Lakshminarayanan Samavedham’s research while affiliated with Birla Institute of Technology and Science - Hyderabad Campus and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (58)


An Improved Industrial Fault Diagnosis Model by Integrating Enhanced Variational Mode Decomposition with Sparse Process Monitoring Method
  • Article

September 2024

·

18 Reads

Reliability Engineering & System Safety

·

·

·

Lakshminarayanan Samavedham

Multidisciplinarity in Undergraduate Education: Journey at the National University of Singapore

September 2024

·

1 Read

From iconic paintings—“Mona Lisa” and “The Last Supper”—to designs for flying machines and ground-breaking studies on optics and perspective, Leonardo da Vinci fused science and art to create works that have become part of humanity’s story. The world is at another transition point now, dealing with several human-triggered crises ranging from climate change, water scarcity, and pandemics to geopolitical issues. These are so-called wicked problems which one cannot hope to solve—presence of several actors (humans, animals, plants and other natural elements) with different objectives and ever-changing behaviour, operating and interacting at different time and length scales produce highly unexpected, emergent behaviour. Such problems defy solutions and can, at best, be understood and managed, calling for the script of a new humanity story. In this context, it is important that education, particularly at the undergraduate level, should be more of a generalist nature.




Deep chemometrics using one‐dimensional convolutional neural networks for predicting crude oil properties from FTIR spectral data

September 2023

·

26 Reads

·

1 Citation

The Canadian Journal of Chemical Engineering

The determination of physicochemical properties of crude oils is a very important and time‐intensive process that needs elaborate laboratory procedures. Over the last few decades, several correlations have been developed to estimate these properties, but they have been very limited in their scope and range. In recent years, methods based on spectral data analysis have been shown to be very promising in characterizing petroleum crude. In this work, the physicochemical properties of crude oils using Fourier transform infrared (FTIR) spectrums are predicted. A total of 107 samples of FTIR spectral data consisting of 6840 wavenumbers is used. One dimensional convolutional neural networks (CNNs) were used employing FTIR spectral data as the one‐dimensional input and Keras and TensorFlow were used for model building. The Root Mean Square Error decreased from 160 to around 60 for viscosity when compared to previous machine learning methods like partial least squares (PLS), principal component regression (PCR), and partial least squares regression with genetic algorithm (PLS‐GA) on the same data. The important hyper‐parameters of the CNN were optimized. In addition, a comparison of results obtained with different neural network architectures is presented. Some common preprocessing techniques were also tested on the spectral data to determine their impact on model performance. To increase interpretability, the intermediate neural network layers were analyzed to reveal what the convolutions represented, and sensitivity analysis was done to gather key insights about the wavenumbers that were the most important for prediction of the crude oil properties using the neural network.




Application of artificial neural network for prediction of 10 crude oil properties

May 2023

·

70 Reads

·

5 Citations

The Canadian Journal of Chemical Engineering

This study aims to develop an industrially reliable and accurate method to estimate crude oil properties from their Fourier transform infrared spectroscopy (FTIR) spectra. We used the complete FTIR spectral data of selected crude oil samples from seven different Canadian oil fields to predict 10 important crude oil properties using artificial neural networks (ANNs). The predicted properties include specific gravity, kinematic viscosity, total acid number, micro carbon content, and production of light and heavy naphtha, Kero, and distillate in oil refineries. The 107 different (65 light oil and 42 heavy/medium oil samples) crude oil samples used in this study came from seven oil fields and reservoirs across Canada. In line with standard practice, we used 80% of the dataset for training the ANN models and used the remaining 20% of the crude oil samples to test the models. In the ANN analysis, the mean squared error (MSE) was used as the loss function in models, and the mean absolute prediction error (MAPE) was used as a reference to compare the performance of different neural networks constructed with different numbers of layers. This work demonstrates that FTIR spectroscopy is a promising technique that provides rapid and accurate estimates for the oil properties of interest to the industry. A comparison of the values predicted by the validated ANN models and their corresponding measured (actual) values showed excellent prediction with the acceptable range of error (below 15%) aimed for by our industry partner for all properties except viscosity, for which building models based on the natural logarithmic values of measured viscosities significantly improved the results.


Fig. 1. Analytical Framework.
Fig. 2. Sequence Plots by Learning Cluster/Typology.
Fig. A1. Dendrogram from Agnes Clustering, Ward Algorithm, OM of Spell Sequence Distance.
Regression Models.
The learning process matter: A sequence analysis perspective of examining procrastination using learning management system
  • Article
  • Full-text available

December 2022

·

153 Reads

·

16 Citations

Computers and Education Open

While procrastination has been found as a crucial indicator that negatively affects performance, no research has investigated using a learning management system (LMS) to formulate a learning typology based on daily learning patterns, grouping students into distinct clusters showing procrastination does affect performance via sequence analysis. This new-fangled approach, shifts away from using an inventory to measure procrastination that avoids self-report bias, poor recall, measurement errors, and researcher selection bias, to directly examining the learning pattern of students based on the exact time acts of procrastination. The findings confirmed the negative effect of academic learning procrastination but also surprisingly found that students who were delayed in their submission were positively related to performance. A relative conception of procrastination is proposed to explain the acts of executing procrastination. An analytical framework with a multimodal emphasis is recommended and outlined for studying procrastination that focuses on sequence analysis as the core model for online education research. The findings also discovered that students strategically used certain studying tactics over time to improve their learning which affects performance that could be discovered from the LMS. These strategies include varying their time length in studying, learning at a specific time of the day & day of the week, and practicing an appropriate launching time length for online learning. As the study was carried out during the COVID-19 period when all the students were restricted to online study, it provided empirical evidence of students in a completely online environment, a baseline for planning future online courses.

Download

Figure 4. The optimization-based methodology that was used to estimate the K values using vendor data.
The dimensions and model parameters for the two PRs.
The estimated RUL values for the filter and the valve seat and the annual diaphragm usage for the different PRs.
Health Monitoring of Pressure Regulating Stations in Gas Distribution Networks Using Mathematical Models

August 2022

·

61 Reads

·

1 Citation

Energies

Many cities have extensive distribution networks that supply natural or town gas to domestic, industrial, and power plant consumers. A typical network may have hundreds of pressure regulating stations that are of different types and capacities, but most legacy networks are sparsely instrumented. The reliability of these stations is the first priority for ensuring uninterrupted gas supplies; hence, condition monitoring and prescriptive maintenance are critical. In this study, mathematical models were developed for two types of commonly used regulators: spring-loaded and lever-type regulators. We also considered three faults that are typically of interest: filter choking, valve seat damage, and diaphragm deterioration. The proposed methodologies used the available measured data and mathematical models to diagnose faults, track prognoses, and estimate the remaining useful life of the regulators. The applicability of our proposed methodologies was demonstrated using real data from an existing distribution network. To facilitate industrial use, the methodologies were packaged into a user-friendly dashboard that could act as an interface with the operational database and display the health status of the regulators.


Citations (31)


... Artificial neural network (ANN), least square support vector machine (LSSVM), and multi-gen genetic programming (MGGP) are among the most effective machine learning and artificial network methods which can grasp the relation between several input and output parameters without needing to capture the involved mechanisms (Esene et al., 2020). Several successful applications of these tools can be cited in the literature (Alizadeh et al., 2023;Bassir and Madani, 2019;Fayyaz et al., 2019;Nguyen et al., 2023;Shoushtari et al., 2020). However, there are no adequate modeling studies that utilize the variety of deterministic tools for forecasting and optimizing the GAGD recovery factor. ...

Reference:

Deterministic tools to predict gas assisted gravity drainage recovery factor
Application of artificial neural network for prediction of 10 crude oil properties
  • Citing Article
  • May 2023

The Canadian Journal of Chemical Engineering

... However, applying analytics to enforce a single learning approach across students may also risk undermining students' agency in self-regulating their learning (Herodotou et al., 2017;Howell et al., 2018) and disregard potentially important differences in student prior preparation, learning preferences, and access to resources . Instructors in this profile can be supported in triangulating dashboard information with other data sources (e.g., observations of and survey responses from students) to make robust instructional decisions, especially in cases where only limited and/ or aggregated LMS activity logs are available or the effectiveness of specific learning patterns remains uncertain (for example, see nuanced findings indicating that not all instances of procrastination are negatively associated with performance, Tan & Samavedham, 2022 and that more LMS checking is not always better, Simon & Randall, 2022). ...

The learning process matter: A sequence analysis perspective of examining procrastination using learning management system

Computers and Education Open

... This should be conducted weekly or bi-weekly, aligning with the daily variability of UFG, to swiftly identify and correct meter faults. Such integrated measures ensure minimized gas losses, promoting sustainability and financial stability within the gas distribution network, marking a best practice adopted globally for its efficiency and effectiveness [113,114]. ...

Health Monitoring of Pressure Regulating Stations in Gas Distribution Networks Using Mathematical Models

Energies

... On the other hand, our knowledge of first principles can be leveraged and exploited to reduce the need for large amounts of data. Therefore, developing hybrid AI models is more appropriate for many chemical engineering applications Mann and Venkatasubramanian, 2021;Mann et al., 2022Mann et al., , 2023aMann et al., ,b, 2024Chakraborty et al., 2021Chakraborty et al., , 2020. The importance of using domain knowledge has become evident even in non-technical areas for LLMs. ...

Mechanism Discovery and Model Identification using Genetic Feature Extraction and Statistical Testing
  • Citing Article
  • May 2020

Computers & Chemical Engineering

... Moreover, it is imperative to validate and periodically enhance ML models to guarantee their precision and suitability for clinical use. Nevertheless, AI approaches are the most promising, as it may be observed from recent publications [18][19][20][21][22]. Additional advantage of using AI methods is the availability of the feature analysis such as Shapley Additive Explanations (SHAP) [23], that can help in better understanding of both the diseases and diagnostics itself [24][25][26]. ...

Multi-Class Diagnosis of Neurodegenerative Diseases: A Neuroimaging Machine Learning based Approach
  • Citing Article
  • May 2019

Industrial & Engineering Chemistry Research

... Classification performances using multiple machine learning approaches have been compared, separately on men and women, based on Regional GM features [12]. Several machine learning and deep learning models have been leveraged on sMRI scans to differentiate PD patients [13][14][15][16][17][18][19]. Thus, most of the state-of-the-art methods classify PD patients from healthy control, but limited attempts have been made towards finding subtypes within the disease. ...

Determination of Imaging Biomarkers to Decipher Disease Trajectories and Differential Diagnosis of Neurodegenerative Diseases (DIsease TreND)
  • Citing Article
  • May 2018

Journal of Neuroscience Methods

... The major issue in such systems reported in the literature is their generalizability. That is, although some of these systems have been reported to have an accuracy as high as 100% in classifying multiple diseases and their sub-types [59][60][61][62][63][64], these systems fail to be as accurate when tested on unseen data from different sources and not used in training. In addition, the research discussed thus far has no embedded explainability, which means that the end user (doctor) has no idea of what is going on within the DL model and what the reason was for a particular diagnosis. ...

Machine Learning-Based Framework for Multi-Class Diagnosis of Neurodegenerative Diseases: A Study on Parkinson’s Disease
  • Citing Article
  • December 2016

IFAC-PapersOnLine

... However, the drawbacks of this controller are that the Jacobean of the PK/PD model based on the multi-layer neural perceptron has insufficient learning that leads to errors in the output and that the propofol control action has a big spike in the response. In addition, the authors in [5] explained an advanced regulatory system based on the PID controller and the model predictive controller (MPC)for first-order plus timedelay approximation of the PK/PD model to modify the level of patients' hypnotic during surgical operation, and they used the trial-and-error method to tune the parameters of the PID controller and the prediction horizon for the weights of the MPC. However, the issue with this controller is that the model is built as a first-order model, while in fact, the PK/PD model is of third-order, which leads the actual BIS to reach the steady state in 10 minutes for a moderately hypnotic state with a value of 50. ...

Advanced regulatory controller for automatic control of anesthesia
  • Citing Article
  • January 2008

... According to the comparison with the simulated annealing algorithm, its performance is better than that of the simulated annealing algorithm, but it performs poorly in terms of overall computing speed. Some researchers [6,7] studied the ant colony algorithm to solve the multiobjective resource allocation problem. e literature optimized the pheromone update system and the probability selection form of the ant colony algorithm and finally compared it with the genetic algorithm in solving the employee assignment problem. ...

A novel optimal experiment design technique based on multi-objective optimization and its application for toxin kinetics model of hemodialysis patients
  • Citing Chapter
  • December 2012

Computer Aided Chemical Engineering

... A necessary condition for the entry of digitalization into patent search and analysis are advances in cloud computer technology with extensive network capacities. In response to the pervasive global demand, computer technology has progressed in recent years, providing computing capacity and processing power at ever lower costs [23]. ...

Data driven fault detection using multi-block PLS based path modeling approach
  • Citing Article
  • December 2012

Computer Aided Chemical Engineering