Identification of abnormal conditions for gold flotation process based on multivariate information fusion and double‐channel convolutional neural network

Abstract Accurate and in‐time working condition identification plays a great role in industrial processes. However, most of the current flotation process identification models only use the characteristics of flotation froth as an identification basis, which often causes identification errors due to the instability of the froth, and a large amount of process data is not fully utilized. In this paper, an abnormal condition identification method based on multivariate information fusion and double‐channel convolutional neural network (double‐channel CNN) is proposed to achieve higher accuracy. First, a double‐channel CNN is used to extract depth features from different distributions of froth images and process data in parallel. Then, double normalized attention mechanism (double normalized AM) and multivariate information fusion methods are used to attach weights to the features and fuse them so as to ensure a higher response of key features and increase the reliability of the identification model. The method shows better performance than existing methods in offline simulations and has been validated online at a mineral processing plant in Shandong.