To investigate the failure behavior of composite open-hole plates under tensile loads, utilizing a high-precision finite element simulation model established based on tensile tests of open-hole plates, a dataset of tensile force-displacement curves was generated in batches. Then, a dual long short-term memory (LSTM) neural network model was proposed to predict the force-displacement curve. The first LSTM model is responsible for extracting input features, while the second LSTM model directly predicts the force-displacement curve. The research results indicate that this model can efficiently and accurately predict the tensile force-displacement curves of open-hole plates. The coefficient of determination (R2) on the test set reaches as high as 0.9755, with the prediction error of key features such as the initial stiffness (E0) being only 1.85% and the prediction error of the maximum load (Fmax) being only 2.16%.