Abstract: The faults of gas valves often happen for reciprocat- ing compressors. Because of the complex mechanism and insu- fficient fault samples, it is often difficult to detect the faults of gas valves effectively for common detection methods. In order to detect the faults of gas valves accurately, the abnormal dete-ction problem to equipment is described and some terms (for e-xample, state space of equipment, self-nonself space and fuzzy space) are introduced based on the negative selection mechani-sm of natural immune system and the negative selection algor-ithm. Consequently, the approach suit for detecting the faults of gas valves is proposed. The result of detection for the common faults of gas valves shows the approach can efficiently detect the faults of gas valves in the way of abnormality curve. This shows the approach is valid. The fault detection approach based on immune mechanism can detect the faults of equipment by learning normal data without fault data. It can efficiently detect the faults of the equipment that lacks fault data.

Key words: Algorithm, Fault detection, Gas valves, Immune mechanism, Reciprocating compressors