STRUCTURAL HEALTH MONITORING USING STATISTICAL SYSTEM IDENTIFICATION TECHNIQUE FROM LIMITED DYNAMIC RESPONSE: Monitoring of structural health is important from safety consideration. System identification techniques, using inverse dynamic approach, are important tools to improve the mathematical models for structural health monitoring. However, uncertainties in the measured data might lead to unreliable identification of damage in structural system. Degree of rigidity in joints of framed structure, which are normally not considered with sufficient accuracy in the finite element models, plays an important role in the analysis. In practice, most of the structures are with semirigid frames having partially restrained joints. Thus, the identification model without considering the accurate joint rigidity may reduce the quality of the estimated parameters. The measurement of dynamic responses at all degrees of freedom of a structure is also not feasible in practice. An analysis methodology is proposed for structural health monitoring of semi-rigid framed structure in the framework of finite element model with limited dynamic responses. The proposed numerical method includes two-stage model updating processes namely model tuning and damage identification. The structural properties viz. axial rigidity, bending rigidity, joint rigidity and mass density per unit length are identified at the element level in the updated models of the system. Damage at the element level is identified in a probabilistic manner, by comparing the identified structural parameters of the updated model of the system with those of the undamaged state. Statistical system identification technique is employed considering the uncertainties associated with the measured dynamic response data. System equivalent reduction and expansion process (SEREP) technique is adopted to simulate the full modal data from the limited dynamic responses measured at selected degrees of freedom experimentally. The proposed model is suitable for practical problem, as it is able to identify the structural parameters with limited modal data of first few modes, measured at selected degrees of freedom. Different numerical examples with various damage scenarios are explored to demonstrate the applicability of the proposed model. The model is able to identify the structural damage with great accuracy from the noisy dynamic responses even if the extent of damage is very small. Experimental studies, on simple cantilever beams, establish the potential of the proposed method for its practical implementation.