Operating Performance Assessment Based on Fusion Model of Multi-Source Heterogeneous Information for the Electro-Fused Magnesium Furnace

For the electro-fused magnesium furnace (EFMF), accurate and effective process operating performance assessment (POPA) is essential to achieve satisfactory comprehensive economic benefits. However, most existing assessment methods rely solely on a single type of data in smelting process of EFMF, ignoring the importance of multi-source heterogeneous information (MSHI) generated with the production. In order to take full advantage of the MSHI, this article presents a novel POPA method based on the fusion model of MSHI (FMMSHI) for the EFMF. First, three assessment sub-models based on the current, sound and image are pre-trained with sparse auto-encoder (SAE), gate recurrent unit (GRU), and convolutional neural networks (CNN), respectively. Considering the fact that MSHI can reflect the process operating performance from multiple perspectives, a comprehensive assessment model which fused the assessment results of three sub-models is developed to obtain the final assessment result. The experiment results show that the proposed FMMSHI-based assessment method has the highest accuracy of all compared methods.