Machine Learning-Based Prediction and Multi-Algorithm Optimization for Reducing Vertical Acceleration and Enhancing Two-Wheeler Ride Comfort

Abstract

The suspension system has a major impact on the rider's comfort and stability and is essential in minimizing vibrations that are conveyed to them. Uncomfortable, exhausting, and perhaps dangerous situations can result from excessive vertical acceleration brought on by uneven roads. In order to analyze and optimize the impact of road surface, speed, and suspension stiffness on vertical acceleration, this study used machine learning regression models and optimization approaches in an experimental examination.

In a total of twenty-seven trials, three separate suspension springs were put through their paces on three separate road surfaces.  To ensure precise data collection, a sitting pad accelerometer and a Svantek FFT analyzer were used to determine the maximum vertical acceleration and RMS acceleration.  Maximum vertical acceleration was predicted using three regression models: Support Vector Regression (SVR), Random Forest Regression (RFR), and Response Surface Methodology (RSM).  In order to find the sweet spot for suspension stiffness, road surface type, and vehicle speed that would minimize maximum vertical acceleration, optimization methods like Genetic Algorithm (GA) and the Nelder-Mead approach were employed.

The findings demonstrate a robust link among suspension stiffness, road surface imperfections, and speed in affecting vertical acceleration. The Random Forest Regression model attained a R² score of 0.30 and a root mean square error (RMSE) of 0.475, whereas the RSM exhibited superior performance with a R² score of 0.66 and an RMSE of 0.563. Nonetheless, SVR demonstrated subpar performance, evidenced by a R² score of -0.61 and an RMSE of 0.722. The refined suspension configuration resulting from these models decreased the maximum vertical acceleration from 1.72 m/s² to 1.656 m/s², enhancing ride comfort by about 3.68%.

This work systematically improves two-wheeler suspension systems for greater ride quality, fatigue reduction, and safety. Automotive engineers, manufacturers, and researchers enhancing car suspension performance in real-world driving situations will benefit from the findings.