Dr Sivakami Sundari M
@act.edu.in
Ass.Prof
Agni College of Technology
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Yamini. S, Sivakami Sundari M, Sekar. S, Senthil Kumar. K, and Divya. S
IEEE
With the emergence of the Montreal and Kyoto Protocols, there's been a prohibition on refrigerants in vapor compression systems that harm the ozone layer and contribute to global warming. While alternative refrigerants have been suggested, they tend to be more energy-intensive than the conventional ones they replace. A promising approach to mitigate this increased energy consumption lies in the use of innovative lubricants that can enhance system efficiency. Nano-lubricants have recently gained attention for their superior performance compared to traditional lubricants in refrigeration systems. However, formulating nano-lubricants and determining their composition with the desired thermophysical and tribological properties pose significant challenges. The development of nano-lubricants requires extensive experimentation, and predicting the properties of these lubricants, especially those that offer optimal performance, demands even more rigorous experimentation. Recently, Artificial Neural Networks (ANNs) have been employed to forecast system performance based on existing experimental data. In this study, we leverage literature data on nano-lubricants of varying compositions to design an ANN architecture. The model is then trained, assessed, and validated for predicting the properties of nano-lubricants with unknown compositions. The ANN's predictions are subsequently compared with actual results to gauge its accuracy and reliability.
P. Malathi, Sivakami Sundari M, S. Karkuzhali, Umamaheswari. B, A Thenmozhi, and Velu Aiyyasamy
IEEE
Particularly on highways, drowsy driving causes a large number of collisions while driving. With the goal to identify driver sleepiness while enhancing roadway security, it is now vital to grasp the situation and take early remedial steps. Using a proposed Image based Learning Strategy for Drowsiness Identification (ILSDI) and cross-validation with the conventional model called Convolutional Neural Network (CNN), the suggested framework offers an approach to assess the degree of fatigue among drivers according to modifications in a driver eyeballs motion. This will assist in deal with the problem concerning roadway protection. In addition, four types of expressions on the face were identified and categorized—open, closed, blinking, and no gazing using ILSDI and CNN models, indicating levels of tiredness. Finding, following, and analyzing the driver’s face and eyes in real-time to calculate a sleepiness index is the goal of this technology, which operates in different lighting circumstances. Avoiding these kinds of accidents is possible with the help of a driver sleepiness monitoring structure, which uses a digital camera and accompanying software to measure the rate of blinking as well as the dimension of the driver’s eyes. The driver sleepiness recognition system can identify when the driver is getting sleepy and sound an alarm if necessary. It is based on an offline implementation of a deep learning algorithm that uses ILSDI and CNN.
Shashikala Karanth, Rao Preethi Venkatachala, and S Sivakami Sundari
Jaypee Brothers Medical Publishing
Objective: To compare the maternal and neonatal outcomes, particularly the incidence of preterm birth, and identify their risk among pregnant women who were seropositive or seronegative for SARS-CoV-2, during the pandemic. Method: Pregnant women who got admitted to an urban tertiary care center for delivery during the period August 1, 2020, to October 30, 2020, and consented to participate in the study were recruited and followed up until delivery. Among 230 women, 73 pregnant women who tested positive for SARS-CoV-2 were included in the positive cohort and the remaining in the negative cohort. Demographic details, symptoms at presentation, gestational age, laboratory tests done, and treatment given were noted. The outcome measures studied were the incidence of preterm birth, gestational age at admission and delivery, risk factors for preterm birth (PTB), obstetrics/medical complications, drugs given, mode of delivery, and neonatal outcomes, such as birth weight, Apgar scores at 1 and 5 minutes, neonatal complications, need for NICU admission, and SARS-CoV-2 positivity. Results: Among the 73 SARS-CoV-2 pregnant women, 95% were asymptomatic. The incidence of preterm birth was similar in the SARS-CoV-2-positive and SARS-CoV-2-negative cohorts (20.5 vs 22.5%). There were four SARS-CoV-2-positive babies in the positive cohort and none in the negative cohort. The distribution of known risk factors of preterm births and other maternal and neonatal outcomes were also comparable between the positive and negative cohorts. Conclusion: There is no increase in incidence of preterm births in SARS-CoV-2-positive compared to SARS-CoV-2-negative cohort, during the pandemic. Majority of them have asymptomatic infection, and good pregnancy outcomes can be anticipated. © The Author(s). 2021.
Sivakami Sundari M, Yamini S, Kalicharan Rath, Senthil Kumar K, and Palaniammal S
IEEE
Successful takeoff of a flight in a busy airport depends on how effectively various queues in the airports are managed. Due to globalization, the number of passengers flying is ever-increasing and due to this, the number of flights handled by busy airports is increasing exponentially. Queuing theory is generally used to design the number of servers in the existing and new systems to optimize the waiting cost. Accordingly, service providers decide whether to increase the number of service stations or not. However in the case of airport operations unlike other queuing systems number of runways cannot be increased to minimize the flight queue length on runways to reduce waiting cost. Flights waiting for takeoff and landing are resulting in high operating costs. Fuel costs during flight waiting on queues to take off and flight flying due to runway congestion for which flight waiting on the sky for landing are phenomenal. In most of the airports only a single runway is available and all the flights depend on that single runway. In this work flights waiting for the runway are considered as a single server queuing system and are modeled using ANN to predict the queue behavior and thereby maximize the use of runways effectively and minimize waiting cost.
Sivakami Sundari M, Senthil Kumar K, Yamini S, and Palaniammal S
ENGG Journals Publications
Successful ranking of a website by Google or electric charging of vehicle, congestion is pervasive in all domains. This implies that the presence of queues everywhere or in various places simultaneously. Under this environment, a good understanding of the relationship between queueing and delay is essential in the design of mathematical queuing models. However, uncertainty is an unavoidable phenomenon in any decision-making process. Good number of mathematical approaches has been presented in the literature to the analysis of queuing. Uncertainty is usually considered as unidimensional in nature that can be handled with probability theory. The objective of queuing analysis is to offer a reasonably satisfactory service to waiting customers. Queuing theory is not an optimization technique. Rather, it determines the measure of performance of waiting lines, such as the average waiting time in the queue and the productivity of the service facility, which can then be used to design the service installation. Assumed systems and systems that are too complicated to be disturbed are often difficult to study by analytical techniques. Simulation is one technique that can be seen successfully utilized for analyzing such systems. Artificial neural networks (ANN) form a branch of artificial intelligence. Neural networks represent a connection of simple processing elements capable of processing information in response to external inputs. In this work, such a Markovian queue is simulated using ANN and presented the result. The result shows that the ANN is capable of solving complex queuing problems.
Sivakami Sundari M*, , Palaniammal S, and
Blue Eyes Intelligence Engineering and Sciences Engineering and Sciences Publication - BEIESP
Successful formulation of queuing models depends on arrival rate, nature of waiting in queues, type of service and customer leaving the system depends on type of arrival, nature of service, number of servers deputed, type of queues, number of customers approaching for service in the system and delay. Kendall notations are popularly used for designating the queuing models like M/M/C/E/D. Various mathematical models have been developed to solve the queuing problem analytically. However solving queuing models with power of computers is the new area of research and this work intends to develop single server infinite capacity queuing system using Artificial Neural Network(ANN). The results of simulation are compared with that of analytical method.