Abhishek Srivastav is a Senior Scientist in the Decision Support and Machine Intelligence Group at UTRC, East Hartford. In his current role he is responsible for devising new algorithms for data-driven diagnostics and prognostics of engineering systems. He has a PhD (2009) in Mechanical Engineering from the Pennsylvania State University, University Park. He has dual masters' degrees in Mathematics (2006) and Mechanical Engineering (2006) also from Penn State, his bachelor degree is in Mechanical Engineering (2003) from the Indian Institute of Technology, Kanpur. His research interests include data mining, machine learning, Graphical Models, Prognostics and Health Management (PHM) and Sensor networks.
Abstract: This paper presents the qualitative nature of communication network operations as abstraction of typical thermodynamic parameters (e.g., order parameter, temperature, and pressure). Specifically, statistical mechanics-inspired models of critical phenomena (e.g., phase transitions and size scaling) for heterogeneous packet transmission are developed in terms of multiple intensive parameters, namely, the external packet load on the network system and the packet transmission probabilities of heterogeneous packet types. Network phase diagrams are constructed based on these traffic parameters, and decision and control strategies are formulated for heterogeneous packet transmission in the network system. In this context, decision functions and control objectives are derived in closed forms, and the pertinent results of test and validation on a simulated network system are presented.
Abstract: Recent literature has reported the theory of symbolic dynamic filtering (SDF) of one-dimensional time-series data and its various applications for anomaly detection
and pattern recognition. This paper extends the theory of SDF in the two-dimensional domain, where symbol sequences are generated from image data (i.e., pixels). Given the symbol
sequence, a probabilistic finite state automaton (PFSA), called the D-Markov machine, is constructed on the principles of Markov random fields to incorporate the spatial information in the local neighborhoods of a pixel. The image analysis algorithm has been experimentally validated on a computer-controlled fatigue test apparatus that is equipped
with a traveling optical microscope and ultrasonic flaw detectors. The surface images of test specimens, made of a polycrystalline alloy, are analyzed to detect and quantify the
evolution of fatigue damage. The results of two-dimensional SDF analysis are in close agreement with those obtained from analysis of one-dimensional time-series data from the ultrasonic sensor, which are simultaneously generated from the same test specimen.
Abstract: Resource aware operation of sensor networks requires adaptive re-organization to dynamically adapt to the operational environment. A complex dynamical system of interacting components (e.g., computer network and social network) is represented as a graph, component states as spins, and interactions as ferromagnetic couplings. Using an Ising-like model, the sensor network is shown to adaptively self-organize based on partial observation, and real-time
monitoring and detection is enabled by adaptive redistribution of limited resources. The algorithm is validated on a test-bed that simulates the operations of a sensor network
for detection of percolating faults (e.g. computer viruses, infectious disease, chemical weapons, and pollution) in an interacting multi-component complex system.
Abstract: This communication introduces a non-intrusive method for void fraction measurement and
identification of two-phase flow regimes, based on ultrasonic sensing. The underlying
algorithm is built upon the recently reported theory of a statistical pattern recognition method
called symbolic dynamic filtering (SDF). The results of experimental validation, generated on
a laboratory test apparatus, show a one-to-one correspondence between the flow measure
derived from SDF and the void fraction measured by a conductivity probe. A sharp change in
the slope of flow measure is found to be in agreement with a transition from fully bubbly flow
to cap-bubbly flow.
Abstract: This article presents an algorithm for adaptive sensor activity scheduling (A-SAS) in distributed sensor networks to enable detection and dynamic footprint tracking of spatial-temporal events. The sensor network is modeled as a Markov random field on a graph, where concepts of Statistical Mechanics are employed to stochastically activate the sensor nodes. Using an Ising-like formulation, the sleep and wake modes of a sensor node are modeled as spins with ferromagnetic neighborhood interactions; and clique potentials are defined to characterize the node behavior. Individual sensor nodes are designed to make local probabilistic decisions based on the most recently sensed parameters and the expected behavior of their neighbors. These local decisions evolve to globally meaningful ensemble behaviors of the sensor network
to adaptively organize for event detection and tracking. The proposed algorithm naturally leads to a distributed implementation without the need for a centralized control. The A-SAS algorithm has been validated for resource-aware target tracking on a simulated sensor field of 600 nodes.
Abstract: This paper presents information-theoretic analysis of time-series data to detect slowly evolving anomalies (i.e., deviations from a nominal operating condition) in dynamical systems. A measure for anomaly detection is formulated based on the concepts derived from information theory and statistical thermodynamics. The underlying algorithm is first tested on a low-dimensional complex dynamical system with a known structure—the Duffing oscillator with slowly changing dissipation. Then, the anomaly detection tool is experimentally validated on test specimens of 7075-T6 aluminum alloy under cyclic loading. The results are presented for both cases and the efficacy of the proposed method is thus demonstrated for systems of known and unknown structures.
Abstract: Maturity of engineering and scientific theories in recent decades has facilitated creation of advanced technology of human-engineered complex (e.g., electro-mechanical,
transportation, and power generation) systems. A vast majority of these systems are often subjected to mechanical vibration. A possible consequence is performance degradation and structural damage that may eventually lead to widespread catastrophic
failures. This chapter presents a recently reported technique of data-driven pattern recognition, called Symbolic Dynamic Filtering (SDF), for online detection of slowly evolving anomalies (i.e., deviation from the nominal characteristics) and the associated
behaviorial uncertainties. The underlying concept of SDF is built upon the principles of Statistical Mechanics, Symbolic Dynamics and Information Theory, where time series data from selected sensor(s) in the fast time scale of the process dynamics are analyzed at discrete epochs in the slow time scale of anomaly evolution. Symbolic dynamic filtering includes pre-processing of time series data using the Hilbert transform. The transformed data is partitioned using the maximum entropy principle to generate the symbol sequences, such that the regions of the data space with more information are partitioned finer and those with sparse information are partitioned coarser. Subsequently,
statistical patterns of evolving anomalies are identified from these symbolic sequences through construction of a (probabilistic) finite-state machine that captures
the system behavior by means of information compression. The concept of SDF has been experimentally validated on a special-purpose computer-controlled multi-degree of freedom mechanical vibration apparatus that is instrumented with two accelerometers
for identification of anomalous patterns due to parametric changes.
Abstract: Data-driven fault diagnosis of a complex system such as an aircraft gas turbine engine requires interpretation of multi-sensor information to assure enhanced performance. This paper proposes feature-level sensor information fusion in the framework of symbolic dynamic filtering. This hierarchical approach involves construction of composite patterns consisting of: (i) atomic patterns extracted from single sensor data and (ii) relational patterns that represent the cross-dependencies among different sensor data. The underlying theories are presented along with necessary assumptions and the proposed method is validated on the NASA C-MAPSS simulation model of aircraft gas turbine engines.
Abstract: A semantic framework for information fusion in sensor networks for object and situation assessment is proposed. The overall vision is to construct machine representations that would enable human-like perceptual understanding of observed scenes via fusion of heterogeneous sensor data. In this regard, a hierarchical framework is proposed that is based on the Data Fusion Information Group (DFIG) model. Unlike a simple set-theoretic information fusion methodology that leads to loss of information, relational dependencies are modeled as cross-machines called relational Probabilistic Finite State Automata using the xD-Markov machine construction. This leads to a tractable approach for modeling composite patterns as structured sets for both object and scene representation. An illustrative example demonstrates the superior capability of the proposed methodology for pattern classification in urban scenarios.
Abstract: This paper presents a statistical mechanics-based approach to investigate critical phenomena and size scaling in communication networks. The qualitative nature of phase transitions in the underlying network systems is characterized; and its static and dynamic critical behaviors are identified. Effects of network size, different routing strategies have been analyzed. In all these analyses, phase transition is considered using a single intensive parameter of the communication network system, namely the external packet load. These problems have been investigated by extensive simulation on the model of a wired communication network.
Abstract: This paper develops a distributed algorithm of decision/awareness propagation in mobile agent systems with a time varying network topology and threshold based agent interaction policy. While message broadcast duration or state updating interval is found to be an actuation parameter for changing time-averaged network topology, the threshold parameter in binary decision policy can be used to trigger or restrain the decision propagation. The influence of (large) seed size on the propagation phenomenon has been exploited to control the threat level threshold, beyond which the awareness propagates throughout the network.
Abstract: A complex network of interdependent components is susceptible to percolating faults. Sensor networks deployed for real-time detection and monitoring of such systems require adaptive re-distribution of resources for an energy-aware operation. This paper presents a statistical mechanical approach to adaptive self-organization of a sensor network for detection and monitoring of percolating faults. A complex dynamical system of interdependent components (e.g. computer and social network) is represented as an Ising-like model where component states are modeled as spins, and interactions as ferromagnetic couplings. Using a recursive prediction and correction methodology the sensor network is shown to adaptively self-organize to the dynamic environment and real-time detection and monitoring is enabled. The algorithm is validated on a test-bed simulating the operation of a sensor network for detection of percolating faults (e.g. computer viruses, infectious disease, chemical weapons, and pollution) in an interacting multi-component complex system.
