The drum strength and total iron content of sinter are important indicators for evaluating sinter quality. Traditional prediction methods often rely on empirical formulas or single-feature-based models,making it difficult to effectively address nonlinearity and time-varying characteristics during sintering process,resulting in limited prediction accuracy. Based on the operating parameters and mixed ingredients of sinter,signal processing and deep learning technology are used to construct an EEMD-EE-LSTM dual-index quality prediction model:the ensemble empirical mode decomposition (EEMD) and the high-information value IMF selection based on energy entropy are introduced,and a long-term short-term memory (LSTM) network with dual output heads is constructed,and the correlation and sensitivity analysis are used. The cooperative prediction of drum strength (TS) and total iron content (TFe) is realized by shared feature extraction and joint loss function. The verification of the actual production data of iron and steel enterprises shows that the accuracy of the proposed method in the prediction of drum strength is as high as 92. 6%,and the accuracy rate in predicting the total iron content reaches as high as 94. 3%. This research not only provides theoretical support for the collaborative prediction of multiple quality indicators during the sintering process,and establishes a complete technical framework that encompasses “signal denoising, and feature optimization, and multi-indicator output modeling”, but also provides generalizable engineering solutions for the intelligent optimization and quality control of complex metallurgical processes.