In the paper, TWS algorithm is presented in terms of the characteristics of a 3-D phased array radar. The algorithm has been made adaptive process based on the characteristics of phased array radar’s variable scan period. The algorithm is also used to settle the problem of inconsistent variable dimensions of different target models, as well as dynamically adjust the covariance matrix of noise to get a better capability of tracking based on to the match level between moving model and system model in the IMM algorithm. Expected achievement has been made in real radar tracking experiment.