Fault monitoring of bioprocess was extremely important to ensure the safety of a reactor and consistently high quality of products. It was difficult to build the accurate mechanistic model of bioprocess, fault monitoring based on rich historical or online database was a effective way. A group of data based on bootstrap method could be re-sampled stochastically, which could improve generalization capability of the model. In this paper, on-line fault monitoring of generalized additive models (GAMs) combining with bootstrap was proposed in glutamate fermentation process. In the experiments, firstly, GAMs and bootstrap were used to decide confidence interval area based on the online and offline normal sampled data from glutamate fermentation experiments. Secondly, GAMs were used to on-line fault monitoring with only time (T), dissolved oxygen (DO), oxygen uptake rate (OUR) and carbon dioxide evolution rate (CER). The proposed method could provide on-line accurate fault alarm, and was help to provide useful information for removing fault and recover abnormal fermentation.