Subsurface imaging of concrete structures using neural network approach

Abstract

A novel artificial neural network (ANN) based approach for the microwave subsurface imaging of reinforced concrete structures is proposed. The proposed technique facilitates the detection of the inner configuration of test structures, and is based on measurement of reflection data using a Ka-band waveguide (WR-28) along with the network analyzer. The waveguide is directly placed in contact with the test structure, and the whole sample is scanned by moving the waveguide holder along its surface in order to measure the reflection data at various positions. The training data for the ANN is generated by simulating the complete measurement setup in the CST Microwave Studio with a typical concrete specimen. The actual measured reflection data is then fed to the previously trained ANN to produce the subsurface image of the test structure. The proposed system is validated by imaging different concrete samples using both simulated and experimental data.