodynamic interaction which results in organized motion. Sometimes harvesting and preserving bacteria is a difficult task. Luckily, the physics associated with these type of problems is amenable to theoretical and numerical techniques. I study these systems using a simulation common simulation technique known as Dissipative Particle Dynamics (DPD).

The following figures show the three different phases observed for different combinations of packing fractions and at different Peclet numbers (Pe). The Peclet number is defined, only for active particles, as the ratio of activity to thermal noise.

Fig a: Random/Meso-turbulent phase (10% active and 90% passive particles and Pe = 0.11)
Fig1b: Vortical phase (10% active and 90% passive particles and Pe = 1.11)
Fig1c: Polar flock (50% active and 50% passive particles and Pe = 1.11)

Karthiga Devi is working on Aerosol Deposition Studies In Human Lung towards Personalized Medicine.
Inter-subject variability in drug deposition is one of the causes of ineffectiveness of a prescribed dose of medicine. Personalized medicine is emerging as a solution to treat some of the respiratory diseases. Lung morphometry has a great influence on the transport and deposition of therapeutic aerosols as well as disease causing micro-organisms. Her study aims to establish the relation between aerosol deposition in the alveolar region and parameters determining the lung morphometry. The first part of the study employs a computational approach to study the possibility of engineering an aerosol size distribution to minimize the effect of inter-subject variability.. For the parametric conditions investigated, model predictions also suggest that particles sizes near 0.01µm and 10µm will result in the least inter subject variation in alveolar deposition.
Determination of lung dimensions of an individual leads to designing the treatment tailored for the specific individual. Thus in the second part of the study, a gamma scintigraphy scan is employed to find the regional deposition and the C/P ratio in healthy subjects. The regional deposition and the C/P ratios are used as input the numerical lung model. The error between the experimental values and the numerical values are minimized following a minimization procedure. Thus the experimental values and the numerical predictions match generating the set of morphometric parameters for the given individual. From this proposed approach, the dimensions of lung were predicted. The same study could be used in patients with respiratory illness, for finding their lung morphometry. This would be indeed a great boon for the medical practioners, where personalized treatment and delivery of drug to a targeted generation in the lung is indispensable.

Nachiketa Janardan worked on the behaviour of sessile drops on an inclined substrate.Three-dimensional sessile drops are ubiquitous in nature. The behaviour of such drops on an inclined, real hysteretic surface has been studied experimentally and theoretically. Surface Evolver has been used to model the shape of a drop. From this study, two critical inclination angles were identified that described the incipient motion of the drop. It was observed that these angles depended on the initial contact angle distribution along the contact line of the drop. It was demonstrated that given the initial conditions that a drop is subjected to, it was possible to predict its behavior when the substrate is inclined. Additionally, a new statistical approach to determining Contact Angle Hysteresis has been developed. The complete three dimensional shape of the sessile drop was generated with only the contact line and the volume as inputs. The distribution of the local contact angles for such drops was generated. A Generalized Extreme Value analysis was performed and the ranges of advancing and receding angles estimated. This is used to yield an approximation of the Contact Angle Hysteresis associated with the liquid-substrate combination.

Using a dynamic variant of the Vicsek model, he showed emergence of a crush from an orderly moving human crowd is a non-equilibrium first order phase transition. It was also reported that this transition can be reversed by modifying the dynamics of a few people, deemed as game changers.Surprisingly, the optimal placement of these game changers is found to be in regions of maximum local crowd speed. The presence of such game changers is effective owing to the discontinuous nature of the underlying phase transition. Thus this generic approach provides (i) strategies to delay crush formation and (ii) paths to recover from a crush, two aspects that are of paramount importance in maintaining safety of mass gatherings of people.

The dynamics of liquid thin films is relevant in many applications such as stability of thin films. Combined effect of spreading and dewetting on liquid substrate helps to understand these processes in detail. This combined effect has been studied through simple spreading experiments. A number of different natural oils - coconut oil, olive oil, gingelly oil and castor oil were used during this study. A drop of the natural oil was placed on the water free surface. Initially, the drop is observed to spontaneously evolve into a thin liquid film through spreading. Because of this spreading process there is sheet thinning. At a critical thickness of around 10 micro meters there is localized sheet thinning on the various places of the spreading film. This local sheet thinning leads to hole nucleation. Then this nucleated hole grows in time due to dewetting. Finally, the liquid bridges separating individual holes were observed to ruptures, thereby forming bigger hole. In the final stage of the spreading and hole formation processes the rim of all the holes are spontaneously break up into a collection of liquid lenses. It was observed that the total length of triple line separating the aqueous phase and oil film is a good measure of emulsification dynamics. To systematically study the effect of dissolution process, spreading experiments were also repeated by saturating the water substrate with the Octanoic acid (OA) where the solubility limit is already known. Spreading experiments of coconut oil on the aqueous solutions of Octanoic acid shows a decrease in spreading rate if the saturation levels of the substrate is increased.

Vignatha is working on the project “Prediction of the likelihood of diabetes” based on glycosylated haemoglobin and anthropometric data(height, weight, body mass index, body adiposity index, age, gender).
As per the American Diabetes Association(ADA), Fasting Plasma Glucose level>=7mmol/l(126mg/dl) and/or 2-hr post glucose level>=11.1 mmol/l(200 mg/dl) is the criteria for diagnosis of diabetes. But this test needs a strict fasting state before blood glucose measurement. So, HbA1c has been incorporated as a diagnostic test for diabetes by ADA as it has the advantage of not requiring a fasting state.A HbA1c value >6.3 has been found to be optimum cut-off value to diagnose diabetes for the rural population of Rayalaseema region of Andhra Pradesh. However this test cannot be used as a confirmatory test as it is giving prediction results with 89% sensitivity and 81% specificity only.
A random forest model built on HbA1c values and anthropometric data is observed to give 100% sensitivity and 100% specificity. This model is as good as the OGTT gold standard test and the diagnosis can be done in a single visit of the patient unlike others. The random forest model built on FPG values and anthropometric data has picked up an additional 84 cases of diagnosed diabetes and missed 1 case when compared to the prediction of diabetes by the FPG cut-off>=126 mg/dl incorporated from the OGTT gold standard test.

The brachistochrone curve for a non-dissipative particle tries to maximize inertia of the particle but for a fluid filled cylinder, increasing inertia would amount to increased dissipative losses. Hence the trade-off between inertia and dissipation plays a vital role in determining the brachistochrone curve of a fluid filled cylinder. This trade-off manifests itself in the form of an integro-differential equation governing the angular acceleration of the cylinder. Here, we compute the brachistochrone curve of a fluid filled cylinder using optimal control principles and investigate the effect of the aforementioned trade-off on the deviation of the brachistochrone curve from that of a non-dissipative particle. Also, we investigate the effects of the non-dimensional parameters of the problem on the shape of the brachistochrone curve. Finally, we analyze the stability of the time varying fluid flow in the cylinder and find an admissible region for the terminal point which would ensure the stability of the fluid flow as the cylinder rolls along the brachistochrone curve.

This study aims to characterize the dynamic behavior of sprays using time series analysis techniques. Phase Doppler particle analyzer has been used to acquire the size-velocity data of the sprays. We study the conditional probability of droplets of a given size or velocity following another droplet. A scalar distance measure used to quantify the distance of the state from a pure random state, indicates that airblast sprays have more ordering towards the edge of spray. Analysis of inter-arrival time of droplets show that droplets which are closely spaced in time coordinate are also closely spaced in size and velocity coordinates. Autocorrelation and partial autocorrelation plots reveal the presence of inherent correlated motion in the spray. This presence of correlated motion renders the spray generation process predictable and hence also justifies the use of time series models. Analysis of residuals using Ljung-Box test revealed the presence of heteroscedasticity in the diameter series and absence of the same in axial velocity series. In order to account for this heteroscedasticity, appropriate GARCH models were used to describe the process. Further Granger causality analysis indicates the direction of causality as $Diameter\rightarrow Velocity$ in the center of the spray and as $Velocity\rightarrow Diameter$ at the edges of the spray. Instantaneous causality analysis was performed at seven radial locations. All seven locations tested positive for the presence of instantaneous causality. In this approach arrival time is not utilized.

The human lungs, a grand design of nature, consists of a sequence of bifurcations known as bronchi. The change of length and diameter of bronchi at each generation of bifurcation gives rise to a wide range of airflow velocities during inhalation. From trachea to alveoli the Reynolds number (Re) varies from 103 to 10-2 generating different types of flows starting from turbulent flow in the trachea to diffusion dominated flow in alveoli. This variation of dimension and Re in the bronchi dictates the deposition of aerosols at different generations which have been studied experimentally.

Effectiveness of pulmonary drug delivery mechanism is contingent on aerosol particles reaching the terminal generations of the lung. Research on aerosol transport in lungs has, however, been mostly restricted to upper branches of the lung. This has resulted in a lack of understanding on transport of aerosols in the deep lung and hence, on effectiveness of aerosols in the pulmonary drug delivery mechanism. The present study aims to develop a fundamental understanding of aerosol transport mechanism in the deep lung with the objective of enabling personalized pulmonary drug delivery through computational and experimental analyses.

Human brain responds very quickly at times of disaster. Crowd under a panic situation may react according to the signals given by the brain and decision making under highly tensed conditions may even lead to loss of lives. s The motive of present work is to understand the dynamics of the panicked crowd and formulate a model that is efficient for safe evacuation.

Fig. Snapshot of a crowd evacuation

The human exhaled breath is predominantly turbulent. During exhalation, the air is forced to flow out of the lung through the windpipe as a result of the contraction of a thin muscle called diaphragm underneath the lungs. The air passes through the throat and oral cavity. The extrathoracic airway is unique for each individual. Turbulent flows from the airway carry the signature of the source of generation i.e., the geometry of the upstream flow region. Therefore, humans can be classified based on the turbulent signatures in their exhaled breath, which becomes the motivation of this study.
The learning process by definition is ‘progressive performance improvement on a specific task’. Machine learning for our task requires powerful information processing algorithms relevant to problems in fluid mechanics. Machine learning algorithms such as logistic regression, decision trees, boosting trees can be used on the features extracted using MFDFA (Multifractal Detrending Fluctuation Analysis) to check whether there is some uniqueness in an individual’s exhalation. These are supervised learning techniques that are used for classification and regression problems, where target values are available beforehand to train the algorithm. Unsupervised machine learning techniques such as K-means clustering and dendrogram can be used in problems where target values are not available to train the machine, which is the case when we make efforts to group humans into different breathing classes.

Kushal Ghate has worked on optimizing the simplex atomizer geometry with a principle focus on improving atomization by reducing resultant drop size and enable a wide dispersion of droplets in the spray chamber. Spray characteristics of several atomizers with different internal geometry (change in convergence angle, divergence angle, insertion of needle inserts) were analyzed and compared with those of conventional simplex atomizer. As a sequel to the extensive studies carried out with a simplex atomizer, he also developed a novel Turbo Swirl Atomizer (TSA) which exhibits better performance than simplex atomizer in many aspects.
A patent has been applied for the concept of TSA which is pending. He has won several prizes for this project including third prize in the student project competition of AERO India show, held at Yelahanka Air force station (Bangalore). He has used several experimental techniques for spray characterization such as backlight shadowgraphy, microscopic imaging, Particle Image Velocimetry (PIV), Laser Doppler Velocimetry (LDV) and Phase Doppler Particle Analyzer (PDPA). His research interests are Liquid atomization and Sprays, Fluid dynamics and Multiphase flows, Laser-based diagnostics and other experimental techniques. Currently, he is working on developing a machine learning algorithm for airblast atomizers.