Data and R codes for the fuzzy FMEA assessment of hydroelectric earth dam failure modes

Published: 20 July 2021| Version 1 | DOI: 10.17632/43d5f6j3xc.1
Jose Roberto Ribas,


It constains two sets of R codes and their corresponding data and rules, which were used to to calculate the Fuzzy Inference System Risk Priority Number (FIS-RPN) under the Fuzzy FMEA approach for an earth dam hydropower plant located in Brazil. In the first set, we computed the RCI through the combined effect of Occurrence and Severity. In the second set, we combined the RCI with Detection to compute the FIS-RPN. In addition, it contains two Excel files, one has the FIS-RPN results of a case study, and the other has the Pearson and Kendall statistics comparing the FIS-RPN estimates of five experts.


Steps to reproduce

Step 1: The earth dam is divided into components, each having a specific function. Possible failure modes are identified for each component; Step 2: Detailed analysis of experts is applied to identify the possible chain of events whose combined effect can cause failure; Step 3: Membership functions can be defined from the user's experience and perspective, but to facilitate this task, standard membership functions, such as TMF and TRMF, are used; Step 4: The instrumentation and procedure for monitoring and early detection of risk situations that can affect the dam depend on the priority of the failure mode and the budget available; Step 5: Preventive actions are identified based on solutions adopted in similar cases, by technical recommendations found in the literature, technical opinions issued by specialized companies, or more simply by suggestions obtained through brainstorming or focus group meetings; The linguistic terms of step 6 and fuzzification of step 7 must be repeated for each of the five input variables that are part of the FMEA process: Occurrence (OCC); Severity (SEV); Detection (DET); Risk Criticality Index (RCI) and Risk Priority Number (RPN); Step 6: The following scores and linguistic terms of the OCC variable: 1 = Extremely unlikely; 2 = Remote; 4 = Occasional; 7 = Reasonably Likely and 10 = Frequent. The subjective estimate of the degree of SEV is more complex, since damage to people, the environment and the economy must be considered. To simplify, generic severity classes with respect to the population were adopted: 1 = No injures; 2 = First aid; 4 = Few injures; 6 = Injures/disabilities; 9 = Few casualties; 10 = Many casualties. The scores according to the possibility of DET are: 1 = Very high; 3 = High; 5 = Moderate; 7 = Low; 9 = Very low; 10 = Undetectable; Step 7: In this step, crisp inputs from the domain are transformed into fuzzy inputs through the membership function. The fuzzification of the context scales is carried out by TMFs and TRMFs assigned by an expert, who can be the owner’s engineer, the project or the plant manager;9 Steps 8,9 and 13: The scores for the OCC, SEV and DET variables are assigned relative to the dam and represent the expectation that the expert has about the risk situation of a given failure mode and its respective root cause; Step 10: In the present method, we adopted the “and” operator to combine the OCC and SEV levels expressed in terms of their linguistic terms. We asked the expert to guess all possible outcomes of such combination, defined as RCI, also referenced in terms of 20 levels ranging from 1 (safe) to 20 (extremely critical); Step 11: The R codes for both phase 1 and phase 2 are performed to calculate the RCI and FIS-RPN, respectively.