Abstract
Failure mode and effect analysis (FMEA) is an effective quality management technique widely used in various industries to improve the reliability and safety of systems, products, processes, and services. In traditional FMEA, the ranking of failure modes is carried out by the risk priority number (RPN), which is calculated by the product of severity (S), occurrence (O), and detection (D). Nevertheless, the normal FMEA has many inherent defects in assessing and ranking failure modes. Therefore, in this paper, we present a new FMEA model, which integrates probabilistic linguistic term sets (PLTSs) and fuzzy Petri nets (FPNs) for the risk assessment and prioritization of failure modes. Specifically, the PLTSs are used to capture the uncertainty of FMEA team members’ subjective judgments, and the FPNs are established to acquire the risk priority of the identified failure modes. Besides, a technique for order preference by similarity to an ideal solution (TOPSIS)-based weighting method is proposed to determine the objective weight of each team member. Finally, a marine-ship system risk assessment example is provided to illustrate the suggested FMEA and a comparative analysis is conducted to assess its effectiveness and usefulness. The results show that the new FMEA approach can produce more reliable and reasonable risk ranking result of failure modes.
Introduction
Failure Mode and Effect Analysis (FMEA) is a cornerstone of quality management across industries, from aerospace to healthcare. Traditionally, engineers prioritize risks using the Risk Priority Number (RPN), a simple product of Severity (S), Occurrence (O), and Detection (D). However, the RPN method suffers from several critical flaws. It forces experts to provide precise integer scores for subjective observations, ignores the relative importance of risk factors, and can produce identical scores for vastly different risk profiles. Furthermore, it often fails to account for the varying expertise levels among team members.
To address these limitations, this research introduces a sophisticated framework that combines Probabilistic Linguistic Term Sets (PLTSs) with Fuzzy Petri Nets (FPNs). This approach moves beyond rigid numerical scoring to better capture the inherent uncertainty of human judgment and provides a mathematically robust, visual reasoning process for risk prioritization.
Methodology
The proposed FMEA model operates through a three-stage process designed to enhance objectivity and resolution in risk assessment.
1. Objective Expert Weighting (TOPSIS)
Rather than assuming all experts contribute equally, the model employs the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS). This method evaluates individual expert judgments against "ideal" (maximum information) and "worst" (minimum information) scenarios. By calculating the distance to these benchmarks, the system assigns an objective weight to each team member, ensuring that more consistent and informative feedback carries more weight in the final analysis.
2. Capturing Uncertainty with PLTS
Human experts often hesitate when assigning a single label to a risk. Probabilistic Linguistic Term Sets (PLTSs) allow experts to express this hesitation. Instead of a binary choice between "High" or "Medium" risk, an expert can state they are 80% confident a risk is "High" and 20% confident it is "Medium." This captures the nuanced belief of the evaluator, preserving data that traditional methods would discard.
3. Risk Reasoning via Probabilistic Linguistic Fuzzy Petri Nets (PLFPNs)
The core innovation is the use of Petri Nets to model risk logic. Petri nets are graphical tools used to represent system behavior and "if-then" rules. In this framework, input places represent risk factors (O, S, D), and transitions represent the reasoning logic leading to a final Risk Priority Value (RPV).
As shown in the basic structure above, a transition $t$ processes information from an input place $a$ to an output place $c$. In practical FMEA applications, this logic is expanded to aggregate multiple risk factors simultaneously.
This multi-input structure allows the model to synthesize Severity, Occurrence, and Detection into a single, nuanced priority value through concurrent reasoning, making the "black box" of fuzzy logic transparent and computationally efficient.
Case Study: Marine-Ship System Risk Assessment
The model was validated using a real-world case study involving a shipboard oil spill. A team of five marine experts evaluated 12 failure modes (FMs), ranging from steering gear breakdown (FM1) to poor coordination (FM12).
Key Results and Insights
- High Resolution Ranking: Unlike traditional RPN, which often results in "clumping" (where multiple failure modes receive the same score), the PLFPN model successfully ranked all 12 failure modes without ties. For instance, while FM2 and FM12 appeared identical under traditional RPN, the new model correctly prioritized FM2 as a higher risk based on the nuanced expert data.
- Critical Failures Identified: The analysis identified "Incomplete information given to local maritime authorities" (FM11) and "List of duties demonstrated in muster list were not properly fulfilled" (FM6) as the most critical risks.
- Stability: Sensitivity analysis across 10 different scenarios showed that while middle-tier rankings shifted with factor weights, the most critical failure (FM11) remained consistently at the top, proving the model's robustness.
Conclusion
By integrating PLTS and Fuzzy Petri Nets, this new FMEA approach provides a more reliable and reasonable risk ranking than traditional methods. While the mathematical complexity is higher, the model's ability to handle expert hesitation and provide clear, visual reasoning makes it a powerful tool for safety-critical industries. Future work will focus on developing user-friendly software tools to bring this advanced logic to non-expert users and exploring dynamic versions of the model for real-time system monitoring.
Summary updated