Kevlar woven fabric is crucial for the lightweight design of aerospace structures due to its unique advantages. However, it is susceptible to damage during operation, making effective damage detection essential for ensuring structural reliability. Nevertheless, conventional sensors used for damage detection have limitations such as complex manufacturing processes and weight. To address this issue, this paper proposed a simple singlemode-multimode-singlemode (SMS) optical fiber sensor to collect vibration signals and construct a damage recognition model based on natural frequency in order to identify the size of damages in Kevlar woven fabrics. To achieve a highly linear and accurate sensor, a length determination criterion for multimode optical fiber segments based on quasi-image, notch, and transmission spectrum loss was proposed. An optical sensor was fabricated based on this criterion. The feasibility and sensitivity of the sensor for vibration signal measurement were verified by constructing a vibration signal measurement device. An experimental study was conducted to detect damage in Kevlar tapes, where the natural frequency changes measured by the SMS sensor were integrated with an optimized multi-variable gray model for damage size detection. The experimental results indicate that the estimated error in damage size is 0.71%. The damage identification system proposed in this paper offers a simple and cost-effective solution for measuring damage in flexible structures using fiber optic sensors.