Metering Pump Operational State Detection Using ML Torque Signals

The patent replaces the mechanical expert-knowledge-based curve analysis system with a machine learning model that automatically processes torque data. The ML model learns patterns from training data and provides adaptive operational information determination without requiring expert intervention, thus reducing device complexity while improving adaptability across different pump types.

The patent changes the approach from fixed threshold comparisons to dynamic parameter adaptation through machine learning. The system uses training data to learn optimal parameters for different pump types, enabling automatic adaptation without manual configuration. This allows the same system to handle multiple pump types effectively.

The patent performs preliminary action by training the machine learning model offline before actual operation. During runtime, the pre-trained model quickly processes torque data without requiring complex real-time computations. This separates the computationally intensive learning phase from the efficient inference phase, improving computational efficiency while maintaining reliability.

The patent uses copying by creating a trained model representation that captures the essential patterns from extensive training data. This model copy can then be efficiently applied to multiple different pump types and operating conditions without requiring the original complex analysis procedures, thus improving both efficiency and reliability.