A Dual-Attention Recurrent Neural Network Method for Deep Cone Thickener Underflow Concentration Prediction

Abstract

This paper focuses on the time series prediction problem for underflow concentration of deep cone thickener. It is commonly used in the industrial sedimentation process. In this paper, we introduce a dual attention neural network method to model both spatial and temporal features of the data collected from multiple sensors in the thickener to predict underflow concentration. The concentration is the key factor for future mining process. This model includes encoder and decoder. Their function is to capture spatial and temporal importance separately from input data, and output more accurate prediction. We also consider the domain knowledge in modeling process. Several supplementary constructed features are examined to enhance the final prediction accuracy in addition to the raw data from sensors. To test the feasibility and efficiency of this method, we select an industrial case based on Industrial Internet of Things (IIoT). This Tailings Thickener is from FLSmidth with multiple sensors. The comparative results support this method has favorable prediction accuracy, which is more than 10% lower than other time series prediction models in some common error indices. We also try to interpret our method with additional ablation experiments for different features and attention mechanisms. By employing mean absolute error index to evaluate the models, experimental result reports that enhanced features and dual-attention modules reduce error of fitting ~5% and ~11%, respectively.

Pruning and Quantization for Deep Neural Network Acceleration: A Survey

“…Convolutional neural networks are a class of feed-forward DNNs that use convolution operations to extract features from a data source. CNNs have been most successfully applied to visual-related tasks however they have found use in natural language processing [89], speech recognition [2], recommendation systems [204], malware detection [213] and industrial sensors time series prediction [251]. To provide a better understanding of optimization techniques, in this section, we introduce the two phases of CNN deployment -training and inference, discuss types of convolution operations, describe batch normalization (BN) as an acceleration technique for training, describe pooling as a technique to reduce complexity, and describe the exponential growth in parameters deployed in modern network structures.…”

Forecasting Copper Electrorefining Cathode Rejection by Means of Recurrent Neural Networks With Attention Mechanism

“…Finally, the Encoder-Decoder architecture has proven to be successful in operational regression settings, such as forecasting operational indicators for thickening processes [20], [21]. However, the value of attention weights as information sources has not been established.…”