Probabilistic priority numbers for failure modes and effects analysis

The purpose of this paper is to propose a method, derived from numerical evaluations on the criteria of security, frequency and detectability, of Failure Modes and Effects Analysis (FMEA), a probabilistic priority measure for potential failures; and to evaluate the use of this method when combined with subjective evaluations to decide on improvement actions.

The method proposed is based on treating the numerical initial measurements as estimates of location parameters of probability distributions. This allows for objectively taking into account the uncertainty inherent in such measurements and to compute probabilities of each potential failure being the most important according to each criterion. These probabilities are then combined into a global quality measure, which can be interpreted as a joint probability of choice of the potential failure.

The results obtained in the cases studied show the suitability of the changes proposed. Thresholds levels proposed for the discretization of the probabilistic scores are also shown to be able to allow for an efficient combination with experts' evaluations.

The approach here developed allows for the introduction of statistical parameters in the first stage of FMEA modeling. Employing a more complete model leads to greater reliability of the methodology.

The more precise modeling asks for a certain degree of practical knowledge of the factors effectively introducing variability in the measurements. This need may be surpassed by the choice of distributions such as those here employed.

The evaluation of the potential failures by the probability of being the most important, according to each criterion, is new in FMEA.