Multi-Cylinder Engine Control Device Vibration Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control devices for multi-cylinder internal combustion engines face challenges in reducing engine vibrations and converging learned values during the first and second learning processes, leading to control interference and impaired value updates due to overlapping components from fuel injection valves.
Innovation Solution
A control device that executes a first learning process to reduce crankshaft rotation deviation and a second learning process based on fuel pressure, with a decreased learning rate for the second learned value until the first learned value converges, allowing for early reduction of engine vibrations and accurate value updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If both the first learning process and the second learning process are executed simultaneously in an idle operating state, then the controllability of fuel injection is improved, but control interference occurs and convergence time increases
Solution Approach 1:
The patent applies preliminary action by executing the first learning process before the second learning process. Specifically, the controller executes the first learning process to learn rotation fluctuation correction values, and only after this learning is complete does it execute the second learning process to learn fuel injection correction values. This sequential approach prevents control interference between the two learning processes while ensuring that both correction values are properly learned, thereby maintaining controllability without extending convergence time.
2Reliability
If both learning processes are executed simultaneously, then fuel injection controllability is improved, but the learned values contain overlapping components that impair accurate updates
Solution Approach 1:
The patent applies segmentation by dividing the learning processes into two distinct sequential stages. The first learning process focuses exclusively on rotation fluctuation correction, while the second learning process focuses exclusively on fuel injection correction. This segmentation prevents overlapping components from being mixed in the learned values, ensuring that each correction value is learned independently and accurately without interference from the other learning process.
3Stability of the object's composition
If the first learning process is executed first, then rotation fluctuation correction is achieved, but fuel injection variations due to manufacturing or deterioration are not corrected
Solution Approach 1:
The patent applies continuity of useful action by executing the second learning process after the first learning process completes. The controller continuously learns rotation fluctuation correction values in the first learning process, and then continuously learns fuel injection correction values in the second learning process. This continuous sequential learning ensures that both rotation fluctuation stability and fuel injection controllability are achieved, with the second learning process correcting fuel injection variations due to manufacturing or deterioration that were not addressed in the first learning process.
Data Source
AI summary
An electronic control unit, in an idle operating state, detects a crankshaft rotation fluctuation in each cylinder using a crank angle sensor, and updates an individual correction value for a control value for each fuel injection valve as a first learned value such that a degree of deviation in the crankshaft rotation fluctuation among the cylinders reduces. The electronic control unit uses a fuel pressure sensor to detect a manner of a fuel pressure fluctuation with fuel injection by each fuel injection valve, and updates an individual correction value for a control value for each fuel injection valve as a second learned value based on a result of comparison between a detected temporal waveform and a basic temporal waveform. In an idle operating state, a learning rate of the second learned value is reduced until the first learned value converges for the first time as compared with after its convergence.