Injector Closure Point Detection via ML Classifier

The patent transforms the closure point in time determination from a direct threshold-based measurement into a probabilistic classification problem. The classifier processes time series data and outputs probability values, allowing the system to adapt to variations in structural parameters, tolerances, and operating conditions by learning optimal decision boundaries from training data.

The patent introduces a classifier as an intermediary between the raw sensor measurements and the closure point determination. This classifier acts as a mediator that processes the time series data through learned patterns, bridging the gap between noisy measurements and reliable closure point identification, thereby improving both precision and reliability.

The patent replaces complex mechanical measurement and adjustment mechanisms with a computational approach. Instead of using highly precise mechanical sensors and manual calibration systems, the invention uses a classifier that processes standard sensor data through learned patterns, achieving high fuel metering precision through software-based intelligence rather than mechanical complexity.

The classifier is trained using training data that enables it to autonomously learn the relationship between time series patterns and closure points. Once trained, the system self-adjusts to different operating conditions without requiring external calibration or adjustment, reducing the need for complex manual intervention while maintaining high precision.