The invention relates to the field of industrial object self-adaptation monitoring, in particular to a noise, correlation and time consumption three-factor coupling dimensionality reduction method, which comprises the steps of: (1) collecting signal samples; (2) calculating feature values and consumed time; (3) setting signal-to-noise ratio normal probability density distribution parameters; (4) setting monitoring signal collecting time, and calculating the time consumption of each feature under the collecting duration condition; (5) setting the total time consumption limit value and the real evaluation level; (6) setting irrelevant feature threshold value, calculating the identification rates of single features, and deleting irrelevant features; (7) setting redundant feature threshold value, calculating the correlation coefficient between two features, and deleting redundant features; (8) carrying out first-turn three-factor coupling dimensionality reduction and selecting candidate queue; and (9) taking out sets from the front part of the candidate queue for carrying out more than two turns of dimensionality reduction, and reading an optimum set and the identification rate of the optimum set. The method has the advantages that the self-adaptation monitoring on industrial objects can be realized, and the identification accuracy rate is high.