Multi-perspective failure mode and effects analysis based on rough number projection

• Published in: Engineering failure analysis 2025-03, Vol.169, p.109192, Article 109192
• Main Authors: An, Hong, Xu, Guosheng, Sun, Zhenghu, Chen, Wenhua, Pan, Jun
• Format: Article
• Language: English
• Subjects: Bayesian best-worst method; Bayesian network; Motorized spindle; Multi-perspective FMEA; Rough number projection
• ISSN: 1350-6307
• DOI: 10.1016/j.engfailanal.2024.109192

•A new FMEA method for conducting a multi-perspective failure mode and effects analysis based on rough number projection.•The weights of risk factors are calculated using the Bayesian BWM.•A new method for calculating a risk decision matrix based on rough number projection.•Electric spindle failure analysis. Failure Mode and Effects Analysis (FMEA) plays a crucial role in reliability analysis and is widely used in the industrial field. Typically, FMEA methods enhance the validity of assessments by simultaneously considering severity (S), occurrence (O), and detection (D). However, this simultaneous consideration can obscure the influence of extreme risk factors on failure modes. For example, in the case of an motorized spindle, although the occurrence and detection rates of spindle deformation are low, the severity is high. When all the risk factors are considered together, this failure mode is often evaluated as medium to low risk, diminishing the significance of extreme risk. A multi-perspective analysis using combinations of SO and SD can better highlight the impact of severity and align the results with expectations. Therefore, to address the aforementioned limitations, this study proposes a multi-perspective FMEA method based on rough number projection. By evaluating different combinations of risk factors (i.e., SO, SD, OD, and SOD), this approach avoids the interaction effects of simultaneous analyses and provides more comprehensive risk information. The evaluation results for different paired risk factors were obtained using the rough number projection method and Rough Number Technique for Order Preference by Similarity to the Ideal Solution (RN-TOPSIS). These results were then used in a Bayesian network to classify risk levels based on the evaluation outcomes. Finally, this method was tested using an motorized spindle case study and compared with other methods to demonstrate its feasibility and effectiveness.