Abstract: The prediction of mechanical properties of hot rolled strips has a very broad application prospect. However, the current mechanical performance prediction model is not accurate and the generalization ability is poor, and the accuracy of the prediction results cannot be evaluated, which limits its effectiveness in practical applications. In order to improve the accuracy of the mechanical performance prediction model and the reliability evaluation of the prediction results, BP neural network is used to transform the current modeling method of direct prediction of mechanical performance results into the modeling of mechanical property deviation distribution prediction between samples, and combined with the model. The degree of dispersion of the prediction results is designed to evaluate the reliability evaluation index, and ensemble learning is used to improve the generalization ability of the model. Through experimental verification, the ensemble learning model has higher prediction accuracy. Further, according to the reliability evaluation index analysis, except for the prediction results of about 3.5% of the samples, there is greater uncertainty, and the yield strength and tensile strength of the remaining samples. The prediction accuracy of the intensity ±30 MPa is 98.45% and 98.97%, and the prediction accuracy of the elongation under the error of ±5% is 99.48%, which effectively improves the accuracy of model prediction and has certain performance in production field applications. Guiding significance.

Key words: deviation prediction, mechanical performance prediction, neural network, ensemble learning, reliability evaluation