Abstract: The temporal constraints decomposition model (TCD), which guarantees the global temporal, can decompose the temporal service composition into two relatively independent processes:constraints decomposition and local optimization. However, the model may lose the feasible combination scheme, which may lead to no solution when the user constraints strength is strong. This paper proposes a constraints strength-aware temporal constraints decomposition model (CIA-TCD), which introduces a relaxation factor into the existing TCD model. When the user constraint strength is weak, the global constraint can be guaranteed; and when the constraint strength is strong, a certain amount of combination scheme can be reserved, thereby increasing the probability that a feasible solution can be found. The experimental analysis shows that the CIA-TCD model has a significantly better probability of finding a feasible combination scheme than the TCD model when the constraint strength is strong.