Coating Composition Property Prediction Using Analytical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining coating composition properties are time-consuming, expensive, and unreliable, particularly in predicting the effects of individual components on overall composition performance, often requiring extensive laboratory tests and failing to account for synergistic effects.
Innovation Solution
A method utilizing a computer-implemented approach that acquires and processes analytical data sets from multiple coating compositions to predict physicochemical properties, employing machine learning models like neural networks to determine optimal component proportions for desired properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional laboratory testing methods are used to determine coating composition properties, then measurement precision is improved, but loss of time and productivity deteriorate
Solution Approach 1:
The patent creates virtual copies of laboratory testing through machine learning models. Instead of physically testing each coating composition in the lab, the system uses trained ML models to predict properties based on compositional data, replacing time-consuming physical experiments with computational predictions that maintain acceptable accuracy
Solution Approach 2:
The patent performs preliminary action by pre-training machine learning models on extensive coating composition data before actual property determination is needed. The models are trained in advance on historical laboratory data, so that when new coating compositions need evaluation, the predictions can be made immediately without repeating the full experimental process
2Reliability
If extensive laboratory tests are conducted to account for synergistic effects of components, then reliability of prediction is improved, but productivity and loss of time worsen
Solution Approach 1:
The patent merges the evaluation of multiple coating components and their synergistic effects into a single integrated machine learning model. Instead of testing each component and interaction separately through extensive laboratory experiments, the ML model processes all compositional data together and predicts the combined effect on coating properties, capturing synergistic interactions computationally
Solution Approach 2:
The patent changes the approach from physical experimentation to computational prediction by transforming compositional parameters into predicted property values through machine learning. The system accepts coating composition parameters as input and generates property predictions, replacing the need for physical testing while maintaining the ability to account for complex component interactions
3Productivity
If machine learning models are used to predict coating composition properties, then productivity is improved, but measurement precision may worsen
Solution Approach 1:
The patent performs preliminary training of machine learning models using extensive historical laboratory data before deployment. This pre-training phase allows the models to learn accurate relationships between composition and properties, ensuring that when the models are used for prediction, they maintain high accuracy while delivering fast results
Solution Approach 2:
The patent implements feedback mechanisms where machine learning predictions can be validated and refined using available experimental data. The system allows for continuous improvement of model accuracy by incorporating new measurement data back into the training process, ensuring that productivity gains do not come at the expense of precision
Data Source
Figure 1
AI summary
The present invention relates to a method for the, in particular predictive, determination of at least one physicochemical coating composition property (YZ) of a first coating composition (XZ), comprising a first coating composition component (AZ) and at least one further coating composition component (BZ) different from the first coating composition component (AZ), with the steps of: - acquisition of a first analysis data set (X2) for a second coating composition, - acquisition of at least one further analysis data set (X3, ... Xn) for at least a third coating composition, and - computer-implemented determination of at least one coating composition property (YZ, ZZ, ...The invention relates to a computer-implemented method for generating a trained machine learning coating composition property determination model (6) for determining at least one physicochemical coating composition property (YZ, ZZ,...) of a coating composition based on the analysis data set (X2) for the second coating composition and the at least one further analysis data set (X3, ... Xn) for the at least one third coating composition. Furthermore, the invention relates to a computer-implemented method for generating a trained machine learning coating composition property determination model (6) for determining at least one physicochemical coating composition property (YZ, ZZ,...) of a coating composition based on the use of analysis data sets (X2 - Xn), as well as a coating composition property determination device.