A New Structural Health Monitoring Strategy Based on PZT Sensors and Convolutional Neural Network

Abstract

Preliminaries convolutional neural network (CNN) applications have recently emerged in structural health monitoring (SHM) systems focusing mostly on vibration analysis. However, the SHM literature shows clearly that there is a lack of application regarding the combination of PZT-(lead zirconate titanate) based method and CNN. Likewise, applications using CNN along with the electromechanical impedance (EMI) technique applied to SHM systems are rare. To encourage this combination, an innovative SHM solution through the combination of the EMI-PZT and CNN is presented here. To accomplish this, the EMI signature is split into several parts followed by computing the Euclidean distances among them to form a RGB (red, green and blue) frame. As a result, we introduce a dataset formed from the EMI-PZT signals of 720 frames, encompassing a total of four types of structural conditions for each PZT. In a case study, the CNN-based method was experimentally evaluated using three PZTs glued onto an aluminum plate. The results reveal an effective pattern classification; yielding a 100% hit rate which outperforms other SHM approaches. Furthermore, the method needs only a small dataset for training the CNN, providing several advantages for industrial applications.

Keywords: CNN; SHM; deep learning; electromechanical impedance; intelligent fault diagnosis; machine learning; piezoelectricity.

Figure 1

A general architecture for the CNN highlighting the layers.

Figure 2

Developed framework for structural damage detection, based on the CNN algorithm, including all three phases.

Figure 3

Representation of the general diagram for the acquisition system (dimensions in millimeters) [27].

Figure 4

Experimental set up including: aluminum plate containing three PZT patches, DAQ (Data Acquisition) and computer running the acquisition software [27].

Figure 5

Block diagram for generating a RGB frame.

Figure 6

Division of the EMI signals for the baseline (top) and unknown (bottom) structural conditions before computing ED.

Figure 7

ED-matrix formed after computing ED from the EMI signatures.

Figure 8

Obtained frame from two random PZT-EMI signatures.

Figure 9

Architecture of the developed CNN to identify structural damage.

Figure 10

Real part of the EMI, for PZT#2, considering various structural conditions (H, D1, D2 and D3).

Figure 11

Set of frames formed from the EMI signatures for PZT#2: (a) baseline with Healthy (H); (b) baseline with damage 1 (D1); (c) baseline with damage 2 (D2); (d) baseline with damage 3 (D3).

Figure 12

Feature maps for the 1st CNN layer after applying 32 kernels into PZT#2 frames for the structural conditions: (a) D1; (b) H.

Figure 13

Feature maps for the 7th CNN layer after applying 64 kernels into PZT#2 frames for the structural conditions: (a) Healthy (H); (b) D1; (c) D2; (d) D3.

Figure 14

Performance analysis of the CNN for PZT#2: (a) training and validation accuracy curve of the model as a function of epoch; (b) training and validation loss curve of the model as a function of epoch; (c) Consumption time versus number of epoch for the training phase.