Engine Calibration Map Optimization via Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The calibration process for engine control algorithms in internal combustion engines is time-consuming and costly due to the need for extensive experimental measurements and complex algorithms, which are prone to measurement errors and imprecision.
Innovation Solution
A method for optimizing calibration maps by individually optimizing each map using competence indices and a 'stretching' procedure to improve continuity, reducing the number of experimental measurements and enhancing precision through the Levenberg Marquardt algorithm when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive experimental measurements are performed to create calibration maps, then the precision of engine control algorithms is improved, but the calibration time and development costs increase significantly
Solution Approach 1:
The calibration process is segmented into two distinct phases: an offline phase where comprehensive experimental measurements are performed to create initial calibration maps, and an online phase where a neural network algorithm rapidly optimizes these maps in real-time based on actual engine sensor data. This segmentation allows extensive measurements to be done once offline, while online calibration becomes fast and efficient.
Solution Approach 2:
Comprehensive experimental measurements and initial calibration maps are performed in advance during the offline phase before actual engine operation. The neural network is pre-trained with this calibration data, so when deployment occurs, the system already has a solid foundation of calibrated information, reducing the need for extensive real-time measurements.
2Measurement precision
If comprehensive experimental measurements are performed to create calibration maps, then the accuracy of control quantity estimation is improved, but the complexity of the calibration process increases
Solution Approach 1:
The traditional mechanical calibration process involving extensive manual measurements and adjustments is replaced with an automated neural network algorithm. The neural network automatically processes sensor data, compares it with calibration maps, and adjusts calibration parameters without requiring manual intervention, thereby reducing process complexity while maintaining or improving accuracy.
Solution Approach 2:
The calibration system becomes self-adjusting through the neural network, which automatically optimizes calibration maps using real-time sensor data from the engine. The system performs self-calibration without requiring external calibration equipment or manual adjustments, reducing the complexity of the overall calibration process.
3Measurement precision
If traditional calibration methods are used with multiple maps and complex algorithms, then the estimation accuracy is maintained, but the calculation time increases
Solution Approach 1:
Traditional calculation-intensive calibration methods are replaced with a neural network-based system. The neural network processes calibration data and sensor inputs much faster than traditional algorithms, achieving both high estimation accuracy and fast calculation speeds suitable for real-time engine control applications.
Solution Approach 2:
The calibration approach transitions from using multiple separate calibration maps with complex interpolation algorithms to a single integrated neural network model. This parameter change in the calibration methodology simplifies the calculation structure while maintaining or improving estimation accuracy through the neural network's ability to learn complex non-linear relationships.
Data Source
AI summary
Described herein is a method for optimizing a plurality of calibration maps for an algorithm of estimation of a control quantity of an internal combustion engine, each of the maps comprising a plurality of calibration values of said control quantity estimated by said algorithm. The optimization method comprises measuring the control quantity, estimating the control quantity, and individually optimizing each calibration map based on the measured control quantity and the estimated control quantity.