Engine Actuator Model Learning Gain for Reliable Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning strategies for internal combustion engine actuators are prone to inaccuracies due to atypical events, leading to fuel mixture deviations and performance issues, which are not effectively addressed by current correction methods.
Innovation Solution
A method for securing the learning function of actuator control models by calculating a difference between stored and current values, determining a confidence index, and using a safety parameter to weight the update of model terms, ensuring reliable convergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the learning function fully learns and overwrites the stored correction value, then the learning speed is improved, but the reliability deteriorates when atypical events occur
Solution Approach 1:
The patent implements dynamic adjustment of the learning gain parameter K based on confidence index and deviation thresholds. The learning gain is not fixed but adapts in real-time: it is reduced when atypical events are detected (low confidence index or high deviation) and increased when conditions are stable (high confidence index and low deviation). This dynamic mechanism resolves the contradiction by making the learning speed variable rather than constant, allowing fast learning during normal operation while preventing erroneous updates during atypical events.
Solution Approach 2:
The patent changes the parameter K (learning gain) based on the confidence index and deviation between stored and current values. When the deviation exceeds a threshold or confidence is low, K is reduced to limit the update magnitude. When conditions are favorable, K is increased to accelerate learning. This parameter change strategy allows the system to optimize both learning speed and reliability by adjusting the learning intensity according to the current operational context.
2Reliability
If the learning function learns only a portion of the current correction value, then the reliability is improved by limiting negative effects of atypical events, but the learning speed deteriorates
Solution Approach 1:
The patent makes the learning gain parameter K dynamic rather than fixed. Instead of always learning only a portion (which would slow down learning), the system adjusts K based on confidence index and deviation metrics. When conditions indicate high reliability (stable operation, low deviation), K is increased to accelerate learning. When reliability is questionable (atypical events, high deviation), K is reduced to prevent erroneous updates. This dynamic approach resolves the contradiction by making learning speed adaptive to reliability conditions.
Solution Approach 2:
The patent implements a feedback mechanism where the confidence index and deviation between stored and current values are continuously monitored. This feedback information is used to adjust the learning gain parameter K in real-time. The feedback loop ensures that the learning process responds to the actual quality of measurements, increasing learning speed when reliable and reducing it when unreliable, thereby resolving the trade-off between speed and reliability.
3Device complexity
If the learning function uses a fixed learning gain parameter, then the device complexity is reduced, but the adaptability deteriorates when operating conditions change
Solution Approach 1:
The patent implements a self-service mechanism where the learning function automatically adjusts its own parameter K based on internal metrics (confidence index, deviation) without requiring external intervention or complex configuration. The system monitors its own performance and adaptively tunes the learning gain, making the mechanism self-regulating. This approach maintains relatively simple device architecture while achieving high adaptability to changing operating conditions.
Solution Approach 2:
The patent changes the learning gain parameter K dynamically based on operating conditions, confidence index, and measurement deviation. Instead of using a fixed parameter, the system adapts the parameter value in real-time according to the current operational context. This parameter change strategy enables the simple learning mechanism to exhibit complex adaptive behavior, resolving the contradiction between device simplicity and operational adaptability.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The present invention relates to a method for improving the reliability of a learning function of a model for controlling actuators of a motor vehicle combustion engine, wherein the method comprises the following steps of calculating (25) a difference between a first value stored by the learning function for the term of the model and a second current value of the term of the model, which value is estimated from instantaneous measurements, of determining (26) a confidence index assigned to the first stored value of the term of the model, of determining (27) the value of a reliability-improving parameter dependent on the difference and on the confidence index, of updating (28) the term of the model by a third value made up of the sum of the parts of the first value and of the second value, during which one of the parts or each of the parts is dependent on the reliability-improving parameter.