In order to provide better technical support for the condition monitoring and health management of the overall performance of main engine, the exhaust gas temperature baseline model based on the combination of deep belief network (DBN) and support vector regression (SVR) was proposed. By calculating the deviation between the measured value and the baseline value, and combining the change of the deviation, the degradation or fault of the components could be judged. Taking the navigation data of a ship from January to June 2015 and the data in the trial report as the initial sample, the final dataset was formed after processing as abnormal value elimination, noise removal and stability point extraction, etc. The DBN was used to extract the features of the dataset, and then the extracted data features were input into the SVR to establish the baseline model of main engine exhaust gas temperature. The results show that by verification of the test set, the DBNSVR model based on parameter optimization can predict the state parameters more accurately than BP model, DBN model and SVR model, and has excellent stability.