Research on the Strain Prediction Method of High-Temperature Pressure Vessel Based on Finite Element and Deep Learning

Abstract

The stress and strain of metals under high-temperature operation have significant implications for the optimal design and safety assessment of high-temperature pressure vessels. However, it is challenging to efficiently evaluate the stress and strain in critical parts of high-temperature pressure vessels. To address these issues, a method combining deep learning and finite element analysis is proposed to predict the total strain of high-temperature pressure vessels. Firstly, finite element simulations are conducted under multiple working conditions and sizes to provide data support for deep learning prediction. Secondly, the WOA algorithm is used to optimize the number of hidden layer neurons and learning rate for LSTM, and the optimized hyperparameters are then assigned to the LSTM network to construct a prediction model that matches the data features more effectively.  Finally, the data obtained from the finite element simulations are fed into the prediction model, and the results show that the prediction accuracy reaches 98.6%, demonstrating good predictive performance. This achievement can provide efficient data support for the optimal design and safety assessment of high-temperature pressure vessels.