For probabilistic monitoring of nonlinear processes, the traditional probabilistic principal component analysis (PPCA)-based monitoring method is generalized through the kernel method. Thus, a probabilistic kernel PCA method is proposed for process monitoring in the present paper. Different from the traditional PPCA method, the new approach can successfully extract the nonlinear relationship between process variables. On the basis of the proposed nonlinear probabilistic monitoring approach, the monitoring performance of nonlinear processes can be effectively improved. To demonstrate the feasibility and efficiency of the proposed method, a case study on the Tennessee Eastman (TE) benchmark process is provided.