Correlation Analysis of Noise, Vibration, and Harshness in a Vehicle Using Driving Data Based on Big Data Analysis...

Area of Science:

• Automotive Engineering
• Data Science
• Machine Learning

Background:

• Traditional vehicle noise, vibration, and harshness (NVH) development is time-consuming and costly.
• Existing methods often overlook the nonlinearity of dynamic characteristics and discard data.
• There is a need for efficient techniques to identify system-NVH factor relationships in automotive development.

Purpose of the Study:

• To present a new vehicle NVH development process utilizing data analysis and machine learning.
• To reduce vehicle development time and cost by accurately identifying system-NVH factor connectivity.
• To incorporate nonlinearity and utilize all available data in the NVH analysis.

Main Methods:

• Exploratory data analysis (EDA) on long-term NVH driving data.
• Application of variable importance, correlation, and sensitivity analyses.
• Utilizing whole big data, including nonlinear dynamic characteristics, without discarding any data.

Main Results:

• Accurate identification of relationships between vehicle systems and NVH factors.
• Efficient pinpointing of areas requiring improvement through correlation and variable importance analysis.
• Quantification of the impact of system NVH changes on cabin noise via sensitivity analysis.

Conclusions:

• The proposed method significantly improves efficiency in vehicle NVH development.
• This approach can accelerate the identification of necessary improvements, reducing overall development time.
• The technique is applicable to other machine learning model validations and represents a future direction for AI-driven automotive development.