About

Dr. Sagar Kamarthi is a Professor of Mechanical and Industrial Engineering and the Director of the Data Analytics Engineering Program at Northeastern University in Boston, MA. He holds MS and PhD degrees from Pennsylvania State University and a BS in Chemical Engineering from Sri Venkateswara University, India. His teaching encompasses data analytics and visualization, with research interests centered on machine learning applications in smart manufacturing and personalized healthcare. He has authored over 200 peer-reviewed research papers and has received multiple awards for his teaching and research excellence, including the 2021 Data Analytics and Information Systems Teaching Award from IISE and the 2020 University Excellence in Teaching Award from Northeastern University. Kamarthi is the founding director of the MS in Data Analytics Engineering Program and the founder and advisor of the MS in Advanced and Intelligent Manufacturing. His research focuses on sensing, diagnostics, and prognostics for manufacturing machines, AI for sustainable manufacturing, machine learning models for personalized medicine, computational methods for pain assessment, and engineering education. He is actively involved in projects such as AI-assisted manufacturing and tele-augmented manufacturing, aiming to enhance human capabilities and improve manufacturing resilience and efficiency.

Research topics

• Computer Science
• Artificial Intelligence
• Engineering
• Machine Learning
• Knowledge management
• Business
• Data science
• Data Mining
• Optics
• Bioinformatics
• Composite material
• Materials science
• Biology
• Physics
• Nanotechnology
• Engineering management
• Medicine
• Manufacturing engineering
• Marketing

Selected publications

• Improvements to Disassembly Lot Sizing With Task Control Through Reinforcement Learning

  Journal of Advanced Manufacturing and Processing · 2025-08-18

  articleOpen accessSenior authorCorresponding

  ABSTRACT This research presents a novel methodology to control disassembly tasks for cost‐efficient component recovery from end‐of‐life products, fostering remanufacturing. Inventory management is an integral part of systems that assemble or disassemble products. Unlike assembly systems, disassembly operations pose a unique challenge, as they can lead to inventory accumulation and risk uncontrolled growth without careful management. Disassembly system inventory management is complex due to various factors, including non‐uniform demand for disassembled components, uncertainty in demands for salvage components, the arrival of different end‐of‐life product variants, end‐of‐life product condition variation, and processing time variation. These complexities often lead to unexpected inventory fluctuations, resulting in high inventory costs, inventory shortages, and customer dissatisfaction due to uncertainty in component availability. These inventory fluctuations can be mitigated if a real‐time decision‐making system supports disassembly processes. This study explores an innovative approach to addressing these complexities and controlling disassembly tasks using Deep Reinforcement Learning (DRL). This approach offers a more effective alternative to traditional methods. Experiments on Quantum‐dot LED (QLED), Organic LED (OLED), and Quantum Dot OLED (QD‐OLED) TV disassembly systems demonstrate the effectiveness of the DRL approach. Compared to the Multiple Elman Neural Networks (MENN) method, the DRL model offers a 21% reduction in inventory accumulation and a 12% improvement in demand satisfaction for the disassembly setup in the study.

  PublisherOA PDFDOI

• Explainable time series features for hard disk drive failure prediction

  Engineering Applications of Artificial Intelligence · 2025-04-11 · 2 citations

  articleOpen accessSenior authorCorresponding

  Reliable data storage is crucial for industry digitalization and cloud infrastructure. To prevent data loss and improve maintenance efficiency in data centers, timely replacement of Hard Disk Drives (HDDs) is critical. HDDs are equipped with Self-Monitoring, Analysis, and Reporting Technology (SMART) to track key performance indicators with empirically established thresholds. In recent years, various machine learning models have utilized SMART time series data for early HDD failure prediction. However, decision-makers need greater transparency and explainability to trust and implement these data-driven models. In this work, we proposed a framework that extracts explainable features from the SMART time series and visualizes the feature impact on short-term HDD failure prediction using SHapley Additive exPlanations (SHAP) analysis. We trained an eXtreme Gradient Boosting (XGBoost) model with information-rich features and evaluated the failure detection rate and false alarm rate. We demonstrated the effectiveness of the proposed approach on Backblaze data for Quarter 1 and Quarter 2 of 2022. The model provided a 74.7% failure detection rate with only a 0.73% false alarm rate on the test data from Quarter 3 of 2022, outperforming an existing explainable model benchmark of a 54.68% failure detection rate and an 11.85% false alarm rate. In addition, the sensitivity analysis optimizes the signal length and the lead time to improve prediction accuracy and inform predictive maintenance policies. The results demonstrate the potential of the proposed framework for effective HDD failure prediction with explainable features. The proposed framework is also applicable to other sensor-based industrial equipment monitoring applications.

  PublisherDOI

• Prediction of Clinical Complication Onset using Neural Point Processes

  arXiv (Cornell University) · 2025-02-18

  preprintOpen access

  Predicting medical events in advance within critical care settings is paramount for patient outcomes and resource management. Utilizing predictive models, healthcare providers can anticipate issues such as cardiac arrest, sepsis, or respiratory failure before they manifest. Recently, there has been a surge in research focusing on forecasting adverse medical event onsets prior to clinical manifestation using machine learning. However, while these models provide temporal prognostic predictions for the occurrence of a specific adverse event of interest within defined time intervals, their interpretability often remains a challenge. In this work, we explore the applicability of neural temporal point processes in the context of adverse event onset prediction, with the aim of explaining clinical pathways and providing interpretable insights. Our experiments span six state-of-the-art neural point processes and six critical care datasets, each focusing on the onset of distinct adverse events. This work represents a novel application class of neural temporal point processes in event prediction.

  PublisherOA PDFDOI

• Part Authentication Through Encrypted Geometric-Magnetic Fingerprint Fusion in Cold Spray Additive Manufacturing

  2025-06-23

  articleOpen access

  Abstract Cold spray additive manufacturing (CSAM) has emerged as a promising cost-effective technology for producing high-performance metal parts, particularly in aerospace applications. However, the widespread adoption of cold spray AM in critical industrial sectors has raised concerns regarding intellectual property protection and supply chain integrity. This paper presents a dual-fingerprint authentication framework for cold spray AM parts that combines inherent geometric features with engineered magnetic patterns. Our system exploits the stochastic nature of the cold spray process, where process variations and particle deposition dynamics generate unique, physically unclonable geometric patterns on each part’s surface. These geometric fingerprints are captured through a laser-based 3D scanning system. Additionally, we introduce a magnetic fingerprinting methodology by strategically embedding nickel deposits within copper main deposits during manufacturing, creating distinct magnetic signatures that can be detected non-destructively when the part is magnetized. The dual-fingerprint feature can be leveraged by various encryption algorithms to preserve fingerprint confidentiality throughout the supply chain while enabling secure authentication. To demonstrate its effectiveness, we implement a learning-with-error (LWE) encryption scheme where the magnetic fingerprint serves as an encryption key for the geometric fingerprint data. The system authenticates parts by comparing newly scanned geometric fingerprints with decrypted reference data using the magnetic key. This dual-factor authentication approach provides a robust defense against counterfeiting while protecting valuable manufacturing intellectual property, thereby securing the entire supply chain for cold spray AM parts. Experimental validation demonstrates the system’s effectiveness in authenticating parts non-destructively while maintaining stringent security standards, making it particularly suitable for aerospace and other critical applications where part authenticity and supply chain integrity are paramount.

  PublisherOA PDFDOI

• Influence of Pre-Existing Pain on the Body’s Response to External Pain Stimuli: Experimental Study

  JMIR Biomedical Engineering · 2025-08-20

  articleOpen accessSenior authorCorresponding

  Background: Accurately assessing pain severity is essential for effective pain treatment and desirable patient outcomes. In clinical settings, pain intensity assessment relies on self-reporting methods, which are subjective to individuals and impractical for noncommunicative or critically ill patients. Previous studies have attempted to measure pain objectively using physiological responses to an external pain stimulus, assuming that the participant is free of internal body pain. However, this approach does not reflect the situation in a clinical setting, where a patient subjected to an external pain stimulus may already be experiencing internal body pain. Objective: This study investigates the hypothesis that an individual's physiological response to external pain varies in the presence of preexisting pain. Methods: We recruited 39 healthy participants aged 22-37 years, including 23 female and 16 male participants. Physiological signals, electrodermal activity, and electromyography were recorded while participants were subject to a combination of preexisting heat pain and cold pain stimuli. Feature engineering methods were applied to extract time-series features, and statistical analysis using ANOVA was conducted to assess significance. Results: We found that the preexisting pain influences the body's physiological responses to an external pain stimulus. Several features-particularly those related to temporal statistics, successive differences, and distributions-showed statistically significant variation across varying preexisting pain conditions, with P values <.05 depending on the feature and stimulus. Conclusions: Our findings suggest that preexisting pain alters the body's physiological response to new pain stimuli, highlighting the importance of considering pain history in objective pain assessment models.

  PublisherOA PDFDOI

• Personalized Learning Paths: LLM-Based Course Recommendations in Manufacturing Education

  2025-08-21

  articleSenior author
  PublisherDOI

• Induction Motor Failure Detection Using Explainable Features from Motor Currents

  2025-01-27 · 3 citations

  articleSenior author

  Induction motors are critical to the reliable operation of various mechanical systems. Motor Current Signature Analysis (MCSA) provides a non-invasive way to monitor motor health and prevent costly diagnostics and repairs resulting from mechanical and electrical faults. The conventional MCSA techniques use the Fast Fourier Transform (FFT) to analyze three-phase voltage and current signals. To improve failure detection accuracy and provide actionable insights, this work proposes a framework for explainable feature extraction and modeling. The framework is designed to extract explainable features from current signal data and train an XGBoost classifier using failure modes as the model response. SHAP (SHapley Additive exPlanations) analysis is then used to visualize the impact of the features on each failure mode. We collected training data using a Machinery Fault Simulator (MFS-Magnum) with a three-phase, 0.5 HP induction motor. Various defects were intentionally introduced into the experimental setup, including inter-turn short circuits, rotor unbalance, and broken rotor bars. By analyzing the relationship between extracted features and each failure mode, this approach facilitates the detection and identification of various impending motor failures.

  PublisherDOI

• Feature Engineering Toolkit for Predictive Analytics in Engineering and Healthcare Informatics

  2025-01-27

  articleSenior author

  This paper presents a feature engineering toolkit that includes complexity-based and recurrence-based feature extraction methods, along with dependency-focused and importance-focused feature selection methods. The toolkit is designed to extract information with predictive power from sensor data, thereby enhancing the capability of diagnostics, prognostics, and health management (PHM). Complexity-based features measure randomness and dynamic variations within the sensor data, while recurrence-based features identify recurring patterns. Dependency-focused and importance-focused feature selection methods improve model performance by identifying the most relevant features for given tasks. This paper demonstrates the toolkit on four applications: wind turbine bearing condition classification, hydraulic systems condition monitoring, CNC tool wear estimation, and respiratory health monitoring. The analytical results exhibit robust prediction accuracy and improved result explainability.

  PublisherDOI

• pyKCN: A Python Tool for Bridging Scientific Knowledge

  arXiv (Cornell University) · 2024-03-24

  preprintOpen accessSenior author

  The study of research trends is pivotal for understanding scientific development on specific topics. Traditionally, this involves keyword analysis within scholarly literature, yet comprehensive tools for such analysis are scarce, especially those capable of parsing large datasets with precision. pyKCN, a Python toolkit, addresses this gap by automating keyword cleaning, extraction and trend analysis from extensive academic corpora. It is equipped with modules for text processing, deduplication, extraction, and advanced keyword co-occurrence and analysis, providing a granular view of research trends. This toolkit stands out by enabling researchers to visualize keyword relationships, thereby identifying seminal works and emerging trends. Its application spans diverse domains, enhancing scholars' capacity to understand developments within their fields. The implications of using pyKCN are significant. It offers an empirical basis for predicting research trends, which can inform funding directions, policy-making, and academic curricula. The code source and details can be found on: https://github.com/zhenyuanlu/pyKCN

  PublisherOA PDFDOI

• A Novel Nonlinear Nonparametric Correlation Measurement With A Case Study on Surface Roughness in Finish Turning

  arXiv (Cornell University) · 2024-06-11

  preprintOpen accessSenior author

  Estimating the correlation coefficient has been a daunting work with the increasing complexity of dataset's pattern. One of the problems in manufacturing applications consists of the estimation of a critical process variable during a machining operation from directly measurable process variables. For example, the prediction of surface roughness of a workpiece during finish turning processes. In this paper, we did exhaustive study on the existing popular correlation coefficients: Pearson correlation coefficient, Spearman's rank correlation coefficient, Kendall's Tau correlation coefficient, Fechner correlation coefficient, and Nonlinear correlation coefficient. However, no one of them can capture all the nonlinear and linear correlations. So, we represent a universal non-linear non-parametric correlation measurement, g-correlation coefficient. Unlike other correlation measurements, g-correlation doesn't require assumptions and pick the dominating patterns of the dataset after examining all the major patterns no matter it is linear or nonlinear. Results of testing on both linearly correlated and non-linearly correlated dataset and comparison with the introduced correlation coefficients in literature show that g-correlation is robust on all the linearly correlated dataset and outperforms for some non-linearly correlated dataset. Results of the application of different correlation concepts to surface roughness assessment show that g-correlation has a central role among all standard concepts of correlation.

  PublisherOA PDFDOI

Recent grants

• Product Monitoring for Holistic Lifecycle Management via Embedded Sensors

  NSF · $300k · 2007–2010

• Developing Integrative Manufacturing and Production Engineering Curricula That Leverage Data Science

  NSF · $2.0M · 2019–2024

Frequent coauthors

• Abe Zeid

  Northeastern University

  49 shared

• Ibrahim Zeid

  Universidad del Noreste

  36 shared

• Srinivasan Radhakrishnan

  Northeastern University

  34 shared

• Stephen Agboola

  Partners HealthCare Connected Health

  31 shared

• Kamal Jethwani

  25 shared

• John Isaacs

  Northeastern University

  23 shared

• Joseph C. Kvedar

  Massachusetts General Hospital

  21 shared

• Surendra M. Gupta

  Lovely Professional University

  20 shared

Education

• MS, Industrial and Manufacturing Engineering

  Pennsylvania State University

  1999

• Phd, Industrial and Manufacturing Engineering

  Pennsylvania State University

  1994

• BTech, Chemical Engineering

  Sri Venkateswara University

  1983

Awards & honors

• 2024 Faculty Research Team Award
• 2022 Institute of Industrial and Systems Engineers (IISE) Fe…
• 2022 Excellence in Mentoring Award
• 2021 Data Analytics & Information Systems (DAIS) Data Analyt…
• 2019 College of Engineering Martin W. Essigmann Outstanding…