A layered working condition perception integrating handcrafted with deep features for froth flotation

To effectively recognise the working condition in froth flotation, many existing methods separately focus on handcrafted features or deep features. Considering that deep features complement handcrafted features with more detailed information, a layered working condition perception method integrating handcrafted features with deep features is developed for zinc flotation. In the proposed method, a fuzzy algorithm layer (FAL) with handcrafted features and a convolutional perception layer (CPL) with deep features are constructed for working condition recognition. Given that similar handcrafted features exist in the boundary between adjacent working conditions, the handcrafted feature-based FAL is limited. Therefore, the layered evaluation agency (LEA) is established to determine whether the working condition needs to be reidentified. If needed, LEA obtains the information of two possible working conditions, and the CPL is applied to reidentify the working condition based on deep features extracted by VGG16 and the support vector machine (SVM) with the specific category designated by LEA. The effectiveness of the proposed framework was evaluated through experiments, and the results confirm the potentiality of the proposed method in working condition recognition.

Automated summary

The method integrates handcrafted features and deep features to recognize the working condition in zinc flotation; a layered evaluation agency is established to determine if reidentification is needed, and deep features with support vector machine are applied for reidentification under specific working conditions.