Gear Grinding Data Modeling for Virtual Rolling Test Prediction

The patent creates a virtual copy (digital twin) of the physical gearing component through data-driven simulation. This virtual model replicates the geometric and functional characteristics of the actual gearing, allowing virtual rolling tests to be performed on the digital twin instead of repeatedly testing physical components. The virtual model is trained using machine learning on data from initial physical tests, enabling accurate prediction of rolling test results without requiring extensive physical testing of every component.

The patent replaces the physical mechanical testing system with a computational simulation system. Instead of using actual rolling test benches to physically roll and measure gearings, the invention uses computer-based simulations with trained machine learning models to predict rolling test outcomes. This substitution eliminates the need for extensive physical infrastructure and manual testing operations while maintaining assessment accuracy.

The patent performs preliminary actions by training the machine learning model once using data from an initial set of physical rolling tests. This training phase creates a pre-configured virtual model that can then be used for rapid predictions without requiring extensive computing resources during actual production testing. The heavy computational work is done upfront during model training, not during each individual prediction.

The patent transforms the complex physical rolling test parameters into simplified input features for the machine learning model. By identifying and extracting only the most relevant geometric and operational parameters that influence rolling test results, the system reduces the dimensionality and complexity of the data processing requirements while maintaining prediction accuracy.

The patent implements feedback mechanisms where the virtual twin model is continuously refined using actual measurement data from physical components. The model learns from discrepancies between predicted and actual rolling test results, automatically adjusting its internal parameters to improve accuracy. This feedback loop enables the system to achieve high prediction precision without requiring manual input of every possible geometric deviation parameter.

The system performs self-calibration and self-improvement by automatically learning from measurement data. Rather than requiring external experts to manually characterize every geometric deviation, the machine learning model autonomously identifies patterns and relationships in the data, extracting relevant information about tool and machine deviations through its training process without requiring explicit parameter specification.