• Basuraj Bhowmik

My talk at EMI-PMC 2021 (Part 1/2)

Modal identification and damage detection – A contemporary real-time approach

Bhowmik_EMI2021
.pdf
Download PDF • 2.69MB

A contemporary approach directed towards key understanding of output response obtained from vibrating systems have led to a class of real-time algorithms based on eigen perturbation approaches (EPA). These algorithms have shown to be less computationally expensive in comparison to other online approaches as these methods utilize only the eigenspace updates with streaming data in real-time.

188 million trips are made across structurally deficient bridges every day in the USA alone. ASCE’s infrastructure report card grades this as a C- ‘average’. Maximization of the useful life of our built infrastructure has its roots firmly embedded in structural health monitoring (SHM).


My talk at the ASCE EMI-PMC showcased an attempt at bridging the gap between sensors and infrastructure. Through this talk - and the Q&A session afterward - the need for online monitoring and maintenance emerged as the need of the hour. While countless researchers have developed sophisticated techniques for the construction of infrastructure, adequate efforts devoted to their maintenance still remain scarce and dispersed. In this context, it becomes essential to understand SHM with the simplest of technicalities.


SHM comprises 5 key steps: First, a dense array of sensors is installed on the system to be monitored. Typically, accelerometers and strain gauges are used here. Next, the acquired readings from these sensors are used to detect the presence of system damage. Once the damage is detected, the location of the damage is then accurately pointed out to the asset monitoring managers which enables quick retrofitting installations. Additionally, the severity of the damage is evaluated, which provides a probabilistic estimate of the remaining useful life of the system.


Now, let us look at the problem. Structures degrade and get overloaded from environmental exposure and increased demands. Decisions based on personal judgment can overestimate damages, leading to unnecessary repair. This is a poor investment of your tax money. Underestimating damages, on the other hand, can cost your life. Real-time eigen perturbation approaches modify traditional algorithms to real-time strategies using a family of first-order eigen perturbation approaches. This ensures early prediction of damage and provides a cheap, hassle-free remote monitoring strategy that has gained prominence in recent times.


When you download this presentation, you can identify a slide with a vehicle traversing over a bridge. The vehicular response streaming in real-time can be transformed using EPA method where the canonical correlation between the basis vectors is maximized. Using damage-sensitive features on the eigenspace updates, the recursive canonical correlation analysis (RCCA) algorithm provides damage event identification in real-time. In the interest of this presentation, the results will be discussed from a point of view of the RCCA algorithm.


In the next part of this blog series (and the presentation in turn), I will talk about the numerical simulations and experimental testbeds devised specifically for the implementation of EPA-based RCCA. Successful verifications have led to the materialization of this work and this presentation received a warm round of favorable reviews and accolades among the speakers.


The following conclusions can now be drawn from the first part of the blog:

  1. Real-time solutions for detecting the presence of system damage are dispersed and an effort is made to bring the work within the ambit of fundamental mechanics.

  2. The use of RCCA has been documented as a case-specific algorithm for built infrastructure monitoring. Essentially, the block covariance matrices of the output response are first formed and a generalized Eigen decomposition is carried out.

  3. Each decomposition set produces eigenspace (i.e., the eigenvectors and eigenvalues) that is updated at each instant of time, instead of updating the covariance matrices. This significantly reduces the computational expense associated with real-time monitoring solutions provided by EPA.


Stay connected for the next part where I'll be providing my thoughts on the real-time analysis using EPA. I promise it to be a more interesting and engaging read for you!




Copyright: Basuraj Bhowmik 2021

95 views0 comments