Intro

Vertically Integrated Project Group

The Internet of Things

as a Methodology for Improving Diagnostics & Prognostics

in Structural Health Monitoring Applications

Managing Professor: Antonios Kontsos antonios.kontsos@drexel.edu

VIP IoT Subgroups
Hardware & Software		Machine Learning/Algorithms
Name	Email	Name	Email
Rakeen Rouf	Rmr327@drexel.edu	Krzysztof Mazur	Kwm38@drexel.edu
Sadman Jafir Islam	Sji29@drexel.edu	Matthew Rantz	Mjr445@drexel.edu
Ishtiaq Shahriar	Is353@drexel.edu	Sarah Malik	Sm3658@drexel.edu
Har Patel	hp377@drexel.edu	Mohammad Adib	mma343@drexel.edu
Abrar Zawad	kz335@drexel.edu	Raj Patel	Rap342@drexel.edu

Project Abstract

Structural Health Monitoring (SHM), defined as the process that involves sensing, computing and decision making to assess the integrity of infrastructure has been plagued by data management and data-driven decision-making challenges. The Internet of Things (IoT) provides a way to decisively address both SHM’s big data problems. The purpose of the IoT project proposed herein is to develop a framework that connects sensor data with processing and modeling to provide diagnostic/prognostic capabilities related to SHM. Specifically, the proposed IoT model will be comprised of 3 components: the Edge, the Fog and the Cloud. The Edge is the bottom level hardware that interfaces with the sensors used and is capable of filtering and transmitting data. The Fog is the hardware that provides an environment where computational algorithms for both diagnostics and prognostics can be implemented leveraging cloud-like properties. The Cloud is used to store data as well as to perform demanding computations such that required large databases and training sets. To accomplish this goal a team of three undergraduate students in majors including mechanical, electrical and computer engineering, as well as computer science and business was formed. The team will leverage existing sensing, computing and data post-processing capabilities at the lab using an actual test case to demonstrate the performance of the proposed IoT framework. 

Data Trends

The Data Trends above exhibit the first major case for an IoT framework. In DIC, it shows some of the 15 different features, in IRT we have 5 different features and in AE we have up to 35 different features. These features come in during tests of up to 10,000 cycles at a variety of acquisition rates. All in all, the “big data problem” of SHM is seen at a small scale. The issue comes with the fact that in aerospace cases any single hit can be an indication of major damage propagation.

General IoT Framework

IoT Data Flow Framework

The focus of the IoT framework in terms of SHM is information management and consolidation performed live by separating workload into varying levels of computational expense and varying levels of informational load. For each structure, a data history in terms of NDT methods and sensors would be required initially. This data history can be simulated as fatigue data obtained from Acoustic Emission (AE), Infrared Thermography (IRT), Digital Image Correlation (DIC) and a load curve obtained from the testing machine itself. This information is trimmed to a manageable level using feature selection and feature reduction techniques, more specifically, correlations between features are compared so that the least correlated features are kept, and then Principal Component Analysis (PCA) is used to further diminish the number of features used.