Multi-time-scale Convolutional Neural Network Soft-Sensing Method Based on Attention Mechanism

Abstract

The invention relates to an attention mechanism-based multi-time-scale convolutional neural network soft-sensing method and the field of soft-sensing technology. Including the following steps: 1. Determine auxiliary variables and perform data processing, select easy-to-measure variables related to difficult-to-measure parameters as auxiliary variables of the soft sensor model and collect time series of auxiliary variables and difficult-to-measure parameters; and then analyze the collected time series Elimination of outliers; 2. The attention mechanism and the selection of the attention area, according to the delay and effective time scale of each auxiliary variable relative to the difficult-to-measure parameter, the attention area is divided; 3. The input of the construction of the soft sensor model, the auxiliary variables The matrix is ​​composed of the time series, and the input of the soft sensor model is determined in combination with the attention area of ​​the attention mechanism; 4. Establish the soft sensor model of the sequential convolutional neural network; 5. Train the soft sensor model of the sequential convolutional neural network; 6. Use step 5 The trained time-series convolutional neural network model estimates difficult parameters in real time.

Description

technical field [0001] The invention relates to a multi-time-scale convolutional neural network soft-sensing method based on an attention mechanism, and belongs to the field of soft-sensing technology. Background technique [0002] In the modern industrial production process, in order to realize energy saving and benefit maximization, it is of great significance to monitor and control the important parameters in the production process in time. Usually, for important parameters in the industrial production process, there are mainly two measurement methods: on-line measurement and off-line measurement. On-line measurement refers to the use of instruments to directly measure parameters, but the equipment is expensive, difficult to maintain, and the accuracy of measurement results is easily affected by on-site working conditions. Off-line measurement refers to the measurement of parameters by the method of offline inspection, but offline inspection often takes a long time, resu...

AI Technical Summary

Problems solved by technology

On-line measurement refers to the use of instruments to directly measure parameters, but the equipment is expensive, difficult to maintain, and the accuracy of measurement results is easily affected by on-site conditions

Offline measurement refers to the measurement of parameters by the method of offline inspection, but offline inspection often takes a long time, resulting in a large delay in the guidance of the production process by the measurement results obtained offline

Images