Composite T-joints are widely used to join composite structures. However, the performance and reliability of composite T-joints rely on the integrity of the adhesive bonds within the joint. The disbond at the interface of the adhesive bond could cause severe failure of the whole structure. Detecting the disbond before it develops through the whole interface will significantly decrease the risk of failure. Vibration and optic fiber strain measurement methods have been studied extensively for health monitoring of the T-joint. However, those methods are not sensitive to the localized defects. The ultrasonic guided wave technique is a promising tool for detecting localized defects. As an ultrasonic guided wave has multiple propagating modes, identifying optimal mode-frequency combinations is critical. In this work, a mode identification algorithm is developed to find mode-frequency combinations for health monitoring of composite T-joints. Theoretical analysis is performed to predict optimal frequency-mode combinations. Finite element analysis is used to verify the prediction.