Aiming at the problem that the fault detection and diagnosis software of a certain type liquid rocket engine lack the function of fault detection in the transient section(start-up and end-up) and fault diagnosis can not give the degree of fault, the fault magnitude is quantitatively determined by introducing the fault factor zm_i on the basis of the original software fault diagnosis algorithm-cosine similarity classification, which realizes qualitative and quantitative analysis of engine fault diagnosis. At the same time, designing and developing the envelope algotithm based on statistics by Python language so that the original software can realize transient section fault detection function of rocket engine. The algorithm of software is validated by simulation test data. And the results show that the optimized fault diagnosis function can calculate the degree of fault and the fault detection function based on envelope algorithm improves the efficiency of rocket engine fault detection.