Ensemble modeling for predicting head vibration based on driving seating conditions: Towards adaptive seating systems

Personalized adaptive seating systems represent a frontier in automotive engineering, promising to enhance driver comfort and safety by adapting to individual characteristics and driving seating conditions. This study investigated the application of ensemble modeling techniques to predict tri-axial...

Full description

Saved in:
Bibliographic Details
Published in:Engineering applications of artificial intelligence Vol. 145; p. 110174
Main Authors: Singh, Amandeep, Nawayseh, Naser, Rakheja, Subhash
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.04.2025
Subjects:
Automotive ergonomics
Ensemble modeling
Explainable artificial intelligence
Feature importance
Personalized seating systems
Postural comfort
Shapley additive explanations
Whole body vibration
Personalized seating systems
Whole body vibration
Feature importance
Automotive ergonomics
Shapley additive explanations
Postural comfort
Explainable artificial intelligence
Ensemble modeling
ISSN:0952-1976
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Personalized adaptive seating systems represent a frontier in automotive engineering, promising to enhance driver comfort and safety by adapting to individual characteristics and driving seating conditions. This study investigated the application of ensemble modeling techniques to predict tri-axial head vibration in drivers along the fore-and-aft (X), lateral (Y), and vertical (Z) axes. We addressed the complex interactions between individual characteristics, seating condition, and whole-body vibration exposure. We analyzed data from 15 male subjects (mean age: 21.93 ± 1.03 years; body mass index: 24.13 ± 2.62 kg/m2) across various seating conditions. Statistical analysis showed significant effects of seating conditions and participant characteristics on acceleration responses across the X, Y, and Z axes (p < 0.05). We employed Random Forest, Bagging, Gradient Boosting, and Adaptive Boosting algorithms to predict tri-axial acceleration responses. The Random Forest model outperformed others, achieving R-Squared values of 0.879, 0.922, and 0.871 for X, Y, and Z axes, respectively. Feature importance analysis demonstrated that combinations of anthropometric and seating variables were most influential, with importance scores up to 0.99 for key feature combinations. The analysis emphasized varying impacts of features across axes, with body mass index and backrest angle showing consistently strong influences. The findings demonstrate the superiority of ensemble methods over traditional approaches in capturing nonlinear relationships affecting driver comfort and safety, offering potential for real-time monitoring and personalized ergonomic interventions in automotive engineering. •Developed interpretable ensemble model for predicting head vibrations.•Random Forest model outperformed others in predicting tri-axial head vibration.•Seating conditions significantly affected acceleration responses.•BMI and backrest angle showed strong influence across all vibration axes.
ISSN:0952-1976
DOI:10.1016/j.engappai.2025.110174