Chirabrata Bhaumik
Chirabrata Bhaumik has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- Arpan Pal
TCS Research, Tata Consultancy Services Limited, Kolkata, India
8 shared publications - Tapas Chakravarty
TCS Research, Kolkata, India
6 shared publications - Abhijeet Gorey
TCS Research
4 shared publications - Amit Swain
TCS Research, Kolkata, India
3 shared publications - Avik Ghose
TCS Research, Kolkata, West Bengal, India
3 shared publications - Mayukh Biswas
TCS Research, Kolkata, India
3 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (10) , 6011904
A Radar-Acoustic Sensor Fusion Unobtrusive System for Imaging High-Impact Regions Inside Opaque Rotating Ball Mills
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (10) , 6011904
A Radar-Acoustic Sensor Fusion Unobtrusive System for Imaging High-Impact Regions Inside Opaque Rotating Ball Mills
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
STARK: A Comprehensive Framework for PLC Structured Text Assistant for Reasoning and Knowledge
Structured Text (ST), a core language in IEC 61131-3 for industrial automation, remains challenging to author manually due to its rigid syntax and domain-specific semantics. We present STARK-Structured Text Assistant for Reasoning and Knowledge, a knowledge augmented system for automatic ST code generation from natural language specifications. STARK integrates a Large Language Model(LLM) with a curated PLC code knowledge base and a dynamic few-shot prompting mechanism. Trained on a customized parallel dataset of Natural Language–ST Code pairs stratified by logical complexity, STARK identifies control intent, retrieves relevant domain knowledge, and synthesizes executable ST code. Evaluation using expert review and LLM-assisted analysis shows that STARK significantly outperforms baseline models in syntactic and semantic correctness for easy and medium complexity tasks. Error visualization reveals improved handling of negations, conditional structures, and temporal operators. Our results demonstrate the effectiveness of structured reasoning and knowledge integration in domain-specific code generation.
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessSTARK: A Comprehensive Framework for PLC Structured Text Assistant for Reasoning and Knowledge
Structured Text (ST), a core language in IEC 61131-3 for industrial automation, remains challenging to author manually due to its rigid syntax and domain-specific semantics. We present STARK-Structured Text Assistant for Reasoning and Knowledge, a knowledge augmented system for automatic ST code generation from natural language specifications. STARK integrates a Large Language Model(LLM) with a curated PLC code knowledge base and a dynamic few-shot prompting mechanism. Trained on a customized parallel dataset of Natural Language–ST Code pairs stratified by logical complexity, STARK identifies control intent, retrieves relevant domain knowledge, and synthesizes executable ST code. Evaluation using expert review and LLM-assisted analysis shows that STARK significantly outperforms baseline models in syntactic and semantic correctness for easy and medium complexity tasks. Error visualization reveals improved handling of negations, conditional structures, and temporal operators. Our results demonstrate the effectiveness of structured reasoning and knowledge integration in domain-specific code generation.
- 28 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (1) , 1-4
A Data-Driven Approach to Leak Identification and Severity Analysis in Pipelines Using Acoustic Sensing and Deep Learning
Gas pipeline leaks in industrial environment can be hazardous, necessitating timely remediation. Owing to the shortcomings of contact-based sensing methods, noncontact localization and diagnostics is necessary. Leak severity, alongside leak localization, is a critical parameter requiring long-term monitoring due to its nonstationary nature. Existing algorithms become computationally expensive in such conditions. Thus, a lightweight data-driven approach is necessary. It has been experimentally found that during training phase, the leak severity gets coupled with leak position due to the directional nature of a leak with respect to the sensor. This work proposes a leak-position-informed neural network training method for leak severity classification. The proposed approach has been compared with various training and testing methods to evaluate their ability to predict leak severity. The proposed method yields accuracy of 93% and 83% in clean and noisy environment over a range of experimental conditions.
- 28 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 10 (1) , 1-4
A Data-Driven Approach to Leak Identification and Severity Analysis in Pipelines Using Acoustic Sensing and Deep Learning
Gas pipeline leaks in industrial environment can be hazardous, necessitating timely remediation. Owing to the shortcomings of contact-based sensing methods, noncontact localization and diagnostics is necessary. Leak severity, alongside leak localization, is a critical parameter requiring long-term monitoring due to its nonstationary nature. Existing algorithms become computationally expensive in such conditions. Thus, a lightweight data-driven approach is necessary. It has been experimentally found that during training phase, the leak severity gets coupled with leak position due to the directional nature of a leak with respect to the sensor. This work proposes a leak-position-informed neural network training method for leak severity classification. The proposed approach has been compared with various training and testing methods to evaluate their ability to predict leak severity. The proposed method yields accuracy of 93% and 83% in clean and noisy environment over a range of experimental conditions.
- 29 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 9 (7) , 1-4
Real-Time Remote Localization of Gas Pipe Leakage Using Acoustic Beamforming and Spatial Filtering—A Lightweight Approach
In-situ monitoring of pipe leak becomes a daunting task, even with state-of-art sensing methods, if the pipe happens to be in an inaccessible indoor location. While acoustic wideband beamforming method exists for the same, the dynamic noise and sensors' susceptibility to malfunction in an industrial environment needs to be addressed. This work presents a joint spectral focusing and fault tolerant paradigm with a novel sound source separation algorithm as a fallback leak diagnostics tool. Computation and communication costs are reduced by processing voluminous array sensor data at the edge, and transmitting the lightweight audio-visual information to the operator for remote localization and diagnostics. The localization accuracy for the proposed method has been found to be within 5% for different experimental conditions.
- 29 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 9 (7) , 1-4
Real-Time Remote Localization of Gas Pipe Leakage Using Acoustic Beamforming and Spatial Filtering—A Lightweight Approach
In-situ monitoring of pipe leak becomes a daunting task, even with state-of-art sensing methods, if the pipe happens to be in an inaccessible indoor location. While acoustic wideband beamforming method exists for the same, the dynamic noise and sensors' susceptibility to malfunction in an industrial environment needs to be addressed. This work presents a joint spectral focusing and fault tolerant paradigm with a novel sound source separation algorithm as a fallback leak diagnostics tool. Computation and communication costs are reduced by processing voluminous array sensor data at the edge, and transmitting the lightweight audio-visual information to the operator for remote localization and diagnostics. The localization accuracy for the proposed method has been found to be within 5% for different experimental conditions.
- 4 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 8 (10) , 1-4
Photoacoustic Sensing System for Noninvasive and Real-Time Measurement of Paint's Viscosity in Flowing Conditions
Inline measurement of paint viscosity in the flowing conditions is extremely important for the paint manufacturing industry. This study proposes a non-invasive, cost-effective and inline method to measure paint's viscosity using frequency domain Photoacoustic (PA) sensing. Through a PA signal, three different frequency and time domain features namely spectral amplitude ratio, acoustic attenuation and acoustic wave velocity are extracted. Due to the lower accuracy (<90%) of the aforementioned features, a novel statistical feature i.e. the harmonic mean is derived from the existing features to enhance the accuracy of the measurement. To mitigate the experimental challenges, the viscosity model is trained from the PA data under static condition and tested for the paint under flowing condition. In the flowing conditions, the accuracy in the measurement is found to be less than 93%. Hence, a correction factor is introduced which considers the doppler shift in the PA wave velocity due to the paint flow. With this correction factor, the accuracy of the viscosity measurement is found to be greater than 95%. The developed viscosity model is validated through the 4-fold cross validation and the results are confirmed for their repeatability and tested with different paint samples.
- 4 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 8 (10) , 1-4
Photoacoustic Sensing System for Noninvasive and Real-Time Measurement of Paint's Viscosity in Flowing Conditions
Inline measurement of paint viscosity in the flowing conditions is extremely important for the paint manufacturing industry. This study proposes a non-invasive, cost-effective and inline method to measure paint's viscosity using frequency domain Photoacoustic (PA) sensing. Through a PA signal, three different frequency and time domain features namely spectral amplitude ratio, acoustic attenuation and acoustic wave velocity are extracted. Due to the lower accuracy (<90%) of the aforementioned features, a novel statistical feature i.e. the harmonic mean is derived from the existing features to enhance the accuracy of the measurement. To mitigate the experimental challenges, the viscosity model is trained from the PA data under static condition and tested for the paint under flowing condition. In the flowing conditions, the accuracy in the measurement is found to be less than 93%. Hence, a correction factor is introduced which considers the doppler shift in the PA wave velocity due to the paint flow. With this correction factor, the accuracy of the viscosity measurement is found to be greater than 95%. The developed viscosity model is validated through the 4-fold cross validation and the results are confirmed for their repeatability and tested with different paint samples.
- Open Access
- 25 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Physics Guided Generative Learning for Domain Adaptable Data Synthesis : Progressive Fault Synthesization for Predictive Machine Maintenance
In this work, we perform Physics guided data synthesis. Proposed method uses seed data from the observed source state / domain for data generation of unobserved target state / domain. Our method adapts the variation of features with respect to the observed source state across unobserved states. This approach uses Physics knowledge and its associated impacts on statistical and signal properties of data. We use generative learning comprising of a variational auto-encoder (VAE) based neural network model. Proposed model has a Physics influenced optimization function to achieve the data generation and adaptation of features across domains. Proposed method aims to overcome the challenges of getting the data for different progressive fault states to achieve improved predictive machine maintenance. We demonstrate the proposed method using real-world gear fault vibration time-series for synthesizing progressive faults from normal to various faulty states.
- 25 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessPhysics Guided Generative Learning for Domain Adaptable Data Synthesis : Progressive Fault Synthesization for Predictive Machine Maintenance
In this work, we perform Physics guided data synthesis. Proposed method uses seed data from the observed source state / domain for data generation of unobserved target state / domain. Our method adapts the variation of features with respect to the observed source state across unobserved states. This approach uses Physics knowledge and its associated impacts on statistical and signal properties of data. We use generative learning comprising of a variational auto-encoder (VAE) based neural network model. Proposed model has a Physics influenced optimization function to achieve the data generation and adaptation of features across domains. Proposed method aims to overcome the challenges of getting the data for different progressive fault states to achieve improved predictive machine maintenance. We demonstrate the proposed method using real-world gear fault vibration time-series for synthesizing progressive faults from normal to various faulty states.
- 22 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 8 (10) , 1-4
Nondestructive Measurement of Microcoating Thickness Using Sweep Frequency Photoacoustic Sensing Technique
Non-destructive and rapid measurement of micro coating thickness is a challenge for most of the paint coating industries. This study proposes a novel sweep frequency photoacoustic sensing technique for the non-destructive estimation of a coating layer thickness. The continuous wave laser diodeused for the sample excitationmakes the overallsystem compact and portable. Also, the usage of frequency sweep enables the technique to be applied to different substrates and different coating layers. The Rosencwaig-Gersho (RG) model is employed to estimate the thickness of the coating layer by measuring the resonance frequency of the photoacoustic wave. Regression analysis is performed to obtain the micro thickness through photoacoustic measurements. The photoacoustic measurements arecompared with the standard (cross-sectional microscopy) thickness measurement technique and the error is found to be less than 5%. The simple computational method and instrumentation used in this technique enable it to be deployed in industries for rapid estimation of the micro thickness.
- 22 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Letters
- Vol. 8 (10) , 1-4
Nondestructive Measurement of Microcoating Thickness Using Sweep Frequency Photoacoustic Sensing Technique
Non-destructive and rapid measurement of micro coating thickness is a challenge for most of the paint coating industries. This study proposes a novel sweep frequency photoacoustic sensing technique for the non-destructive estimation of a coating layer thickness. The continuous wave laser diodeused for the sample excitationmakes the overallsystem compact and portable. Also, the usage of frequency sweep enables the technique to be applied to different substrates and different coating layers. The Rosencwaig-Gersho (RG) model is employed to estimate the thickness of the coating layer by measuring the resonance frequency of the photoacoustic wave. Regression analysis is performed to obtain the micro thickness through photoacoustic measurements. The photoacoustic measurements arecompared with the standard (cross-sectional microscopy) thickness measurement technique and the error is found to be less than 5%. The simple computational method and instrumentation used in this technique enable it to be deployed in industries for rapid estimation of the micro thickness.
- 15 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (6) , 7364-7374
A Novel Approach for Quantifying Impact and Abrasive Wear by Short-Range mmWave InSAR
Runtime inspection of surface wear is imperative to ascertain the production performance of a mill. Mill liners are vulnerable to impact and abrasive wear, subsequently leading to unplanned breakdown and increased downtime. This remains an open-problem owing to the dust-ridden environment of a grinding mill that challenges existing wear profiling systems. In this scenario, millimeter-waves (mm-wave) can propagate through dust and illuminate a visually occluded wear zone. This paper presents a unique approach to detect and quantify wear using short-range mm-wave interferometric synthetic aperture radar (InSAR). We propose a set of algorithms to reduce defocusing of SAR images in short range; automated identification of anchor points for precise InSAR image co-registration; and lastly quantification of surface wear from interferograms. Our experiments demonstrate the improvement in signal-to-noise ratio (SNR) of SAR images and efficacy of the proposed co-registration scheme in interferometric processing. Wear estimation using the resultant interferograms has been carried out with an MSE of 0.266.
- 15 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (6) , 7364-7374
A Novel Approach for Quantifying Impact and Abrasive Wear by Short-Range mmWave InSAR
Runtime inspection of surface wear is imperative to ascertain the production performance of a mill. Mill liners are vulnerable to impact and abrasive wear, subsequently leading to unplanned breakdown and increased downtime. This remains an open-problem owing to the dust-ridden environment of a grinding mill that challenges existing wear profiling systems. In this scenario, millimeter-waves (mm-wave) can propagate through dust and illuminate a visually occluded wear zone. This paper presents a unique approach to detect and quantify wear using short-range mm-wave interferometric synthetic aperture radar (InSAR). We propose a set of algorithms to reduce defocusing of SAR images in short range; automated identification of anchor points for precise InSAR image co-registration; and lastly quantification of surface wear from interferograms. Our experiments demonstrate the improvement in signal-to-noise ratio (SNR) of SAR images and efficacy of the proposed co-registration scheme in interferometric processing. Wear estimation using the resultant interferograms has been carried out with an MSE of 0.266.
- 8 June 2022
- Published by Springer Nature
- p. 247-255
Application of Photoacoustic Sensing in Depicting Viscosity Information of Lubrication Oil
- 8 June 2022
- Published by Springer Nature
- p. 247-255
Application of Photoacoustic Sensing in Depicting Viscosity Information of Lubrication Oil