Babji Srinivasan

Events such as pipe burst, pipe clogging, excess demand often disrupt the water distribution network operation. As a result, water utilities have to bear an extra financial burden. The knowledge of such events could help the water supply board formulate and operate the network with minimal loss. This work aims to identify the nature of such uncertain events with limited operational information using simple measures designed to monitor the network locally at nodes. Instead of acquiring information from all the nodes, a few nodes are identified for sensor placement by solving a Mixed Integer Linear Programming (MILP) problem. The objectives for solving MILP are to minimise total sensor cost and maximise the sensitivity of measurements to any given fault. Sensitivity is calculated from operational data generated by performing Monte Carlo simulations. These simulations generate data for each fault of different magnitude at different locations. Gradient boosting trees are trained using the limited operational information from the network. Accuracy of identification of faults of 78% on the passive network and over 81% on active networks was obtained using limited features with extreme gradient boosting (XGBoost) model.

A new technique is developed for assessing the performance of linear and nonlinear Kalman filter based state estimators. The proposed metric will indicate the performance of these state estimators which will be primarily influenced by: (i) difference between the model dynamics and process dynamics and, (ii) various approximations of the nonlinear plant dynamics used in nonlinear Kalman filters. Currently, there exists no such quantification method to analyze the performance of linear and nonlinear Kalman filters, a key requirement for improvement and a practical benchmark for comparison of these state estimation algorithms. The proposed technique uses the generalized Hurst exponent of the prediction errors (difference in measured output and a posteriori estimates) obtained from the state estimators to quantify the performance. This technique could be implemented on-line as it requires only plant operating data and the predicted outputs (from the linear and nonlinear Kalman filters) to assess the performance. Several simulation studies demonstrate the applicability of the proposed performance metric to both linear and non-linear Kalman filters. © 2014 American Automatic Control Council.

Internet of Things (IoT) is a key enabler of Industry 4.0 with networked devices providing sensor data to help manage, automate, streamline and optimize assets, operations and processes. In such industrial IoT settings, reliability and process experts spend a considerable amount of time in creating accurate ground-truth data to assist with the inferencing capabilities of Artificial Intelligence (AI) engines. This process can be time-consuming and sometimes inaccurate, depending on the complexity of data. Accurate expert annotated data is the foundation for many AI applications because data needs to be classified on several bases, for instance into 'normal', 'abnormal' or 'pre-abnormal' states. Such problem formulations can be appropriately addressed using Active Learning (AL) techniques. We propose an AL framework capable of handling two practical challenges: oracle uncertainty and quantification of model performance in the absence of ground truth. Consequently, the proposed approach addresses uncertainties within AL techniques by fusing information pertaining to expertise levels of the human annotators and their confidence levels corresponding to the annotation provided.

Smart grids integrate computing and communication infrastructure with conventional power grids to improve situational awareness, control, and safety. Several technologies such as automatic fault detection, automated reconfiguration, and outage management require close network monitoring. Therefore, utilities utilize sensing equipment such as PMUs (phasor measurement units), smart meters, and bellwether meters to obtain grid measurements. However, the expansion in sensing equipment results in an increased strain on existing communication infrastructure. Prior works overcome this problem by exploiting the sparsity of power consumption data in the Haar, Hankel, and Toeplitz transformation bases to achieve sub-Nyquist compression. However, data-driven dictionaries enable superior compression ratios and reconstruction accuracy by learning the sparsifying basis. Therefore, this work proposes using dictionary learning to learn the sparsifying basis of smart meter data. The smart meter data sent to the data centers are compressed using a random projection matrix prior to transmission. These measurements are aggregated to obtain the compressed measurements at the primary nodes. Compressive sensing-based estimators are then utilized to estimate the system states. This approach was validated on the IEEE 33-node distribution system and showed superior reconstruction accuracy over conventional transformation bases and over-complete dictionaries. Voltage magnitude and angle estimation error less than 0.3% mean absolute percentage error and 0.04 degree mean absolute error, respectively, were achieved at compression ratios as high as eight.

Operator competence is critical to ensure safe and profitable operation in hazard intensive chemical process industries. Human errors account for the majority of the accidents in these industries. Therefore, it is imperative to develop methodologies to assess the competence of operators to minimize human errors. Traditional approaches to elicit operator competence are based on subjective measures and/or measures derived primarily from the process and operator actions. These approaches ignore the cognitive aspects of operators such as perception, decision-making strategy, and workload, which are crucial for improving performance. Recent development in sensor technology has enabled the researchers to measure human cognitive behavior objectively. Sensors such as eye-Tracking, electroencephalography (EEG) and galvanic skin resistance (GSR) are found to provide intrinsic human characteristics that cannot be measured otherwise. In this paper, we discuss how eye-Tracking can be used to capture control room operators' cognition and help infer their competence. Eye-Tracking provides information about the location of a person's gaze, which serves as a trace of their attention allocation. Eye gaze based metrics show a strong correlation with operator cognitive behavior such as situation awareness and cognitive workload during process monitoring tasks.

The Polymer Electrolyte Membrane Fuel Cell is a widely researched fuel cell, and a highly potential candidate for alternate power generation. However, technical issues such as membrane flooding and drying prevent its deployment in many applications. Electrochemical Impedance Spectroscopy (EIS) is a very powerful technique that is used to isolate flooding and drying of the fuel cell from operating data. Such information about the state of operation of the cell is critical to deciding necessary control actions to maintain the health and performance of the cell. However, the time taken in obtaining measurements in EIS can be large enough to allow the cell to flood or dry beyond irreparable damage, rendering it a mere postmortem technique. Moreover, after long durations of perturbation, the cell takes a considerable amount of time to return to its regular operation. A new technique is proposed that uses the concept of EIS, but is computationally faster and gives results comparable with those of traditional EIS. This technique is based on perturbing the cell with a small chirp signal containing large number of frequencies instead of series of small sinusoids at different frequencies. Simulation results on isolation of flooding and drying based on Fast EIS are illustrated and future work directions are indicated.

Cyberinfrastructure (e.g., sensors, actuators and the associated communication network) has become an integral part of our modern power grid. While these cyber technologies enhance situational awareness and operational efficiency, they also expose the physical system to cyber-attacks. In this paper, we consider the problem of transmission system state estimation based on measurements from a number of PMUs. In this context, a PMU data integrity attack, Man-in-the-Middle (MitM) attack that can potentially cause a severe impact on the grid is analyzed. Specifically, we propose a novel method based on an alternate expectation-maximization framework to mitigate the effects of these attacks on the state estimation process. Numerical tests are conducted on IEEE-14, 30 and 118 bus systems with different attack scenarios to validate the developed method. Unlike existing works, the proposed algorithm provides accurate state estimates without any prior knowledge of the location of the attack or the number of meters being attacked.

Machine vision and AI-based techniques hold significant promise for automating the analysis of extensive sports video datasets to uncover longitudinal performance trends. This study introduces an innovative pipeline tailored for the analysis of lengthy top-view boxing training session videos, recorded in uncontrolled natural settings and featuring multiple athletes. Our primary focus lies in capturing the movement patterns of boxers within the ring. Within this research, we present Histotracker, an intelligent rule-based tracking module that connects segmented objects across frames using cosine similarity. Distinguishing itself from existing trackers, this module possesses the unique ability to backtrack and correlate frames with the highest association to maintain continuous tracking information. When compared to various standard approaches, our proposed Histotracker demonstrates remarkable results, boasting a MOTA score of 0.95 In approximately 69% of the total bout videos, there were no occurrences of Identity Switching or Identity Update. These findings hold immense promise for advancing the application of automated video analytics in diverse combat sports.

Sports analytics is a field of study that utilizes camera and sensor data to monitor the athlete’s performance and health to optimize the player's strategy and increase the success rate. Coaches rely on analytics to scout opponents and optimize play calls in gameplay. With the advancement in artificial intelligence, accessible and in-depth data collection has been enabled. The well-grounded technique for performance evaluation in sports analytics is Human Pose Estimation (HPE). Our focus is on real-time action recognition in combat sports like boxing. Existing state-of-the-art deep learning models are heavily parameterized, so can’t be used in real-time in any low-end devices. Apart from this, fine-grained classification in highly dynamic activities in sports are typically performed using sensors only. Our proposed Machine Learning based pipeline provides real-time fine-grained solution for 14 boxing punch types of classification using RGB video only. Our approach includes the implementation of three novel and generalized motion dynamics features that encode velocity as well as acceleration of the pose sequences., 1) Unified-Axis Angular Encoding (UAE), 2) 2D Motion Dynamics Descriptors (2DMDD), 3) Fifth-order Angular Encoding (FAE). We employed classical machine learning algorithms I.e., Support Vector Machine (SVM), Random Forest (RF), and K Nearest Neighbours (KNN) to make a lightweight model and test it on YouTube videos. The average accuracies of pipeline using the proposed features are found to be 55%, 92% and 84% for UAE, 2DMDD, and FAE respectively. Using KNN, we have achieved 99% accuracy on 10-fold cross-validation by using FAE features.