Ignition Timing Control Using Neural Network Knocking Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ignition timing control systems for internal combustion engines face challenges in accurately detecting knocking intensity due to mechanical vibrations, leading to false determinations and excessive retarding of ignition timing, especially when using learned neural networks that are not pre-trained for specific engine vibrations, resulting in reduced engine output.
Innovation Solution
An ignition timing control device that employs a storage unit with a pre-learned normal signal generation model and neural networks to remove unlearned noise components from knocking sensor outputs, allowing for accurate estimation of knocking intensity without a pressure sensor, and adjusts ignition timing based on predicted values and differences in estimated knocking intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pressure sensor is used to detect knocking intensity, then measurement precision is improved, but device complexity and cost increase significantly
Solution Approach 1:
The patent creates a virtual copy of the pressure sensor's measurement capability by training a neural network to replicate the relationship between knocking sensor vibrations and pressure sensor readings. The learned model reproduces the pressure sensor's knocking intensity detection function using only the knocking sensor's output, eliminating the need for the actual pressure sensor while maintaining measurement precision.
2Device complexity
If a learned neural network is used to estimate knocking intensity from knocking sensor output, then device complexity is reduced, but measurement precision deteriorates due to unlearned engine vibrations
Solution Approach 1:
The patent extracts and removes the harmful unlearned vibration components from the knocking sensor output before feeding it to the neural network. By separating the useful knocking signal from the harmful mechanical vibration noise, the system maintains high measurement precision while using only the knocking sensor and neural network, without requiring a pressure sensor.
3Reliability
If ignition timing is excessively retarded to prevent false knocking detection, then reliability is improved, but productivity decreases due to reduced engine output
Solution Approach 1:
The patent implements a feedback mechanism where the neural network continuously monitors the knocking sensor output and adjusts the ignition timing based on the estimated knocking intensity. The system only retards ignition timing when actual knocking is detected, not during normal mechanical vibrations, thereby maintaining both reliability in knocking detection and productivity by avoiding excessive ignition timing retardation.
Data Source
AI summary
An ignition timing control device includes a storage device that stores a normal signal generation model configured to output, upon receiving an output value of a knocking sensor, a noise-removed output value therefrom which an unlearned noise component value has been removed, and a first learned neural network pre-learned to output, upon receiving one of the output value of the knocking sensor and the noise-removed output value, an estimated value of a knocking intensity representative value, and a processor that acquires the estimated value by inputting the output value of the knocking sensor to the normal signal generation model and inputting the noise-removed output value to the first learned neural network, and executes retarding control of an ignition timing based on the acquired estimated value.