Silvia Carpitella

A combined multi-criteria approach to support FMECA analyses: A real-world case

Abstract The paper proposes an approach that combines reliability analyses and multi-criteria decision methods to optimize maintenance activities of complex systems. A failure mode, effects, and criticality analysis (FMECA) is initially performed and the fuzzy TOPSIS (FTOPSIS) method is then applied to rank previously identified failure modes. For prioritization, failure modes are assessed against three evaluation criteria that differ from those traditionally involved in risk priority number (RPN) computation (i.e. severity, occurrence and detection). Two criteria refer to the maintenance management reflecting the operational time taken by the maintenance activity performed after the occurrence of a specific fault, and the way such an action is executed. The third criterion reflects the classical frequency of the occurrence of faults. To further develop previous research, the analytic hierarchy process (AHP) is herein applied to weight evaluation criteria and a group of experts is involved with aspects associated with the considered criteria. The approach is applied to a real-world case study, showing that the obtained results represent a significant driver in planning maintenance activities. To test the influence of criteria weights on ranking results, a sensitivity analysis is carried out by varying the vector of criteria weights obtained from the group decision process.

k-out-of-n systems: An exact formula for the stationary availability and multi-objective configuration design based on mathematical programming and TOPSIS

Reliability and availability analyses are recognized as essential for guiding decision makers in the implementation of actions addressed to improve the technical and economical performance of complex systems. For industrial systems with reparable components, the most interesting parameter used to drive maintenance is the stationary availability. In this regard, the present paper proposes an exact formula for computing the system stationary availability of a k-out-of-n system. Such a formula is proved to be in agreement with the fundamental theorem of Markov chains. Then, a multi-objective mathematical model is formulated for choosing the optimal system configuration design. The Pareto front is developed using the Lexicographic Goal Programming (LGP) method, and the TOPSIS method is successively implemented to choose the k-out-of-n configuration that represents the best compromise between the considered objective functions. A numerical example is provided. Stationary availability formula proposed for k-out-of-n systems.Validation of the proposed formula by Markov chains.Design of a k-out-of-n system.Selection of Pareto solution by means of TOPSIS multi-criteria method.

Consistent clustering of entries in large pairwise comparison matrices

Abstract In multi-attribute decision making the number of decision elements under consideration may be huge, especially for complex, real-world problems. Typically these elements are clustered and then the clusters organized hierarchically to reduce the number of elements to be simultaneously handled. These decomposition methodologies are intended to bring the problem within the cognitive ability of decision makers. However, such methodologies have disadvantages, and it may happen that such a priori clustering is not clear, and/or the problem has previously been addressed without any grouping action. This is the situation for the case study we address, in which a panel of experts gives opinions about the operation of 15 previously established district metered areas in a real water distribution system. Large pairwise comparison matrices may also be found when building comparisons of elements using large bodies of information. In this paper, we address a consistent compression of an AHP comparison matrix that collapses the judgments corresponding to a given number of compared elements. As a result, an a posteriori clustering of various elements becomes possible. In our case study, such a clustering offers several added benefits, including the identification of hidden or unknown criteria to cluster the considered elements of the problem.