Small Sample-Oriented Prediction Method of Mechanical Properties for Hot Rolled Strip Steel Based on Model Independent Element Learning

In this paper, a new method based on Model-Agnostic Meta-Learning (MAML) is proposed to address the small sample problem in predicting the mechanical properties of hot rolled strip steel. Traditional prediction models rely on large amounts of data, and when data is limited, the prediction accuracy and generalization ability are insufficient. By training on multiple related tasks, MAML can quickly adapt to new tasks with limited data, making it suitable for dealing with small sample problems.In this study, the chemical composition, processing parameters, and mechanical properties of hot rolled strip steel were collected, cleaned, and preprocessed. Linear regression, BP neural network, LASSO regression, Ridge regression, and convolutional neural network models were trained using the MAML algorithm. The experimental results show that the prediction accuracy and adaptability of models with the MAML algorithm are significantly better than traditional methods, especially in terms of rapid adjustment to maintain prediction accuracy when production conditions change. This paper verifies the effectiveness of MAML in industrial forecasting and provides a new approach and method for forecasting the production processes of other materials.