A damage detection algorithm based on a model-free technique and surface strain data is developed to detect debonding damage in stiffened composite laminates. The model-free detection algorithm is based on the robust regression analysis and the strain data are acquired by using the digital image correlation technique. A series of experiments are carried out to establish a database composed of surface strain data from healthy and damaged specimens with different damage sizes. The experimental data are employed to evaluate the damage detection algorithm. The results show that the algorithm can timely detect the damage irrespective of the damage size. The influence of key parameters that may affect the performance of the algorithm such as the number of sensors, arrangement of sensors, and the noise level is also studied. A stable and reliable detectability is found when certain number of sensors is utilized with a limited noise level. However, there is no clear correlation between the arrangement of the sensors and the performance of the algorithm.