Fuel Injector Correction Map Adaptation for Precise Metering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fuel metering systems in internal combustion engines, such as diesel engines, face inaccuracies due to variations among fuel injectors and age-related changes, leading to deviations between actual and target fuel quantities, which are challenging to correct over time.
Innovation Solution
A method utilizing a training dataset and correction dataset to adjust correction values based on virtual sensors, employing machine learning models and physical cause-effect relationships to determine accurate fuel metering by assigning status to fields in the datasets, and applying mean and transfer training values to adjust correction values across various operating points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If correction values are adjusted using virtual sensors and machine learning models, then fuel metering precision is improved over time, but system complexity increases
Solution Approach 1:
The system continuously monitors actual fuel quantity metered through virtual sensors and compares it with target values, then automatically adjusts correction values in the correction map based on the deviation. This closed-loop feedback mechanism enables continuous improvement of fuel metering precision without manual intervention, resolving the contradiction by automating the correction process despite increased system complexity.
Solution Approach 2:
The fuel injection system performs self-correction by using its own operational data to train machine learning models and update correction values autonomously. The system serves itself by continuously learning from actual measurements and adjusting its own parameters, eliminating the need for external calibration equipment or manual adjustments, thus improving precision while managing complexity through self-sufficiency.
2Reliability
If correction values are continuously adjusted over time, then accuracy is maintained throughout service life, but computational resources increase
Solution Approach 1:
The system applies correction values selectively based on operating conditions and only adjusts correction values when necessary, rather than continuously recalculating all parameters. The machine learning models are trained incrementally using only relevant operational data, performing partial updates that maintain accuracy while conserving computational resources during normal operation.
Solution Approach 2:
The system performs preliminary training of machine learning models during periods of low computational demand, such as during engine idle or shutdown periods, so that correction values are ready for immediate application during high-demand operation. This preliminary preparation ensures accuracy is maintained without requiring intensive computational resources during critical fuel injection phases.
3Device complexity
If virtual sensors are used to detect actual flow quantity, then direct measurement complexity is reduced, but measurement precision may be affected by calculation errors
Solution Approach 1:
The system introduces an intermediary correction map that translates between target fuel quantities and actual metered quantities based on historical operational data and machine learning predictions. This intermediary layer compensates for calculation errors in virtual sensors by applying learned correction factors, maintaining measurement precision while avoiding the complexity of direct physical measurement of actual fuel flow.
Data Source
AI summary
A method for adjusting correction values for metering fuel using at least one fuel injector, from a high-pressure accumulator into a combustion chamber of an internal combustion engine, using training and correction datasets. The method includes: in the correction dataset, adjusting the correction values based on training values of the training dataset, the correction value in each active neighboring field being adjusted based on a mean neighboring field training value, the correction value in each field of each field region that includes an active field being adjusted based on a mean field training value, wherein the correction value in each inactive field is adjusted based on a transfer training value, the transfer training value being determined according to at least one correlation rule, based on the training values of the active neighboring fields; and providing the correction map having the adjusted correction values for further use.