<p>Damage detection, such as cracks, in beam-like structures remains a significant challenge, particularly in scenarios where baseline data from the intact structure are often unavailable in real-world applications. This study addresses this issue by introducing a new damage index, referred to as a refined normalized curvature damage factor (RNCDF), aimed at accurately identifying and localizing cracks without requiring any prior reference information. The proposed method utilizes mode shape curvature derived directly from the damaged structure and applies a Euclidean norm-based transformation (norm-2) to enhance the sensitivity of the damage identification. By eliminating the dependence on baseline data, this approach provides a simple computation approach and cost-effective solution for early damage detection. The methodology was validated through both numerical simulations and experimental tests on a cantilever beam suffering from single- and double-crack scenarios. The results confirmed the method’s high accuracy in localizing crack positions and its robustness to suppress the noise over the traditional curvature-based techniques. These findings highlight the potential of the proposed method as a reliable and practical tool for structural health monitoring, particularly in applications where baseline data are unavailable or difficult to be obtained.</p>