A method based on hidden graphic evaluation and review technique (HGERT) for system reliability prediction and assessment is proposed for complex equipment with degradation failure. Nodes in GERT network represents state transmission from a normal state to a failure state in a failure process. The performance-characteristic-parameters of hidden states in the system are quantized and input through the sensors, so that each deteriorated state node in GERT network can be expanded. Viterbi algorithm is used to decode and study the model parameters, and then the state transition probability of degradation failure system is determined. Working stress, environment load and other factors which influence the performance of equipment operation are considered as input, performance degradation of critical indicators as output, a general control model reflecting equipment performance degradation can be established and then we use degradation as a random variable transmitted between nodes in GERT network. The parameters are analyzed by means of signal flow graph-theory, the performance degradation level in different states can be monitored, which provides support for fault prognosis and health management. Feasibility and rationality of the model is proved by analyzing the performance degradation of a milling machine.