Assessment of Geometrical Features of Internal Flaws with Artificial Neural Network

摘要: In nondestructive testing (NDT), geometrical features of a flaw embedded in the material such as its location, length, and orientation are critical factors to assess severity make post-manufacturing decisions improve design. this study, artificial intelligence (AI) based NDT approach was applied ultrasonic oscillograms obtained from virtual estimate flaw. First, numerical model specimen constructed using acoustic finite element analysis (FEA) produce signals. The validated by comparing simulated signals produced with experimental data actual tests. Then, 750 models containing flaws different locations, lengths, angles were generated FEA. Next, divided into 3 datasets: 525 for training, 113 validation, 112 testing. Training inputs network parameters extracted fitting them sine functions. Lastly, evaluate performance, outputs including flaw’s angle compared desired values all datasets. Deviations calculated regression analysis. Statistical also performed measuring root mean square error (RMSE) efficiency. RMSE x-location, y-location, estimations 0.09 mm, 0.19 mm, 0.46 mm, 0.75°, efficiencies 0.9229, 0.9466, 0.9140, 0.9154, respectively dataset. Results suggest that proposed AI-based method has potential interpret material.