A fault diagnosis method for liquid-propellant rocket engines was proposed based on the dynamic cloud-BP(back propagation) neural network in the way of the integration of cloud model and BP neural network. The Cloud transform method was used to identify the network configuration and to extract the cloud features. And a unit time-delay was also introduced into the input layer to describe the dynamic characteristics of the engine. Results with test data show that the method can isolate the existed 3 fault modes precisely. A 0 expectation

0.2 standard deviation noise was used to simulate the entironmental noise and stochastic noise

and the method can still detect and classify the fault accurately acount to lasting-rule. The method can run in real-time with the single processing time being 1.124×10-4s.