RSBO based LSTM network for pulsed eddy current steel wall-thinning inspection

Corrosion-induced wall thinning in steel pressure vessels presents a critical threat to the safety and reliability of industrial operations, requiring effective inspection methods. In this paper, the random searching Bayesian optimization (RSBO) based long-and short-term memory (LSTM) network is proposed to estimate the wall thinning. To improve computational efficiency and reduce data redundancy, data augmentation is firstly conducted by selecting the sample interval with more features. Then the RSBO based LSTM model is designed to estimate the thickness. With the designed estimator, it can automatically tune the hyperparameters for the LSTM network which is friendly for the users. Experiments have been carried out and, from the experimental results, the proposed model has a better performance and can achieve the thickness estimation with the average MAPE of 5.54% and MAE of 1.98 mm.