Pseudo Neural Network Torque Control via Root Solving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing engine control systems face difficulties in implementing torque prediction models due to the need for complex mathematical inversion, especially when using highly non-linear Artificial Neural Networks (ANNs), which can be cumbersome and difficult to implement effectively for responsive vehicle control.
Innovation Solution
The method involves creating a 'Pseudo ANN' by reducing the number of variables in the ANN to a single variable, allowing for torque control without relying on full mathematical inversion, using a simplified mathematical expression to solve for the root value, thereby enabling torque prediction and control through air-per-cylinder (APC) or spark control variables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a highly non-linear ANN torque prediction model is used, then prediction accuracy is improved, but the complexity of mathematical inversion increases making control implementation difficult
Solution Approach 1:
The patent segments the complex multi-variable ANN inversion problem into simpler sub-problems by fixing most variables and solving for one variable at a time. This is achieved through creating a pseudo-NN where n-1 variables are held constant and only one variable is solved for iteratively, breaking down the intractable full inversion into manageable sequential steps.
Solution Approach 2:
The patent transitions from solving a high-dimensional inversion problem (multiple independent variables) to solving a series of one-dimensional root-finding problems. By reducing the problem dimensionality from n variables to 1 variable at a time, the mathematical complexity is dramatically reduced while maintaining prediction accuracy.
2Productivity
If traditional torque prediction models are used, then implementation is simpler, but responsive vehicle control cannot be achieved
Solution Approach 1:
The patent performs preliminary action by pre-training the ANN model offline with comprehensive data, storing the trained weights and architecture. This preliminary training phase separates the complex model development from the real-time control phase, allowing simple iterative root-solving to be used during actual vehicle operation without sacrificing accuracy.
Solution Approach 2:
The patent introduces dynamic iterative root-solving that adapts to real-time operating conditions. Instead of using static lookup tables or fixed models, the system dynamically adjusts control variables through iterative computation, allowing responsive adaptation to changing vehicle states while maintaining implementation simplicity.
3Reliability
If full mathematical inversion of ANN is performed, then complete torque control is achieved, but computational burden becomes excessive
Solution Approach 1:
The patent extracts and isolates the essential control variable from the complex ANN model by fixing other variables at their current operating values. This extraction approach takes out only the necessary computational element (one variable root-finding) from the full inversion problem, achieving sufficient control authority without the excessive computational burden of complete inversion.
Data Source
AI summary
In various embodiments, methods, systems, and vehicle apparatuses are provided. A method for implementing torque control using a Neural Network (NN) for a torque prediction model to receive a set of measured vehicle operating inputs associated with torque prediction; substituting a set of multiple independent variables into the torque prediction model so that the NN is then taking the form of a simplified pseudo-NN that contains a reduced variable set of one independent variable; processing, the set of measured vehicle operating inputs by the pseudo-NN based on the NN prediction model by using only one independent variable in a pseudo-NN's simplified mathematical expression; and solving for at least one root of the pseudo-NN's simplified mathematical expression by obtaining a root value without having to rely on an inversion operation of a mathematical expression that consists of an entire set of independent variables.