Impeller Clutch Control via Predictive Analytical Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic control systems for impeller clutches in earth-moving machines are inefficient due to reliance on heuristic rule-based controllers and linear algorithms, which fail to adapt to varying material densities and conditions, leading to suboptimal torque and wheel speed management.
Innovation Solution
A time-step predictive analytical model is trained using fluidic pressure information to modulate impeller clutch engagement, allowing for adaptive control through a machine learning framework that emulates human operator responses to dynamic conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If heuristic rule-based controllers are used for impeller clutch control, then the system is simple to implement, but the control accuracy and adaptability to varying material densities are insufficient
Solution Approach 1:
The patent replaces traditional mechanical rule-based control systems with a machine learning-based analytical model. The system uses neural networks and other ML algorithms to predict optimal clutch engagement values based on real-time sensor data, substituting deterministic mechanical control logic with adaptive data-driven models that can handle the nonlinearity and variability of earth-moving conditions.
Solution Approach 2:
The system dynamically adjusts clutch engagement parameters based on predicted material density and machine operating conditions. The analytical model continuously updates control parameters (clutch engagement value) based on changing conditions, allowing the system to adapt to varying material densities and maintain optimal performance across different operating scenarios.
2Device complexity
If linear algorithms are used for automatic control, then the control system is computationally simple, but the system cannot adapt to varying material densities and conditions
Solution Approach 1:
The patent implements a dynamic control system using machine learning models that continuously adapt to changing conditions. The analytical model processes real-time sensor data and dynamically adjusts clutch engagement predictions based on current material density, machine speed, and operating conditions, enabling the system to handle highly variable and non-stationary earth-moving environments.
Solution Approach 2:
The system incorporates feedback loops where sensor measurements of actual machine performance and material interaction are fed back into the analytical model. This feedback enables the model to continuously refine its predictions and adapt to changing conditions, with the model learning from actual system responses to improve future control decisions.
3Device complexity
If conventional control systems are used, then the system structure is simple, but the system cannot provide fully-automated impeller control for varying work pile conditions
Solution Approach 1:
The patent implements a self-learning control system where the analytical model automatically improves its performance through continuous operation. The system uses collected operational data to retrain and refine its predictions without requiring manual intervention or reprogramming, enabling the controller to autonomously adapt to new conditions and materials encountered during operation.
Solution Approach 2:
The analytical model serves multiple functions: it predicts clutch engagement values, estimates material density, adapts to different operating conditions, and can be retrained with new data. This multi-functional approach allows a single system to handle diverse earth-moving conditions and materials, providing fully automated control across various scenarios without requiring separate control logic for each condition.
Data Source
AI summary
A method includes obtaining first fluidic pressure information indicative of fluidic pressure in a hydraulic cylinder of a machine as a function of time, and generating a first training set including second fluidic pressure information associated with impeller clutch engagement values. A training system associated with the machine computes a first plurality of test system response values based on the first training set, compares first plurality of test system response values with a plurality of observed response values, and determines a first response error. The training system determines whether the first response error is less than or equal to a threshold error value, generates a time-step predictive analytical model associated with the first training set, and provides the time-step predictive analytical model to an electronic control module of the machine. The time-step predictive analytical model is usable to control machine torque and/or an impeller clutch of the machine.