The invention discloses a method and a system for predicating self-adaptive disk array failure, which is characterized in that: based on a back-propagation neutral network, historical data of a parameter capable of characterizing the health state of disk array is collected by means of self-learning mechanism of the neutral network, so the rule characterizing the variation of the parameter along with time can be summarized by an approximation method, further predicating the value of the parameter. In the method, three parameter collecting policies from low level to high level, i.e. the first, second and third parameter collecting policies, are set in total, the number of the parameters increases gradually as the time granularity decreases gradually, and a universal calculation model for calculating the health state of disk array according to predicated parameter values and parameter failure thresholds are provided, therefore, the failure condition of disk array can be effectively predicated. The self-adaptive disk array failure predicating system comprises an information acquisition module, a self-adaptive configuration module, a data storage module, a self-adaptive index predication module, a parameter failure threshold module, a user management module and a user interface module.