The aim is this paper is to study fault diagnosis in a continuous chemical process. An experimental system is built to be the research base and a model is proposed and trained to carry out the fault diagnosis in the process. A Bayesian network model with two-layer nodes structure is designed and Maximum likelihood estimation (MLE) is used to amend the conditional probability table (CPT) given by expert knowledge. Then a Monte Carlo method is applied to simplify the inference rules and the data samples collected from the experimental system has been used to test the model.