Abstract: This article analyzes the deficiencies of discreteness and subjectivity of the current method of risk priority number (RPN) in dealing with the three key information factors of severity, occurrence and detection of failure. Based on information entropy theory and maximum entropy inference, it demonstrates that when being used to evaluate the degree of uncertainty of risks, the substantial irrationalities of the RPN method are its dependency on the mean and the variance of the RPN function so that it cannot integrate the risk factors in proper proportion. Meanwhile, considering the uncertainty of multi-dimensional information, the problem of how to select the indexes for risk evaluation and allocate their weights for forecasting risk events is analyzed. For the realization of measuring the degree of risks consistently, a new definition and calculation by the name of the risk possibility number(RPoN)is proposed, which is different from the RPN. The formulation of the RPoN is based on and deduced from the continuity, monotonicity, and additivity assumptions of axioms of information entropy theory. The property of consistency of the RPoN approach is also proved when it is used to evaluate the degree of risks in the process of statistical decision-making. Experimental results suggest that the RPoN approach is better than the RPN method on quantitative calculation and prioritization of risks.

Key words: risk priority number, information theory, multivariable control system, entropy, measurements, risk possibility number