An improved principle components analysis method(PCA) was proposed for fault dia gn osis and faulty signal reconstruction in a maglev train, which had multiple s e nsors and redundant signals. In systems with phase shifting signals, the main co mponents acquired by the traditional PCA method were not optimal, so the process i ng effects were not convincing. To overcome this flaw, an improved idea was put fo rward and followed by a strict demonstration in theory. What’s more, the practi ca l realization process of the idea was also acquired in the demonstration proces s . Finally, adopting the squared prediction error principle, the traditional and improved PCA methods were both used in the maglev system fault detection. Whethe r the system has faulty signals or not, the conclusion is that the improved PCA m ethod can increase the probability of faulty sensor identification and decrease the error.