Torque Converter Oil Pressure Estimation Using ML Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing oil pressure estimation techniques for vehicles with torque converters and lockup clutches face challenges in accurately controlling oil pressure due to deviations between instructed and actual oil pressure values, requiring complex relational expressions that are time-consuming to derive.
Innovation Solution
An oil pressure estimation device using machine learning-based mapping data to calculate estimated differential pressure, incorporating input variables such as instruction differential pressure, acceleration, accelerator operation, vehicle speed, gear ratio, and oil temperature, allowing for accurate estimation without the need to derive complex relational expressions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex relational expressions are derived to accurately estimate oil pressure, then measurement precision improves, but device complexity and time consumption increase
Solution Approach 1:
The patent replaces complex mathematical relational expressions with a neural network-based machine learning model. The neural network learns the mapping between instruction differential pressure and actual differential pressure through training data, eliminating the need for manual derivation of complex pressure relationship equations while maintaining high estimation accuracy.
Solution Approach 2:
The patent transforms the problem from solving complex differential equations to using a trained neural network model that takes instruction differential pressure as input and outputs estimated actual differential pressure. This parameter transformation approach converts a mathematically intensive problem into a computationally efficient machine learning inference task.
2Measurement precision
If complex relational expressions are derived to accurately estimate oil pressure, then measurement precision improves, but loss of time increases
Solution Approach 1:
The patent performs preliminary training of the neural network model using historical data before actual operation. During the training phase, the system learns the complex pressure relationships from training datasets containing instruction differential pressures and corresponding actual differential pressures. Once trained, the model can rapidly estimate actual pressure during real-time control without requiring time-consuming calculations.
Solution Approach 2:
The patent substitutes time-consuming analytical derivation of pressure relationships with pre-trained machine learning inference. The neural network model, after being trained offline, provides rapid predictions during real-time operation, eliminating the need for complex real-time calculations while maintaining high accuracy.
3Ease of operation
If machine learning mapping is used to estimate differential pressure, then ease of operation improves, but device complexity increases
Solution Approach 1:
The patent creates a virtual model (neural network) that copies the behavior of the actual torque converter pressure system. The trained neural network serves as a digital twin that replicates the complex pressure relationships without requiring physical sensors or complex calculation hardware, simplifying the estimation process while encapsulating the complexity within the trained model.
Data Source
AI summary
An oil pressure estimation device for calculating an estimated differential pressure that is an estimated value of a differential pressure between two oil chambers generated in a torque converter including the two oil chambers and a lockup clutch includes a storage device and an execution device. The storage device stores mapping data defining a mapping, the mapping outputting as an output variable an estimated differential pressure variable indicating the estimated differential pressure, in response to input of an input variable, and the mapping having been trained by machine learning. The mapping includes an instruction differential pressure variable indicating the instruction differential pressure as one of a plurality of the input variables. The execution device executes an acquisition process of acquiring a value of the input variable and a calculation process of inputting the value of the input variable into the mapping to calculate a value of the output variable.