The drum strength of sinter is one of the important indicators reflecting the quality of sinter during the sintering process, and its accurate prediction can improve the control accuracy and efficiency of the production process, and reduce the production cost and resource waste. However, in actual production, there are some difficulties in predicting the drum strength of sinter, such as limited data volume and poor data quality. Therefore, in order to improve the prediction accuracy, the Generative Adversarial Network (GAN) is used to augment the original dataset to solve the problem of limited data volume. Then, the Echo State Network (ESN) optimized by Sparrow Search Algorithm (SSA) is used to construct the prediction model. Compared with traditional neural networks, ESN has better stability and generalization ability, and can quickly train and adapt to new data. The prediction accuracy and efficiency of the model are verified by experiments, and compared with other prediction algorithms. The results show that the expanded dataset and ESN model can significantly improve the prediction accuracy, and the mean absolute percentage error is reduced from 1.41% to 1.06%.