<p>A software-defined vehicle (SDV) utilizes a centralized control computer and a new electrical/electronic architecture to enhance vehicle performance, safety, and user experience. In this study, we discuss a prognostics and health management (PHM) strategy for detecting failures caused by degradation in legacy polymer suspension components, which are indispensable to SDVs, and parts whose polymer properties significantly affect operational performance. In Part 1, the domain knowledge of target components as well as the characteristics of the physical model and failure modes was analyzed. In this paper, which is Part 2, an empirical investigation was conducted based on measurement data related to the failure phenomena of target components. Various signal processing techniques were selected and applied to extract key features with high discriminative power from external noise signals (e.g., road excitations, driving conditions). This serves as the first step in developing physics-informed signature learning models for AI that incorporate domain expertise. Through efficient preprocessing of the raw signal data, this approach provides concrete measures for completing a pipeline to achieve high diagnostic accuracy with minimal computational resources.</p> Graphical Abstract <p></p>