Real-time fault detection and fault diagnosis is a key part of building smart factories. Principal component analysis (PCA) and kernel principal component analysis (KPCA) are the two most commonly used methods for process monitoring. Nevertheless, PCA cannot deal with nonlinear data and KPCA cannot be applied for large data sets. To tackle the above two problems, a novel fault detection and fault diagnosis method based on Gaussian-Bernoulli restricted Boltzmann machine (GBRBM) is proposed in this paper. The key idea of our approach is using GBRBM as a nonlinear dimensionality reduction technique and applying the reconstruction error to establish monitoring SPE statistic. The upper limit of the SPE is estimated by kernel density estimation (KDE). If the SPE statistic of an online sample exceeds the control limit, then the fault is detected and the contribution plot is utilized to diagnose the fault. Tennessee Eastman (TE) process is applied to evaluate the fault detection and diagnosis performance of the proposed method. The effectiveness of the proposed GBRBM monitoring method is improved by saving significant memory requirement while obtaining the highest average false detection rate and comparable fault diagnosis reliability, relative to PCA and KPCA techniques.