Research on Prediction of Bearing Performance of Bottom Water Seal Based on Neural Network

Artificial neural networks have made great progress in the field of engineering. Due to the stress relaxation characteristics of the material and the large deformation characteristics at over-limit compression, the mechanical characteristics of rubber water seals under load cannot be clarified. This may easily lead to seal failure, which can lead to engineering diseases. A neural network prediction model was constructed to evaluate the bearing performance of the bottom water seal. This neural network prediction model is trained by learning training samples, which are obtained through load tests. The prediction model can be used to analyze the complex relationship between the length and compression of the bottom water seal and the loaded reaction force, which enables the prediction of the bottom water seal's bearing capacity and ensures the sealing effect. Compared with the conventional fitting method, the neural network prediction model has better discrimination accuracy. The minimum mean square error (MSE) of the prediction model is 0.094, and the R2 value is 99.99%, which can provide an important technical guarantee for the reliable operation of hydropower projects.