Ultrasonic signal classification and porosity testing for CFRP materials via artificial neural network

Ultrasonic testing is one of the most commonly used non-destructive (NDT) methods to detect porosity in carbon fiber reinforced polymer (CFRP) material. However, the ultrasonic testing requires well-trained technicians and their high concentration during the testing process to avoid human errors. The artificial neural network (ANN) models provide an efficient and accurate way to reduce testing time and effort and reduce the human factor and uncertainty during ultrasonic testing. In this work, CFRP samples with 12 different kinds of porosity levels and thicknesses were prepared for the experiment. X-ray CT testing and through-transmission ultrasonic (TTU) testing are applied to measure the material porosity. Based on the porosity data obtained by these two testing methods, a backpropagation (BP) neural network was developed and trained to analyze and predict the sample porosity. In our experiment, the accuracy of our ANN-based method can reach up to 97.22%, and the average prediction result is 88.02%. Our experimental results demonstrate that our new method is quite promising to predict porosity testing results in CFRP materials.