CN-121999936-A - Paint simulation prediction method and system

CN121999936ACN 121999936 ACN121999936 ACN 121999936ACN-121999936-A

Abstract

The invention provides a paint simulation prediction method and a system, wherein the paint simulation prediction method comprises the steps of constructing a mapping relation database between process parameters, characteristic parameters, evaluation parameters and quality identifiers by adopting stable data intervals after unstable data intervals are identified and removed based on normal distribution analysis, calculating a first-stage Bayesian judgment point, a second-stage Bayesian judgment point and a third-stage Bayesian judgment point or calculating a first-stage Bayesian reverse judgment point, a second-stage Bayesian reverse judgment point and a third-stage Bayesian reverse judgment point by adopting a hierarchical Bayesian estimation theory based on stable data in the established mapping relation database, and outputting simulation prediction results. The invention guarantees the authenticity and the fitting degree of the mapping relation based on normal distribution screening data, improves the accuracy and the representativeness of the simulation result by adopting hierarchical Bayesian estimation, and realizes the function of forward predicting the quality of the paint and reversely carrying out simulation prediction on the technological parameters of the paint.

Inventors

HU FANGMING
SU MENGXING
XUE JUNHAO
HONG YUEHUI
ZHANG CHI
LIANG YING
LOU LI
LI KUANGYU

Assignees

洛阳船舶材料研究所(中国船舶集团有限公司第七二五研究所)

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The paint simulation prediction method is characterized by comprising the following steps of: s1, collecting original data Receiving, at a computer system comprising a processor, raw data relating to a coating, the raw data comprising process parameters, feature parameters, evaluation parameters, and a quality identification; s2, establishing a mapping relation database Carrying out normal distribution analysis on the characteristic parameters collected in the step S1, identifying and removing unstable data intervals, reserving stable data intervals, and constructing a mapping relation database among process parameters, characteristic parameters, evaluation parameters and quality identifiers based on the stable data intervals; S3, calculating based on Bayesian estimation theory Based on the stable data in the mapping relation database established in the step S2, inputting known technological parameters at a computer system comprising a processor, adopting a hierarchical Bayesian estimation theory to calculate a first-stage Bayesian decision point, a second-stage Bayesian decision point and a third-stage Bayesian decision point, or inputting a known quality identifier to calculate a third-stage Bayesian reverse decision point, a second-stage Bayesian reverse decision point and a first-stage Bayesian reverse decision point; S4, outputting simulation prediction results When the known process parameters are input in the step S3, the simulation prediction result is a coating quality identifier, and when the known quality identifier is input in the step S3, the simulation prediction result is a coating process parameter.
2. The paint simulation prediction method of claim 1, wherein when the outputted simulation prediction result is a paint quality identifier, step S3 includes: The method comprises the steps of calculating a first-level Bayesian decision point, wherein the first-level Bayesian decision point identifies the mapping probability of a characteristic parameter corresponding to the technological parameter, calculating a second-level Bayesian decision point, wherein the second-level Bayesian decision point identifies the mapping probability of an evaluation parameter corresponding to the characteristic parameter, and calculating a third-level Bayesian decision point, wherein the third-level Bayesian decision point identifies the mapping probability of a quality identifier corresponding to the evaluation parameter, wherein the output of the first-level Bayesian decision point is used as the input of the second-level Bayesian decision point, and the output of the second-level Bayesian decision point is used as the input of the third-level Bayesian decision point.
3. The paint simulation prediction method of claim 1, wherein when the outputted simulation prediction result is a paint process parameter, the step S3 includes: The method comprises the steps of determining a quality identifier, calculating a third-level Bayesian reverse determination point, wherein the third-level Bayesian reverse determination point is used for identifying the mapping probability of the evaluation parameter corresponding to the quality identifier, calculating a second-level Bayesian reverse determination point, wherein the second-level Bayesian reverse determination point is used for identifying the mapping probability of the characteristic parameter corresponding to the evaluation parameter, and calculating a first-level Bayesian reverse determination point, wherein the first-level Bayesian reverse determination point is used for identifying the mapping probability of the process parameter corresponding to the characteristic parameter, the output of the third-level Bayesian reverse determination point is used as the input of the second-level Bayesian reverse determination point, and the output of the second-level Bayesian reverse determination point is used as the input of the first-level Bayesian reverse determination point.
4. A paint simulation prediction method according to claim 3, wherein when the simulation prediction result is a paint process parameter satisfying a manually constrained feature parameter, in step S3, the output of the third-stage bayesian reverse decision point and the manually constrained feature parameter are used as the input of the second-stage bayesian reverse decision point at the same time.
5. The paint simulation prediction method according to claim 1, wherein in the step S1, the process parameters include an auxiliary agent type, a diluent type, and a stirring time, the characteristic parameters include fineness and viscosity, the evaluation parameters include appearance and performance, and the quality marks include pass and fail.
6. The method according to claim 1, wherein in step S2, after the stable data interval screened by the user is input at the computer system including the processor, the computer system automatically selects and eliminates the unstable data interval and the rest stable data interval, or the computer system automatically selects and eliminates the data interval with the duty ratio close to 0, and the stable data interval is reserved.
7. A simulated predictive system for paint comprising: the system comprises a database, a data storage and a data storage, wherein the database is used for storing raw data related to the paint and a mapping relation database, the raw data comprises process parameters, characteristic parameters, evaluation parameters and quality identifiers, and the mapping relation database is constructed based on stable data intervals after normal distribution screening; The processor is communicated with the database and programmed to receive raw data related to the paint, perform normal distribution analysis on the characteristic parameters, identify and reject unstable data intervals, reserve stable data intervals, and construct a mapping relation database among process parameters, characteristic parameters, evaluation parameters and quality identifiers based on the stable data intervals; The method comprises the steps of adopting data in a mapping relation database to calculate a simulation prediction result based on a hierarchical Bayesian estimation theory, importing the data in the mapping relation database into software to calculate a first-level Bayesian decision point which identifies the mapping probability of a characteristic parameter corresponding to an input process parameter, calculating a second-level Bayesian decision point which identifies the mapping probability of an evaluation parameter corresponding to the characteristic parameter, wherein the output of the first-level Bayesian decision point is used as the input of the second-level Bayesian decision point; The processor is programmed to support a bi-directional output mode including a quality prediction mode based on a forward Bayesian decision point and a formulation design mode based on a Bayesian reverse decision point, wherein an output of the third level Bayesian reverse decision point is used as an input of the second level Bayesian reverse decision point and an output of the second level Bayesian reverse decision point is used as an input of the first level Bayesian reverse decision point; And the man-machine interaction module is used for receiving data input by a user and outputting simulation prediction results.
8. The simulation prediction system of claim 7, wherein the simulation prediction result is a quality identifier or a process parameter, wherein the simulation prediction result is output as the quality identifier when the man-machine interaction module receives the process parameter input by the user, and wherein the simulation prediction result is output as the process parameter when the man-machine interaction module receives the quality identifier input by the user.
9. The simulation prediction system of claim 7, wherein the human-computer interaction module dynamically interacts with the first level bayesian decision point, the second level bayesian decision point and the third level bayesian decision point to receive one or more of a process parameter, a feature parameter and a manually constrained feature parameter input by a user and output a process parameter or quality identification prediction result.
10. The simulation prediction system of any one of claims 7 to 9, wherein the man-machine interaction module further receives a stable data interval input by a user, and is configured to identify and reject an unstable data interval.

Description

Paint simulation prediction method and system Technical Field The invention relates to the field of quality prediction and formula design, in particular to a paint simulation prediction method and system. Background The quality of the paint is used as a key material in the fields of industrial production, architectural decoration, product protection and the like, and directly influences the appearance performance, the durability and the market competitiveness of the product. In the process of coating research and development and production, how to accurately predict the quality of a coating and design a coating formula with high efficiency is a core problem faced by the industry for a long time, namely, the traditional coating research and development needs to rely on experience of technicians to carry out multiple rounds of trial and error, the period is long, the trial and error cost is high, the research and development cost is further improved by verifying the formula accuracy of at least three batches of products, meanwhile, complex nonlinear relations exist between process parameters (such as auxiliary agent types, diluent types and stirring time), characteristic parameters (such as fineness and viscosity) of the coating, coating evaluation parameters (such as appearance and performance) and final quality, and accurate corresponding relations are difficult to establish by simply relying on experience or simple experiments, so that the stability of formula design and quality control is insufficient. Patent application number CN105051762 discloses a bayesian processing based paint formulation determination system and method that identifies body pigment and refined pigment types by analyzing spectrophotometric or image data of a target paint, thereby generating a matching formulation. Although the technology adopts the Bayes theory to improve the accuracy of pigment identification, the technology does not relate to the field of quality prediction, and the accuracy of early-stage data processing is lacking, and the technology is not matched with actual production scenes. Patent application number CN110288199 discloses a product quality prediction method, although the prediction accuracy is improved through data screening and a neural network integrated model, the quality prediction can only be realized, a formula cannot be reversely designed, and the reality of data and the fitting degree of a production scene are limited. Therefore, an integrated technical scheme capable of realizing accurate prediction of coating quality and efficient design of a formula through scientific data analysis and probabilistic reasoning is needed. Disclosure of Invention In view of the above, the present invention aims to provide a simulation prediction method and system capable of combining actual production and simultaneously realizing coating quality prediction and formulation design, so as to solve the problems of inaccurate quality prediction and limited adhesion degree with actual production in the prior art. In order to achieve the above purpose, the technical scheme of the invention is realized as follows: the invention provides a paint simulation prediction method, which comprises the following steps: s1, collecting original data Receiving, at a computer system comprising a processor, raw data relating to a coating, the raw data comprising process parameters, feature parameters, evaluation parameters, and a quality identification; s2, establishing a mapping relation database Carrying out normal distribution analysis on the characteristic parameters collected in the step S1, identifying and removing unstable data intervals, reserving stable data intervals, and constructing a mapping relation database among process parameters, characteristic parameters, evaluation parameters and quality identifiers based on the stable data intervals; S3, calculating based on Bayesian estimation theory Based on the stable data in the mapping relation database established in the step S2, inputting known technological parameters at a computer system comprising a processor, adopting a hierarchical Bayesian estimation theory to calculate a first-stage Bayesian decision point, a second-stage Bayesian decision point and a third-stage Bayesian decision point, or inputting a known quality identifier to calculate a third-stage Bayesian reverse decision point, a second-stage Bayesian reverse decision point and a first-stage Bayesian reverse decision point; S4, outputting simulation prediction results When the known process parameters are input in the step S3, the simulation prediction result is a coating quality identifier, and when the known quality identifier is input in the step S3, the simulation prediction result is a coating process parameter. In the invention, the paint simulation prediction method comprises a paint quality prediction path or a formula design path. As shown in FIG. 1, the invention integrates the original data in the big d