Recycled Polymer Composite Composition for Target Property Optimization

The patent applies preliminary action by pre-establishing learned models that predict physical properties from composition parameters. These models are trained in advance on existing data, allowing rapid prediction without repeated experimentation. The system uses the learned models to directly determine optimal compositions based on desired physical properties, eliminating the need for time-consuming trial-and-error experiments.

The patent uses copying by creating virtual recycled polymer composite materials with predicted physical properties that replicate real materials. The learned models copy the relationship between composition and physical properties from training data, allowing the system to predict properties of virtual compositions without physically producing them. This enables efficient optimization by working with digital representations rather than physical materials.

The patent introduces learned models as intermediaries between the desired physical properties and the composition parameters. These models mediate the complex relationship by providing direct predictions of physical properties from composition data, and vice versa. The optimization unit uses these intermediary models to automatically determine blend ratios without requiring complex manual analysis, simplifying the overall process.

The patent applies parameter changes by transforming the complex problem of analyzing multiple characteristics into optimization of key parameters. The system changes the approach from analyzing numerous individual characteristics to optimizing composition parameters (mass ratios of recycled polymer, glass fiber, rubber) that directly control physical properties. The learned models handle the complexity of parameter relationships, allowing straightforward optimization.