Flexible Piezoelectric PZT Actuator/Sensor for Damage Classification of Rail Structures Based on Convolutional Neural Network

As the railway comes to the intelligent maintenance stage from the massive construction stage in China, new sensors and data fusion algorithms are required for structural health monitoring (SHM) and damage detection of rail structures. Flexible piezoelectric PZT actuators/sensors (FPAS) laminated onto the rail waist are designed for damage identification using the active detection method based on Lamb guided waves. Finite element modeling and experimental results show that the feature data of the echo signals from FPAS vary with different damage sizes or shapes, which is effective for damage classification and recognition of the rail. Convolutional Neural Network is adopted for damage recognition of rail structures based on FPAS with high accuracy (> 90%), which has great potential in field tests of rail structures. FPAS with intelligent algorithms will be a research direction for SHM and damage recognition, improving the performance of rail structures.