Unlock instant, AI-driven research and patent intelligence for your innovation.

A Nonlinear Dynamic Industrial Process Product Prediction Method Based on Spatiotemporal Attention Networks

A nonlinear dynamic and attention technology, applied in neural learning methods, biological neural network models, data processing applications, etc., can solve the problem of not fully considering the temporal dynamics of the correlation of key quality variables, and improve the accuracy sexual effect

Active Publication Date: 2022-06-24
CENT SOUTH UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the problem that the prior art does not fully consider the nonlinearity of the industrial process, the correlation between input variables and key quality variables, and the dynamics of the industrial process in time, and provides a method based on space and time attention. The method of long-short-term memory network to predict the product quality of industrial nonlinear dynamic process realizes the accurate prediction of key quality variables in the production process

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Nonlinear Dynamic Industrial Process Product Prediction Method Based on Spatiotemporal Attention Networks
  • A Nonlinear Dynamic Industrial Process Product Prediction Method Based on Spatiotemporal Attention Networks
  • A Nonlinear Dynamic Industrial Process Product Prediction Method Based on Spatiotemporal Attention Networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0158] as the flow chart figure 1 As shown, the initial boiling point of jet fuel in the hydrocracking process is predicted as follows:

[0159] Step (1), select 43 variables (as shown in Table 1) that have an impact on the initial boiling point of jet fuel from the hydrocracking process as input variables, and extract 536 off-line assays at 8 o'clock and 20 o'clock every day for 268 days. samples.

[0160] Step (2), standardize the dispersion of the data collected in step (1) to obtain a new data set, and the transformation function is:

[0161]

[0162] where x min is the minimum value of the dataset, x max is the maximum value of the dataset. Dispersion standardization is a linear transformation of the original data, so that the results fall into the [0,1] interval;

[0163] The first 450 samples are used as the training set to train the model parameters, and the remaining 86 samples are used as the test set to test the prediction performance of the model. First, th...

Embodiment 2

[0214] as the flow chart Figure 9 As shown, the C4 concentration of the debutanizer is predicted as follows:

[0215] In step (1), seven variables (shown in Table 3) that have an impact on the C4 concentration are selected from the debutanizer as input variables, and sampling is performed every 10 minutes to obtain a total of 1700 samples.

[0216] Step (2), standardize the dispersion of the data collected in step (1) to obtain a new data set, and the transformation function is:

[0217]

[0218] where x min is the minimum value of the dataset, x max is the maximum value of the dataset. Dispersion standardization is a linear transformation of the original data, so that the results fall into the [0,1] interval;

[0219] The first 1500 samples are used as the training set to train the model parameters, and the remaining 200 samples are used as the test set to test the prediction performance of the model. First, the input and output matrices of the training set are obtain...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention belongs to the field of industrial process control, and specifically relates to a method for predicting product quality in industrial nonlinear dynamic processes based on long-term and short-term memory networks based on space and time attention, which specifically includes the steps of: selecting key variables that have an impact on product quality, and The input variables and quality variables are continuously and uniformly sampled; standardize the deviation of the sampled original data set; determine the training set data; determine the parameters and hyperparameters of the network, and train the long-short-term memory network based on spatial and temporal attention; network to get the predicted values ​​of quality variables. The invention can not only adaptively select the input variable related to the quality variable, but also deal with the dynamic characteristics in the industrial process, and greatly improve the accuracy of the soft sensor model.

Description

technical field [0001] The invention relates to the field of industrial process prediction and control, in particular to a method for predicting the quality of industrial nonlinear dynamic process products based on a long-term and short-term memory network based on space and time attention. Background technique [0002] Soft sensing technology is widely used in modern industrial processes because of its rapid response and low maintenance cost. It monitors some key quality variables to predict key quality variables and achieve the purpose of ensuring product quality and production safety. [0003] The current soft measurement technology mainly includes principal component regression analysis, partial least squares regression, support vector regression, artificial neural network algorithm and so on. However, this kind of shallow network cannot mine the nonlinear features in the process data well, so the prediction performance is also limited. The proposal of deep neural netwo...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06N3/04G06N3/08G06Q10/06G06Q50/04
CPCG06N3/084G06Q10/06395G06Q50/04G06N3/044G06N3/045Y02P90/30
Inventor 袁小锋李林王雅琳阳春华桂卫华
Owner CENT SOUTH UNIV