Defects Clustering for Mineral Sintering Surface Based on Multi-source Data Fusion

During sintering of minerals, multiple categories of defects often appear on the surface of material due to hysteresis of control. Fusion of data from multiple sources is considered in this paper in a bid to identify different types of defects and to improve the utilization of collected data. Specifically, this paper establishes a heterogeneous multi-source dataset based on sintering data. And a novel deep learning approach for multi-source data fusion is introduced and extended. Based on constructed multi-source dataset and associated algorithm, the downstream task of clustering is carried out. Affiliated experimental results show that the multi-source information clustering process for mineral sintered surface defects proposed in this paper outperforms unimodal approach, and relative performance metrics have reached the level of engineering applications.