The invention discloses a method for diagnosing a compressor-associated failure based on a Bayesian network. Based on a three-layer Bayesian diagnosis network of working condition operation, failure and symptom, a simple and intuitive graphic mode is used for expatiating the complicated causality and conditional relation between the failure, the reason and the symptom thereof, performing diagnostic reasoning under a finite, incomplete and uncertain information condition and completely describing the probability of the system failure, so qualitative information and quantitative information can be integrated, the function of fully utilizing multi-source information is realized, and the testing sample size for performing the failure diagnosis is reduced by fully utilizing the priori information and historical information of a specific diagnosis object; and the accuracy and the efficiency for the failure diagnosis can be greatly improved by establishing a proper Bayesian diagnosis network and determining a rational prior probability.