Anuradha Thangavelu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anuradha Thangavelu has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Selligoundanur Subramaniyam Sivaraju
Department of Electrical and Electronics Engineering, RVS College of Engineering and Technology, Kannampalayam, Coimbatore, India
5 shared publications - Srithar Samidurai
Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, India
3 shared publications - Subramaniam Usha
Kongu Engineering College
2 shared publications - A. Senthilnathan
Electrical and Electronics Engineering, Dr.N.G.P. Institute of Technology, Coimbatore, India
1 shared publication Arun Tigadi
Constituent College of KLE Technological University, Hubballi
1 shared publication- Balamurugan Kaliappan
Sri Ramakrishna Engineering College
1 shared publication
Latest Publications
- 22 July 2026
- Published by Springer Nature in SN Computer Science
- Vol. 7 (6) , 646
Energy-Efficient Hybrid Metaheuristic Resource Allocation Model for Dynamic Virtual Machine Scheduling in Cloud Computing Platforms
- 22 July 2026
- Published by Springer Nature in SN Computer Science
- Vol. 7 (6) , 646
Energy-Efficient Hybrid Metaheuristic Resource Allocation Model for Dynamic Virtual Machine Scheduling in Cloud Computing Platforms
- Open Access
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Matéria (Rio de Janeiro)
- Vol. 31
An optimized neuro-fuzzy control approach for pH neutralization in industrial wastewater treatment systems
Industrial wastewater treatment systems often experience difficulties in regulating pH, as the neutralization process is highly nonlinear and time-varying. Conventional controllers such as PI and PID struggle to perform reliably under such variable conditions. To alleviate the limitations of overshoot, slow response, and instability of the response surface models, an ANFIS-based adaptive control strategy optimized using Chicken Swarm Optimization (CSO) is employed. The uniqueness of this study lies in the tuning of ANFIS parameters and fuzzy membership functions using CSO. This enables the adaptive controller to self-tune in response to process variations and disturbances in real time. The CSO-ANFIS hybrid approach combines neural network (NN) learning capability with fuzzy logic interpretability, which is further enhanced by CSO’s global search efficiency. The simulation results show that the proposed CSO-ANFIS controller reduces the percentage overshoot to 0.4712% compared with 4.9163% in the PI controller and achieves a faster settling time of 20.7680 s under disturbance conditions. These improvements demonstrate enhanced stability, tracking accuracy, and robustness for industrial wastewater pH control.
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Matéria (Rio de Janeiro)
- Vol. 31
Open AccessAn optimized neuro-fuzzy control approach for pH neutralization in industrial wastewater treatment systems
Industrial wastewater treatment systems often experience difficulties in regulating pH, as the neutralization process is highly nonlinear and time-varying. Conventional controllers such as PI and PID struggle to perform reliably under such variable conditions. To alleviate the limitations of overshoot, slow response, and instability of the response surface models, an ANFIS-based adaptive control strategy optimized using Chicken Swarm Optimization (CSO) is employed. The uniqueness of this study lies in the tuning of ANFIS parameters and fuzzy membership functions using CSO. This enables the adaptive controller to self-tune in response to process variations and disturbances in real time. The CSO-ANFIS hybrid approach combines neural network (NN) learning capability with fuzzy logic interpretability, which is further enhanced by CSO’s global search efficiency. The simulation results show that the proposed CSO-ANFIS controller reduces the percentage overshoot to 0.4712% compared with 4.9163% in the PI controller and achieves a faster settling time of 20.7680 s under disturbance conditions. These improvements demonstrate enhanced stability, tracking accuracy, and robustness for industrial wastewater pH control.
- 30 October 2025
- Published by Springer Nature in Iranian Journal of Science and Technology, Transactions of Electrical Engineering
- Vol. 50 (2) , 1779-1801
State-of-Health Prediction for Electric Vehicle Battery Systems with Quantum Spatial Graph Convolutional Neural Network Optimized Using Dollmaker Optimization Algorithm
- 30 October 2025
- Published by Springer Nature in Iranian Journal of Science and Technology, Transactions of Electrical Engineering
- Vol. 50 (2) , 1779-1801
State-of-Health Prediction for Electric Vehicle Battery Systems with Quantum Spatial Graph Convolutional Neural Network Optimized Using Dollmaker Optimization Algorithm
- 1 August 2025
- Published by Springer Nature
- p. 260-274
Analysis and Design of PFC with SEPIC Converter: A PQ Theory-Based Current Generation Approach
- 1 August 2025
- Published by Springer Nature
- p. 260-274
Analysis and Design of PFC with SEPIC Converter: A PQ Theory-Based Current Generation Approach
- 23 July 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
An Effective Image Denoising Model Using Improved Deep Learning Techniques with Optimization Algorithm
Physicians and radiologists utilize medical image processing to diagnose diseases. However, medical images can be compromised by noise introduced through various imaging processes, leading to diminished image quality. This degradation manifests as blurred boundaries, suppressed edges, and loss of structural details. Preserving edges and details is crucial for accurate disease diagnosis. Therefore, medical image denoising is essential in aiding physicians with diagnosis. MRI, CT, X-ray, and ultrasound images are examples of common medical image types. To overcome the above problems, this work implemented an Improved Convolutional Neural Network (ICNN) with the Self-Improved Orca Predation Algorithm (SI-OPA) for image denoising for medical image denoising. ICNN has been used to treat noise-distorted medical images. The CNN’s weights are modified by SI-OPA, increasing the precision of denoising. Comprehensive evaluations concern cutting-edge denoising approaches, encompassing conventional, and deep learning-based methods, emphasizing denoising efficacy, computing efficiency, and picture detail retention. Compared to traditional image denoising algorithms, the proposed method offers superior noise suppression, as evidenced by improved accuracy, Structural Similarity (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Standard Deviation (STD) metrics. Additionally, experimental findings support the suggested denoising algorithm’s viability and efficiency. Finally, the findings show that whereas the current approaches only reach 86%, 82%, 78%, and 65% of the study results, the suggested model achieves a high 94%. Accuracy is 91%, SSIM is 0.91, PSNR is 45.75%, and STD metrics is 43.89%.
- 23 July 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
An Effective Image Denoising Model Using Improved Deep Learning Techniques with Optimization Algorithm
Physicians and radiologists utilize medical image processing to diagnose diseases. However, medical images can be compromised by noise introduced through various imaging processes, leading to diminished image quality. This degradation manifests as blurred boundaries, suppressed edges, and loss of structural details. Preserving edges and details is crucial for accurate disease diagnosis. Therefore, medical image denoising is essential in aiding physicians with diagnosis. MRI, CT, X-ray, and ultrasound images are examples of common medical image types. To overcome the above problems, this work implemented an Improved Convolutional Neural Network (ICNN) with the Self-Improved Orca Predation Algorithm (SI-OPA) for image denoising for medical image denoising. ICNN has been used to treat noise-distorted medical images. The CNN’s weights are modified by SI-OPA, increasing the precision of denoising. Comprehensive evaluations concern cutting-edge denoising approaches, encompassing conventional, and deep learning-based methods, emphasizing denoising efficacy, computing efficiency, and picture detail retention. Compared to traditional image denoising algorithms, the proposed method offers superior noise suppression, as evidenced by improved accuracy, Structural Similarity (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Standard Deviation (STD) metrics. Additionally, experimental findings support the suggested denoising algorithm’s viability and efficiency. Finally, the findings show that whereas the current approaches only reach 86%, 82%, 78%, and 65% of the study results, the suggested model achieves a high 94%. Accuracy is 91%, SSIM is 0.91, PSNR is 45.75%, and STD metrics is 43.89%.
- Open Access
- 25 June 2025
- Published by MDPI AG in World Electric Vehicle Journal
- Vol. 16 (7) , 355
Reduction of Current Harmonics in BLDC Motors Using the Proposed Sigmoid Trapezoidal Current Hysteresis Control
Brushless DC (BLDC) motors are widely used in applications such as Electric Vehicles (EVs) due to their high efficiency, low maintenance, and favorable torque-to-mass ratio. However, one major challenge in BLDC motors is the presence of current harmonics, which can lead to increased noise, vibration, and reduced efficiency, particularly at low speeds or light loads. These harmonics primarily arise from abrupt current transitions during phase commutation. To address this, thispaper presents an innovative approach that combines the Proposed Sigmoid Trapezoidal Current Model with hysteresis control to reduce current harmonics. The model facilitates smooth current changes by applying a sigmoid function, replacing sharp transitions with gradual ones, thus significantly minimizing harmonic distortion. Additionally, hysteresis PWM control enhances the system by precisely regulating the current and dynamically adjusting the switching frequency to maintain the current within a defined range. Simulation results confirm the effectiveness of this method, showing substantial reductions in current harmonics, speed ripple, and torque ripple. Specifically, the proposed method reduces torque ripple by 81% compared to traditional Electronic Commutation Control and improves torque ripple by 30% compared to the conventional method.
- 25 June 2025
- Published by MDPI AG in World Electric Vehicle Journal
- Vol. 16 (7) , 355
Open AccessReduction of Current Harmonics in BLDC Motors Using the Proposed Sigmoid Trapezoidal Current Hysteresis Control
Brushless DC (BLDC) motors are widely used in applications such as Electric Vehicles (EVs) due to their high efficiency, low maintenance, and favorable torque-to-mass ratio. However, one major challenge in BLDC motors is the presence of current harmonics, which can lead to increased noise, vibration, and reduced efficiency, particularly at low speeds or light loads. These harmonics primarily arise from abrupt current transitions during phase commutation. To address this, thispaper presents an innovative approach that combines the Proposed Sigmoid Trapezoidal Current Model with hysteresis control to reduce current harmonics. The model facilitates smooth current changes by applying a sigmoid function, replacing sharp transitions with gradual ones, thus significantly minimizing harmonic distortion. Additionally, hysteresis PWM control enhances the system by precisely regulating the current and dynamically adjusting the switching frequency to maintain the current within a defined range. Simulation results confirm the effectiveness of this method, showing substantial reductions in current harmonics, speed ripple, and torque ripple. Specifically, the proposed method reduces torque ripple by 81% compared to traditional Electronic Commutation Control and improves torque ripple by 30% compared to the conventional method.
- 11 January 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Modeling, Simulation, and Scientific Computing
- Vol. 17 (03)
Cloud-based edge fusion for smart grid powered by blockchain technology
In the new generation of power grids, the smart grid (SG) integrates sophisticated characteristics, including situation awareness, two-way communication, and distributed energy supplies. Integrated SG uses various operational metrics, including devices with sensors, meters, and renewable power sources. There are several challenges when securely disposing and storing electricity data acquired from an SG. It is vulnerable to cyberattacks due to its digitization and integration of an increasing number of links. Issues with latency, security, privacy, and excessive bandwidth consumption will arise when this enormous amount of data is transmitted directly to the cloud. Edge computing (EC) solves this problem by moving data processing to the network’s periphery, close to the embedded devices. With improved data processing speeds, a more responsive and resilient grid may be achieved, instantly responding to energy demand and supply changes. EC reduces the volume of sensitive data sent to central servers, reducing potential security breaches. Data may be better protected from intrusions by being analyzed locally and only pertinent information transferred to the cloud. Thus, a blockchain is an intriguing SG paradigm solution with many benefits. The SG’s decentralization and improved cybersecurity have prompted a lot of work into using blockchain technology; since it is well-known that data saved in the blockchain is immutable, it is crucial to find foolproof ways to verify data are accurate and comply with high-quality standards before storing it in the blockchain. A practical solution for storing precise power data that enables the safe execution of adaptable transactions is a Cloud-Edge Fusion Blockchain model for the smart grid (CEFBM-SG). Consequently, the SG’s dependability, resilience, and scalability will be improved as the number of distributed energy sources (DERs) connected to it increases. Utilizing the idea of computing at the edge to enhance responsiveness and dependability. Executed security analyses and performance evaluations demonstrate CEFBM-SG’s exceptional security and efficiency.
- 11 January 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Modeling, Simulation, and Scientific Computing
- Vol. 17 (03)
Cloud-based edge fusion for smart grid powered by blockchain technology
In the new generation of power grids, the smart grid (SG) integrates sophisticated characteristics, including situation awareness, two-way communication, and distributed energy supplies. Integrated SG uses various operational metrics, including devices with sensors, meters, and renewable power sources. There are several challenges when securely disposing and storing electricity data acquired from an SG. It is vulnerable to cyberattacks due to its digitization and integration of an increasing number of links. Issues with latency, security, privacy, and excessive bandwidth consumption will arise when this enormous amount of data is transmitted directly to the cloud. Edge computing (EC) solves this problem by moving data processing to the network’s periphery, close to the embedded devices. With improved data processing speeds, a more responsive and resilient grid may be achieved, instantly responding to energy demand and supply changes. EC reduces the volume of sensitive data sent to central servers, reducing potential security breaches. Data may be better protected from intrusions by being analyzed locally and only pertinent information transferred to the cloud. Thus, a blockchain is an intriguing SG paradigm solution with many benefits. The SG’s decentralization and improved cybersecurity have prompted a lot of work into using blockchain technology; since it is well-known that data saved in the blockchain is immutable, it is crucial to find foolproof ways to verify data are accurate and comply with high-quality standards before storing it in the blockchain. A practical solution for storing precise power data that enables the safe execution of adaptable transactions is a Cloud-Edge Fusion Blockchain model for the smart grid (CEFBM-SG). Consequently, the SG’s dependability, resilience, and scalability will be improved as the number of distributed energy sources (DERs) connected to it increases. Utilizing the idea of computing at the edge to enhance responsiveness and dependability. Executed security analyses and performance evaluations demonstrate CEFBM-SG’s exceptional security and efficiency.
- Open Access
- 15 December 2024
- Published by Yayasan Riset dan Pengembangan Intelektual in Journal of Applied Engineering and Technological Science (JAETS)
- Vol. 6 (1) , 593-609
Deep Learning Techniques for MRI Image-Based Performance Analysis of Brain Tumor Classification
Brain tumors can produce symptoms and indicators due to direct tissue death, localized invasion of the brain, or aftereffects from increased intracranial pressure. In order to identify images from the publicly available image dataset, this work combined multiple image feature sources using deep learning algorithms. The architecture of most classic convolutional neural networks (CNNs) consists of convolution modification and max-pooling of layers connected with several completely linked layers. The steps used in this system are pre-processing, segmentation, feature extraction, and classification. The preprocessing procedures of this investigation were used by the modified trimmed median filtering approach. U-Net segmentation is used to carry out the segmentation process. Features are then extracted using the wavelet transform method. In this study, MRI images of brain tumors, including meningnant and benign tumors, were detected and classified using the proposed CNN-based VGG16 model. The convolutional neural network (CNN) architectures employed in this investigation were guided by the VGG-16. The outcomes are assessed in terms of accuracy, precision, recall, and F1-score after the suggested model has been simulated. According on the test findings, the recommended approach may lead to 96.9% maximum recall, 97.4% maximum F1-score, 98.45% maximum accuracy, and 98.1% maximum precision.
- 15 December 2024
- Published by Yayasan Riset dan Pengembangan Intelektual in Journal of Applied Engineering and Technological Science (JAETS)
- Vol. 6 (1) , 593-609
Open AccessDeep Learning Techniques for MRI Image-Based Performance Analysis of Brain Tumor Classification
Brain tumors can produce symptoms and indicators due to direct tissue death, localized invasion of the brain, or aftereffects from increased intracranial pressure. In order to identify images from the publicly available image dataset, this work combined multiple image feature sources using deep learning algorithms. The architecture of most classic convolutional neural networks (CNNs) consists of convolution modification and max-pooling of layers connected with several completely linked layers. The steps used in this system are pre-processing, segmentation, feature extraction, and classification. The preprocessing procedures of this investigation were used by the modified trimmed median filtering approach. U-Net segmentation is used to carry out the segmentation process. Features are then extracted using the wavelet transform method. In this study, MRI images of brain tumors, including meningnant and benign tumors, were detected and classified using the proposed CNN-based VGG16 model. The convolutional neural network (CNN) architectures employed in this investigation were guided by the VGG-16. The outcomes are assessed in terms of accuracy, precision, recall, and F1-score after the suggested model has been simulated. According on the test findings, the recommended approach may lead to 96.9% maximum recall, 97.4% maximum F1-score, 98.45% maximum accuracy, and 98.1% maximum precision.