Title: Machine Learning Based Concrete Property Monitoring Through Nano-sensors

Author(s): Yenfang Su

Publication: Web Session

Volume:

Issue:

Appears on pages(s):

Keywords:

DOI:

Date: 4/1/2021

Abstract:
Electromechanical impedance (EMI) method coupled with piezoelectric nano-sensor is proved as a promising technology for assessing the concrete strength. Based on the unique piezoelectric effect, the EMI signature of the piezoelectric nano-sensor can catch the properties changes of concrete during the hydration process. In this study, large-scale slab (8-feet by 12-feet) experiments were conducted to monitoring the in-situ mechanical property of concrete. The factors, including mixture design, ambient temperature, and humidity, have been taken into consideration to develop a dataset for training the machine learning model. The EMI signal of the slab was recorded at a very early age from 6th hour to 12th hour with the 2 hours’ interval, and at an early age on the 1st, 3rd, 7th day. We compared the prediction performance for various machine learning algorithms, including artificial neuron network (ANN) and convolutional neuron network (CNN), with different network architectures. The result indicates that the established prediction function via a machine learning approach can effectively predict the concrete strength. Consequently, the machine-learning based EMI method can be used as potential non-destructive testing (NDT) method for in-situ monitoring of concrete compressive strength gain at a very early age.