Mixed modeling method and system based on combination of process priors and data-driven model

A data-driven model, combined with process technology, applied in electrical digital data processing, special data processing applications, instruments, etc., can solve problems such as simple model structure, high constraints on modeling objects, and discarding useful information.

Inactive Publication Date: 2015-09-16
EAST CHINA UNIV OF SCI & TECH
View PDF4 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Some scholars have also proposed data-driven modeling methods that integrate process mechanisms (such as proxy model methods with mechanisms, neural network models for output gain detection, modeling methods for combined probability density estimation, and monotonic neural networks), but comprehensive modeling method

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
  • Mixed modeling method and system based on combination of process priors and data-driven model
  • Mixed modeling method and system based on combination of process priors and data-driven model
  • Mixed modeling method and system based on combination of process priors and data-driven model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0087] see image 3 , the present invention proposes a hybrid modeling method combining process priors and data-driven models, the specific steps are as follows:

[0088] [Step S1] Select a suitable data-driven model and a suitable model structure from known data-driven models, establish a mathematical relational expression of the corresponding model, and arrange all model parameters in a certain order.

[0089] The model mathematical relationship expression of step S1 Arrange par with model parameters: select a data-driven model, generally we often choose BP neural network model, response surface model, support vector machine, etc.

[0090] For the BP neural network model, its model input and output relationship is:

[0091] y ^ l ( p a r , X ) = f h ...

Embodiment 2

[0146] The description of the following examples will help to understand the present invention, but does not limit the content of the present invention. see figure 2 , this embodiment realizes the estimation of the average particle size of the nanofiltration membrane sol based on the hydrolysis temperature, the amount of glycerin added and the amount of complexing agent added, and the influence of input variables on the particle size of the sol is shown in Table 1. The nanofiltration membrane sol preparation process studied in this example is a steady state process, and the process data are collected from 46 groups of samples obtained in the experiment (sol particle size stability performance data acquisition period is long (about one month), and a large number of samples cannot be obtained). Other variable parameters are as follows: the molar ratio of the precursor is Zr:Ti=4, the molar ratio of the precursor to water is 1:555, the sol concentration is 0.1mol / L, and the comp...

Embodiment 3

[0159] The description of the following examples will help to understand the present invention, but does not limit the content of the present invention. The analysis object of this example is a reactor of the acetylene hydrogenation reactor of ethylene production unit (such as image 3 colored circle). In this paper, the feed rate of C2 fraction, the feed rate of hydrogen and the temperature difference between the inlet and outlet of the first-stage reaction tank are selected as the easy-to-measure (independent variables) of the soft sensor, and the soft sensor model is established to estimate the concentration of acetylene at the outlet of the first-stage reaction tank.

[0160] The acetylene soft measurement of a section of the reaction tank studied in this example, the process data are collected from 40 groups of samples obtained from the experiment. Divide 40 groups of samples into 2 parts, randomly select 30 groups as training samples, and 10 groups as test samples. Sel...

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 present invention discloses a mixed modeling method and system based on the combination of process priors and a data-driven model. The method comprises: selecting a data-driven model and an appropriate model structure from known data-driven models so as to establish a mathematic relation expression corresponding to the data-driven model, and arraying all model parameters in a certain order; verifying the process priors of the model so as to obtain a constraint equation of the degree of detected model against the process priors; comparing the model output of the sample with observed values, and establishing an optimal target equation for detecting the degree of the model fitting the training sample; combining the constraint equation with the optimal target equation so as to establish a constraint-optimization problem; solving the optimal parameter solutions by adopting an intelligent algorithm of constraint processing; adopting the obtained optimal parameter solutions as the model parameters of S1, and bringing the obtained optimal parameter solutions in the original model so as to predict or optimize the model. Through the adoption of the mixed modeling method disclosed by the present invention, the model which more conforms to priori knowledge can be obtained in the neural network training of a small amount of data samples, so that the over-fit phenomenon can be avoided.

Description

technical field [0001] The invention belongs to the field of chemical process modeling, and relates to a process modeling method, in particular to a hybrid modeling method combining process prior and data-driven models; at the same time, the invention also relates to a method combining process prior and data-driven A mixture of models. Background technique [0002] The traditional mechanism information modeling method must have a sufficient understanding of the modeled system, and establish an accurate description model through formulas such as mass-energy equations and reaction kinetic equations. However, the actual industrial process has high complexity and many reaction mechanisms, and it is difficult to accurately grasp it. As a result, the mechanism information modeling is often difficult to meet the accuracy requirements of the process system model. Data-driven methods have developed rapidly and are widely used in chemical process modeling and optimization. The advan...

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
IPC IPC(8): G06F17/50
Inventor 李绍军成祥杨一航许文夕郑文静
Owner EAST CHINA UNIV OF SCI & TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap