Reducing Harmful Effects on Human Health of Car Vibrations using a Combination of Unsupervised and Supervised Machine Learning Algorithm

Abstract

Car vibrations are typically common, but it’s crucial to determine the level of shaking that signifies a potential threat to human health. In this research, the machine learning algorithm was employed to predict potential failures in car suspension systems, with the goal of enhancing vehicle reliability and safety. A combination of unsupervised and supervised machine learning algorithms is applied to classify data, isolate outlier, and predict fault diagnosis outcomes in the simulation context with a dataset of a car suspension system.  In general, either unsupervised or supervised learning techniques can contribute to a valuable predictive maintenance when dealing with larger datasets in a car suspension system either usual or unusual conditions. Therefore, the results may not cover unknow patterns within unusual conditions, such as high amplitude sinusoidal and step road profile. In this research, a combination of unsupervised and supervised learning techniques is proposed to identify car suspension issues caused by poorly maintained vehicles, potentially impacting human health. The findings from the simulation indicate that a combination of unsupervised and supervised machine learning algorithms can effectively classify fault diagnosis and isolate unknow patterns for future training models. Moreover, the identification of non-groupable elements using a combination of these learning techniques is illustrated through outlier detection on unknown patterns. Therefore, the research results enable engineers to assess devices for maintenance and repair needs, aiming to reduce or eliminate car vibrations, which can pose serious harm to human health.