Big Data Analytics in Uncertainty Quantification: Application to Structural Diagnosis and Prognosis

Abstract

This study investigates the use of big data analytics in uncertainty quantification and applies the proposed framework to structural diagnosis and prognosis. With smart sensor technology making progress and low-cost online monitoring becoming increasingly possible, large quantities of data can be acquired during monitoring, thus exceeding the capacity of traditional data analytics techniques. The authors explore a software application technique to parallelize data analytics and efficiently handle the high volume, velocity, and variety of sensor data. Next, both forward and inverse problems in uncertainty quantification are investigated with this efficient computational approach. The authors use Bayesian methods for the inverse problem of diagnosis and parallelize numerical integration techniques such as Markov-chain Monte Carlo simulation and particle filter. To predict damage growth and the structure’s remaining useful life (forward problem), Monte Carlo simulation is used to propagate the uncertainties (both aleatory and epistemic) to the future state. The software approach is again applied to drive the parallelization of multiple finite-element analysis (FEA) runs, thus greatly saving on the computational cost. The proposed techniques are illustrated for the efficient diagnosis and prognosis of alkali-silica reactions in a concrete structure.