Research is being conducted to quantify fault trees (FTs) using physics-informed neural networks (PINNs). Based on the previous study, this work attempted to quantify FTs considering common cause failures (CCFs). The binomial failure rate (BFR) model is used to obtain the failure rate data required for continuous-time Markov chain (CTMC) which takes a role of an objective function in PINN. For better training performance, it is necessary to reduce the size of Q matrix of CTMC. The proposed method compares the PINNs among an original Q matrix, a reduced Q matrix, and an exact solution. The proposed method not only reduced learning time, but also showed improved accuracy.