The complex characteristics of multi-unit and strong corrections among different units in modern industrial processes have imposed great challenges to process modeling and monitoring. To solve these issues, a new multi-block graph convolutional network monitoring model based on Pearson correlation coefficient is proposed (MBGCN) for multi-block process monitoring in this study. Firstly, graph data is constructed to describe the intricate correlations of the high-dimensional process data. The adjacency matrix of the graph structure is initiated via K-Nearest Neighbor, and a graph autoencoder is utilized to automatically learn the interactions between samples. Secondly, the Pearson correlation coefficient is employed to divide the process into sub-blocks. Graph convolutional neural network is applied to learn effective feature representations for constructing monitoring statistics. Finally, the monitoring results of each sub-block are fused through Bayesian inference. The experimental results in Tennessee Eastman Process illustrate the effectiveness and superiority of MBGCN in multi-block process fault detection.