Communicated to EWSHM, UK, July 2018
DAMAGE DETECTION IN PRESENCE OF VARYING TEMPERATURE THROUGH RESIDUAL ERROR MODELLING APPROACH WITH DUAL NEURAL NETWORK
Modal property of structural system gets affected not only due to the presence of damage but also due to a variation of environmental agents like temperature, humidity etc. Detection of damage through modal parameter comparison thus may lead to false predictions. This article presents a two-stage data-driven approach in which damage detection and localization are performed in consequence by combining Auto-Associative Neural Network and RBF environment for classification.
Presented in ICRESH, IIT Madras, January 2019
DESIGN AND HEALTH MONITORING OF TENSEGRITY STRUCTURES: AN OVERVIEW
The very first step to do, when you're eager to work in a new field is to go through its available literature. When the field is relatively newer, it is difficult to gather and skim through the important stuff. To ease this ache, this paper earnestly attempts to compile most of the works that are relevant to the topic at hand.
This work has been published as a book chapter in Reliability, Safety and Hazard Assessment for Risk-Based Technologies, Springer, Singapore.
Published in IFAC Conference
CORRENTROPY BASED IPKF FILTER FOR PARAMETER ESTIMATION IN PRESENCE OF NON-STATIONARY NOISE PROCESS
Existing filtering based structural health monitoring (SHM) algorithms assume a constant noise environment which does not always conform to the reality as noise is hardly stationary. Thus to ensure optimal solution even with non-stationary noise processes, the assumed statistical noise models have to be updated periodically. This work incorporates a modification in the existing Interacting Particle-Kalman Filter (IPKF) to enhance its detection capability in the presence of non-stationary noise processes.
Presented in EC 2018, JU
PLATE DAMAGE DETECTION UNDER VARYING TEMPERATURE USING DUAL NEURAL NETWORK
The structural modal property gets affected not only due to damage but also variation in ambient temperature, humidity etc. Detection of damage through modal comparison thus may lead to false predictions. This article presents a two-stage data-driven approach in which damage detection and localization are performed in consequence. The proposed algorithm is further validated using numerical experiments.
Published in Recent Advances in Computational Mechanics and Simulations
DAMAGE DETECTION IN PRESENCE OF VARYING TEMPERATURE USING MODE SHAPE AND A TWO-STEP NEURAL NETWORK
The dynamic characteristics of any structural system get affected not only due to damage but also from variations in ambient uncertainty. Thus, false positive or negative alarm may be signalled if temperature effects are not taken care off. The difficulty lies in correlating response measurements to corresponding damage patterns in the presence of varying temperature. This study employs machine learning algorithm to filter out the temperature effect from the measured mode shapes. A two-stage data-driven approach has been developed in which damage detection and localization are performed in consequence. For detection, a model to correlate mode shapes and temperature is formulated using an Auto-Associative Neural Network (AANN) and a temperature-invariant prediction error is defined as Novelty Index (NI). NIs are further classified to corresponding damage cases employing a fully connected layer network. With numerical experiments, the algorithm presented excellent efficiency and robustness against varying temperature in detecting damage.
Published in Journal of Structural Control and Health Monitoring
Online structural damage identification technique using constrained dual extended Kalman filter
Periodic health assessment of large civil engineering structures is an effective way to ensure safe performance all through their service lives. Dynamic response‐based structural health assessment can only be performed under normal/ambient operating conditions. Existing Kalman filter‐based parameter identification algorithms that consider parameters as the only states require the measurements to be sufficiently clean in order to achieve precise estimation. On the other hand, appending parameters in an extended state vector in order to jointly estimate states and parameters is reported to have convergence issues. In this article, a constrained version of the dual extended Kalman filtering (cDEKF) technique is employed in which two concurrent extended Kalman filters simultaneously filter the measurement response (as states) and estimate the elements of state transition matrix (as parameters).
Published in Journal of Civil Structural Health Monitoring (JCSHM)
ONE-DIMENSIONAL CONVOLUTIONAL NEURAL NETWORK-BASED DAMAGE DETECTION IN STRUCTURAL JOINTS
Structural health monitoring research traditionally focuses on detecting damage in members excluding the possibility of weakened joint conditions. Efficient model-based joint damage detection algorithms demand computationally expensive model that may affect the promptness of detection. Deep learning techniques have recently come up as an efficient alternative to this cause. This article proposes an output-only approach for joint damage detection in which a 1D-Convolutional Neural Network (CNN) has been introduced to locate weakened joints in semi-rigid frames. Numerical validation is performed on a 2D-steel frame under different damage locations and severities followed by experimental validation on a steel frame structure.
Published in Journal of Mechanical Systems and Signal Processing
STRUCTURAL HEALTH MONITORING WITH NON-LINEAR SENSOR MEASUREMENTS ROBUST TO UNKNOWN NON_STATIONARY INPUT FORCING
Bayesian filtering based structural health monitoring algorithms typically assume stationary white Gaussian noise models to represent an unknown input forcing. However, typical structural damages occur mostly under the action of extreme loading conditions, like earthquake or high wind/waves, which are characteristically non-stationary and non-Gaussian. Clearly, this invalidates this basic assumption, causing these algorithms to perform poorly under non-stationary noise conditions. This paper extends an existing interacting filtering algorithm to efficiently estimate structural damages while being robust to unknown non-stationary non-Gaussian input forcing. Furthermore, this approach is generalized beyond linear measurements to encompass the case of non-linear measurements such as strains.
Published in Journal of Bridge Engineering
Bridge Damage Detection in Presence of Varying Temperature Using Two-Step Neural Network Approach
The dynamic properties of bridges can be affected not only through damage but also from ambient uncertainty. False-positive or negative alarms may be raised if environmental effects are not considered in the detection algorithm. This article presents a two-step data-driven approach that can incorporate temperature effects in vibration-based damage detection and localization, provided the temperature is also measured. To detect the occurrence of damage, prediction errors of an autoassociative neural network (AANN) framework are first employed as a temperature-invariant novelty index (NI).
Published in Journal for Acta Mechanica
Progressive damage identification using dual extended Kalman filter
Existing Kalman filter-based parameter identification algorithms estimate the system parameters as either sole states or a subset of augmented states. While the former approach requires the measurement to be sufficiently clean, the latter is reported to have numerical stability issues. Since the parameters are estimated in both these approaches in an optimal sense, in the presence of a significant variation in parameters (due to damage), the estimates may often diverge. In this article, we propose an online health monitoring scheme powered by dual extended Kalman filtering technique to simultaneously estimate the system parameters along with the response states of a reduced-order system. To capacitate damage localization beyond sensor resolution, the proposed method employs location-based structural properties as system parameter.
Published in Journal for Mechanical Systems and Signal Processing
Estimation of local failure in tensegrity using Interacting Particle-Ensemble Kalman Filter
Tensegrities form a special case of truss, wherein compression members (struts/bars) float within a network of tension members (cables). Tensegrities are characterized by the presence of at least one infinitesimal mechanism stabilized with member pre-stress to ensure equilibrium. Over prolonged usage, the cables may lose their pre-stress while the bars may buckle, get damaged, or corrode, affecting the structural stiffness leading to change in the measured dynamic properties. Upon loading, a tensegrity structure may change its form through altering its member pre-stress affecting its global stiffness, even in the absence of damage.
Published in SHMII-10 2021–10th International Conference on Structural Health Monitoring of Intelligent Infrastructure
Robust Interacting Particle-Kalman Filter based structural damage estimation using dynamic strain measurements under non-stationary excitation -an experimental study
Sensor types and their positioning is a major factor in structural health monitoring (SHM) to ensure certainty in estimation. While acceleration has predominantly been employed for damage detection, they are known to be costly and not frame invariant (except for moderately accurate GPS based accelerometers). A thorough monitoring of a real life structure requires dense instrumentation which might become expensive with costly sensor types. Further, damages mostly occur at rare events, like seismic base excitation, for which typical accelerometers are not proper. This study employs strain as a cheaper alternative for damage sensitive measurement that is also frame invariant.
Published in Journal for Mechanical Systems and Signal Processing
Seismic-induced damage detection through parallel force and parameter estimation using an improved interacting Particle-Kalman filter
Standard filtering techniques for structural parameter estimation assume that the input force is either known or can be replicated using a known white Gaussian model. Unfortunately for structures subjected to seismic excitation, the input time history is unknown and also no previously known representative model is available. This invalidates the aforementioned idealization. To identify seismic induced damage in such structures using filtering techniques, force must therefore also be estimated. In this paper, the input force is considered to be an additional state that is estimated in parallel to the structural parameters. Two concurrent filters are employed for parameters and force respectively. For the parameters, an interacting Particle-Kalman filter is used to target systems with correlated noise.
Published in Journal of Hydrology
Uncertainty quantification using the particle filter for non-stationary hydrological frequency analysis
Recent changes in climate, anthropogenic activities and land-use patterns have significantly altered the hydrological cycle and thus led to the presence of non-stationarity in hydrological data series. Existing conventional approaches for hydrological frequency analysis (HFA) have commonly overlooked non-stationarity and consequently they might produce false estimates of hydrological design events. Assessing the effect of non-stationarity through uncertainty quantification is a potentially feasible approach for HFA. This paper proposed to incorporate the particle filter (PF) into HFA (here flood frequency analysis (FFA) was exemplified) (named PF-FFA) for quantifying prediction uncertainty in flood quantile estimates. The feasibility of the PF-FFA was verified through comparison with the conventional L-moment based FFA (LM-FFA) as well as the random sampling based FFA (RS-FFA) in terms of both accuracy and precision respectively using several selected evaluation indices.
Published in European Workshop on Structural Health Monitoring
Damage Detection in Tensegrity Using Interacting Particle-Ensemble Kalman Filter
Tensegrity structures form a special class of truss with dedicated cables and bars, that take tension and compression, respectively. To ensure equilibrium, the tensegrity members are required to be prestressed. Over prolonged usage, the cables may lose their prestress while bars may buckle, affecting the structural stiffness as well as its dynamic properties. The stiffness of tensegrities also vary with the load even in the absence of damage. This can potentially mask the effect of damage leading to a false impression of tensegrity health. This poses a major challenge in tensegrity health monitoring especially when the load is stochastic and unknown.