Knock Detection Using Dynamic Injection Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knock detecting devices face challenges in ensuring accurate knock detection across various fuel injection conditions, particularly when injection-valve noise overlaps with the knock determination period, and fail to adapt to changing conditions such as multiple-stage injections or different cylinder configurations.
Innovation Solution
A knock detecting device that includes a first calculation unit for frequency analysis of knock sensor signals, a second calculation unit for determining a background level, a storage unit for associating fuel injection numbers with learning values, and a knock determination unit that uses a knock index to differentiate between knock and noise signals based on the ratio of frequency components and learning values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the off-timing of fuel injection is masked from the knock determination period to avoid injection-valve noise, then knock detection accuracy is improved, but knock occurring in the mask period cannot be detected
Solution Approach 1:
The fuel injection process is segmented into multiple stages (first-stage injection and second-stage injection), allowing separate learning and determination periods to be defined for each stage. This enables the system to handle different injection events independently, preventing noise from one stage from interfering with knock detection in another stage.
Solution Approach 2:
The system performs preliminary learning of injection-valve noise characteristics before actual knock determination. By pre-learning the noise patterns during a learning period and storing them as reference data, the system can subsequently subtract this learned noise from the knock determination signals, enabling accurate knock detection even when injection timing overlaps with the determination period.
2Ease of operation
If a fixed learning value is used for knock determination, then the system is simple to operate, but knock detection accuracy varies with different fuel injection conditions
Solution Approach 1:
The learning value is made dynamic rather than fixed. The system automatically adjusts the learning value based on the detected number of fuel injections by selecting from multiple pre-stored learning values corresponding to different injection conditions (1 injection, 2 injections, 3 or more injections). This dynamic adaptation maintains high knock detection accuracy across varying injection conditions without requiring manual intervention.
Solution Approach 2:
The system changes the learning value parameter according to the injection condition parameter (number of injections). By storing multiple learning values in the storage unit and selecting the appropriate one based on the current injection count, the system adapts to different operational conditions while maintaining simple automatic operation.
3Measurement precision
If the learning period excludes the knock determination period to learn injection-valve noise, then noise learning accuracy is improved, but the system cannot learn under actual knock determination conditions
Solution Approach 1:
The learning period is segmented into multiple sub-periods corresponding to different injection stages. The learning unit learns injection-valve noise characteristics during specific sub-periods (e.g., first learning sub-period for first-stage injection, second learning sub-period for second-stage injection) while excluding other periods. This segmented approach allows the system to learn noise characteristics under various actual operating conditions and store multiple learning values for different injection scenarios.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Ensures accurate knock detection irrespective of fuel injection conditions, even when injection-valve noise occurs during the knock determination period, by learning and adapting to the specific conditions of each cylinder and injection stage, thereby preventing misjudgment of operational changes as knock occurrences.
Implementation Method 1
a knock sensor, which detects vibrations of an internal combustion engine
Implementation Method 2
a first calculation unit that calculates frequency components by subjecting signals output from a knock sensor to frequency analysis
Data Source
AI summary
Provided is a knock detecting device capable of ensuring knock detection accuracy irrespective of fuel injection conditions even when a period during which injection-valve noise occurs overlaps with a knock determination period. An ECU 9 subjects signals output from a knock sensor, which detects vibrations of an internal combustion engine, to frequency analysis to calculate frequency components (501). The ECU 9 calculates a background level that indicates the average of the frequency components (503). The ECU 9 stores, in association with each other, the number of fuel injections, which indicates the number of fuel injections in a predetermined time period during one combustion cycle, with a learning value, which indicates a frequency-component correction amount (507). The ECU 9 determines the presence or absence of a knock (504) on the basis of a knock index, which indicates the ratio of the difference between each frequency component and the learning value corresponding to the number of fuel injections, to the background level (503).