Abstarct: The extensive literature on structural health monitoring (SHM) has documented the critical importance of detecting damage in civil engineering systems at the earliest possible time. Bridges is one of most important infrastructure, hence monitoring the condition of bridges and other civil infrastructures is necessary in an effort to develop cost-effective lifecycle maintenance strategies. In this study, a novel method for intelligent damage detection of steel bridge, baxsed on least square support vector machine (LS-SVM), is proposed. Damage is modeled as a reduction of elasticity modulus as key parameters in stiffness matrix of structural elements. Results are compared with same system baxsed on radial basis function neural network (RBFNN). Severity and location of damages are obtained baxsed on variations of mode shape between the analytical models and the responses measured in damaged models. To demonstrate the ability of proposed method for detection of damage in bridge, different types of steel bridge are considered. The results show that the proposed method baxsed on LS-SVM has better results in detecting the exact locations and the severity of damages in comparison with RBFNN, without affecting from noise. Furthermore, baxsed on result, the mode shape differences or the mode shape ratios between before and after damage are more sensitive to damage than the mode shapes themselves.