Gas-Mass Flow Correction Using Non-Parametric Gaussian Process
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Combustion engines face pulsation errors in gas-mass flow measurements due to cyclical operation, leading to inaccurate sensor readings, as existing correction methods are limited by parameter-dependent models and restricted input quantities, resulting in interpolation errors and reduced precision.
Innovation Solution
A data-based, non-parametric function model using a Gaussian process is employed to calculate a correction quantity, allowing for a higher number of input quantities and excluding interpolation errors, with a separate hardware unit for rapid calculation, applying the correction to smoothed sensor signals in an additive or multiplicative manner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a 2D program map or combination of 2D program map and 1D interpolation is used for correction, then the device complexity is reduced, but the measurement precision deteriorates due to interpolation errors and limited data points
Solution Approach 1:
The patent replaces the traditional mechanical interpolation-based correction system with a data-based non-parametric function model (neural network). This substitution eliminates the need for rigid 2D program maps and 1D interpolation methods, allowing the system to learn complex correction patterns from training data and provide accurate corrections without interpolation errors, thereby resolving the contradiction between device complexity and measurement precision.
Solution Approach 2:
The patent changes the correction approach from using a limited number of input quantities (maximally three) in traditional methods to utilizing a higher number of input quantities in the non-parametric function model. This parameter change enables the model to capture more aspects of the pulsation behavior and provide more accurate corrections across the entire parameter space, improving measurement precision without being constrained by the complexity limitations of earlier methods.
2Device complexity
If only maximally three input quantities are employed for ascertaining correction quantity, then the device complexity is reduced, but the measurement precision deteriorates due to limited parameter space coverage
Solution Approach 1:
The patent implements a universal correction system using a non-parametric function model that can process a higher number of input quantities simultaneously. This multi-functional approach allows the single correction system to handle various operating conditions and parameter combinations effectively, providing accurate corrections across the entire parameter space without requiring separate correction functions for different regions, thus improving measurement precision while maintaining manageable device complexity.
3Ease of manufacture
If traditional correction functions with defined data points in a raster are used, then the ease of manufacture is improved, but the measurement precision deteriorates due to interpolation errors in regions with limited data points
Solution Approach 1:
The patent replaces the traditional raster-based correction function implementation with a data-based non-parametric function model (neural network). This substitution eliminates the need to manually define data points in a raster structure and perform interpolations, allowing the system to learn optimal correction patterns directly from training data. This approach maintains ease of manufacture through standardized neural network implementation while dramatically improving measurement precision by eliminating interpolation errors entirely.
Data Source
AI summary
A method for ascertaining a mean value of a gas-mass flow in a combustion engine. The method includes measuring a gas-mass flow impinged upon by a pulsation, smoothing a sensor signal obtained by the measurement, applying a correction quantity to the smoothed sensor signal in order to obtain the mean value of the gas-mass flow, and ascertaining the correction quantity as a function of the operating state of the combustion engine with the aid of a data-based, non-parametric function model.