Vehicle Noise Inspection Using Neural Estimation of Vibration Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise inspection methods for vehicles are imprecise due to the difficulty in measuring vibration transfer characteristics of multiple vehicle parts associated with noise, as many parts cannot be removed for measurement, making it challenging to identify the source of unpleasant noise in a vehicle cabin.
Innovation Solution
A vehicle noise inspection apparatus utilizing a neural network that learns from measured values of original sound characteristics, evaluation sound characteristics, and vibration transfer characteristics, allowing for the estimation of vibration transfer characteristics without removing parts, and determining the cause of noise by analyzing these values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vibration transfer characteristic measurement is performed for each vehicle part associated with noise, then measurement precision is improved, but device complexity and ease of operation deteriorate due to the need to remove multiple parts for measurement
Solution Approach 1:
The patent creates a virtual model (copy) of the vehicle's vibration transfer characteristics through finite element analysis. Instead of physically measuring each component, the system builds a computational model that replicates the vibration behavior of the entire vehicle structure, allowing noise source identification through simulation rather than physical disassembly and measurement of multiple parts
Solution Approach 2:
The patent replaces the mechanical measurement system (physical sensors, actuators, and disassembly procedures) with a computational mechanics approach using finite element analysis. The vibration transfer characteristics are determined through numerical simulation rather than physical measurement, eliminating the need to remove and measure each component individually
2Measurement precision
If vibration transfer characteristic measurement is performed for each vehicle part associated with noise, then measurement precision is improved, but device complexity increases due to the need for multiple measurement setups and part removal procedures
Solution Approach 1:
The patent merges the measurement and analysis functions into a single integrated finite element model. Instead of using separate measurement systems for each component, the vibration transfer characteristics of all components are combined into one comprehensive computational model that can be analyzed simultaneously, reducing overall system complexity while maintaining measurement precision
Solution Approach 2:
The finite element model serves multiple functions: it represents the structural geometry, material properties, vibration characteristics, and transfer paths of all vehicle components in a single universal framework. This multi-functional model eliminates the need for separate measurement setups for each component, reducing device complexity while enabling precise noise inspection
Data Source
AI summary
A storage device of a noise inspection apparatus is configured to store a neural network machine-learned to receive, as inputs, an original sound characteristic value indicating a characteristic of sound generated by a transmission and an evaluation sound characteristic value indicating a characteristic of sound that reaches a vehicle cabin, and output a route part characteristic value that is a value indicating a characteristic of a vibration transfer of a vehicle part positioned on a vibration transfer route from the transmission to the vehicle cabin. An execution device of the noise inspection apparatus is configured to calculate, as an estimated value of the route part characteristic value, an output of the neural network that has received, as inputs, measured values of the original sound characteristic value and the evaluation sound characteristic value.