Neural Network Knocking Strength Estimation for Engine Vibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knocking detection systems in internal combustion engines face challenges in accurately distinguishing between knocking vibrations and mechanical vibrations, leading to incorrect knocking strength judgments, while pressure sensors provide precise detection but are expensive and prone to deposit buildup, making them unsuitable for widespread use.
Innovation Solution
A knocking detection system utilizing a neural network to estimate knocking strength from the output values of a pressure sensor, where the neural network is trained using data from both a knocking sensor detecting engine body vibrations and a pressure sensor, allowing for precise knocking strength estimation without the need for a pressure sensor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a knocking sensor is used to detect vibration of the engine body, then the knocking strength can be detected, but mechanical vibrations from valve seating and fuel injector operations cause false readings and reduce measurement precision
Solution Approach 1:
The patent extracts the harmful mechanical vibration components from the knocking sensor signal by comparing it with reference vibration patterns from valve seating and fuel injector operations. This allows isolation of the true knocking signal for accurate strength measurement.
Solution Approach 2:
The patent introduces a neural network as an intermediary that processes both the knocking sensor signal and pressure sensor data to estimate knocking strength. This intermediary learns to distinguish true knocking from mechanical vibrations through training, resolving the interference problem.
2Measurement precision
If a pressure sensor is used to detect pressure inside the combustion chamber, then knocking strength can be precisely detected without mechanical vibration interference, but the sensor becomes extremely expensive and prone to deposit buildup
Solution Approach 1:
The patent creates a virtual copy of the pressure sensor's knocking detection capability by training a neural network to replicate its function using data from both the pressure sensor and knocking sensor. This allows the system to achieve pressure-based accuracy without requiring the expensive physical pressure sensor during operation.
Solution Approach 2:
The patent replaces the expensive, maintenance-prone pressure sensor with a combination of a durable knocking sensor and a software-based neural network estimator. This substitution uses cheaper, more reliable hardware components while maintaining detection accuracy through intelligent processing.
3Duration of action of moving object
If a pressure sensor is used for long period operation, then continuous monitoring is possible, but deposit buildup on the sensor changes combustion characteristics and reduces reliability
Solution Approach 1:
The patent performs preliminary training of the neural network using data from the pressure sensor before deployment. During this training phase, the pressure sensor is used to establish the relationship between pressure fluctuations and knocking strength. Once trained, the system operates using only the durable knocking sensor, avoiding the deposit buildup problem during long-term operation.
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
This approach enables accurate detection of knocking strength with high precision, avoiding the limitations of mechanical vibrations interference and the high cost and maintenance issues associated with pressure sensors, thus providing a cost-effective solution for precise knocking detection.
Implementation Method 1
weights of the neural network are learned by the learning device by using the value representing vibration of the engine body detected by the knocking sensor as an input value of the neural network
Implementation Method 2
a knocking sensor detecting vibration of an engine body
Implementation Method 3
a pressure sensor detecting a pressure inside a combustion chamber of the internal combustion engine
Data Source
AI summary
A knocking sensor detecting vibration of an engine body and a pressure sensor detecting a pressure of a combustion chamber are provided. A value representing the knocking strength is acquired from output values of the pressure sensor. Weights of a neural network are learned using a value representing the vibration of the engine body detected by the knocking sensor as an input value of the neural network and using the acquired value representing the knocking strength as training data. The value representing the knocking strength is estimated from the output values of the knocking sensor by using the learned neural network.