Health Indicators Construction and Remaining Useful Life Estimation for Concrete Structures Using Deep Neural Networks

Abstract

Remaining useful life (RUL) prognosis is one of the most important techniques in concrete structure health management. This technique evaluates the concrete structure strength through determining the advent of failure, which is very helpful to reduce maintenance costs and extend structure life. Degradation information with the capability of reflecting structure health can be considered as a principal factor to achieve better prognosis performance. In traditional data-driven RUL prognosis, there are drawbacks in which features are manually extracted and threshold is defined to mark the specimen’s breakdown. To overcome these limitations, this paper presents an innovative SAE-DNN structure capable of automatic health indicator (HI) construction from raw signals. HI curves constructed by SAE-DNN have much better fitness metrics than HI curves constructed from statistical parameters such as RMS, Kurtosis, Sknewness, etc. In the next stage, HI curves constructed from training degradation data are then used to train a long short-term memory recurrent neural network (LSTM-RNN). The LSTM-RNN is utilized as a RUL predictor since its special gates allow it to learn long-term dependencies even when the training data is limited. Model construction, verification, and comparison are performed on experimental reinforced concrete (RC) beam data. Experimental results indicates that LSTM-RNN generally estimates more accurate RULs of concrete beams than GRU-RNN and simple RNN with the average prediction error cycles was less than half compared to those of the simple RNN.

A Deep-Learning-Based Health Indicator Constructor Using Kullback–Leibler Divergence for Predicting the Remaining Useful Life of Concrete Structures

“…Due to their lower complexity and wider application toward different systems and faults, data-based methods have been more favorable in recent years, especially with the rise of artificial intelligence [ 43 ]. Notable mentions in this category are statistical projection [ 39 , 44 , 45 ], deep learning models [ 12 , 46 , 47 , 48 ], and evolutionary computation [ 33 , 49 ], etc. Different studies following these methods have achieved promising results in HI construction for the RUL prognosis task.…”

“…Concerning the HI construction for RUL prognosis, the evaluation can be generally divided into two categories: the investigation of the HI’s intrinsic nature from the construction result (fitness analysis) and the performance of HI in the RUL prognosis tasks. Fitness analysis is often performed with the following metrics: monotonicity [ 12 , 41 ] (measurement of the monotonic trend in HI), trendability [ 12 , 41 ] (the correlation of HI and time), and scale similarity [ 12 , 41 , 48 ] (similarity of HI ranges), etc. The purpose of this type of evaluation is to self-reflect the HI properties via low computational complexity without concern of the prognosis task.…”

“…By predicting the future state of the structure, a user can approximate the time when a failure might occur, thus adjusting the usage and preparing a maintenance plan accordingly in advance. The most recent studies [ 11 , 12 , 13 , 14 , 15 , 16 ] regarding this topic focus on the development of an autonomous system that is capable of extracting feature(s) from data and constructing indicators based on these extracted features. A common scheme for this approach is as follows: initially, sensors are deployed, which are then utilized to collect and send the real-time data to the central brain where data processing and health indicator construction occur.…”

“…As nondestructive methods for SHM have become dominant in recent years, different methods have been investigated for concrete SHM such as ultrasonic [ 5 , 17 , 18 ], vibration [ 19 , 20 ], image processing [ 3 , 21 , 22 ], and acoustic emission (AE) [ 4 , 12 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 ]. Even though ultrasonic testing can detect internal defects and their sizes, it is susceptible to complicated part geometry and certain materials, such as austenitic steel, which can mask defects by causing attenuation.…”

Deep learning-based prediction for time-dependent chloride penetration in concrete exposed to coastal environment

“…In recent years, with advancements in neural network technology, deep learning has emerged as a research hotspot in various industries. In the field of civil engineering, its applications primarily focus on damage identification and prediction of deformation and settlement [ [17] , [18] , [19] , [20] , [21] , [22] , [23] , [24] ]. For instance, Yang et al [ 21 ] utilized both statistical and deep learning models to predict deformation in a concrete dam based on actual detection data.…”