FMEA[VisualPro Tech Insight] Leveraging AI for AIAG-VDA FMEA Part 2: Establishing Objective Risk Rating Criteria

 Failure Mode and Effects Analysis (FMEA) is a classic yet enduring methodology widely recognized for its utility. It systematically predicts potential failures in products or processes and analyzes their effects to mitigate risk, thereby saving both cost and time. While the automotive industry traditionally adapted FMEA to individual company standards, the globalization of the sector drove the need for a unified global standard. Consequently, the Automotive Industry Action Group (AIAG) and the German Association of the Automotive Industry (VDA) harmonized their approaches, publishing the AIAG-VDA FMEA handbook in 2019. <What is AIAG-VDA FMEA?>

<The Official AIAG-VDA FMEA Handbook>

 Today, we will discuss how VWAY’s VisualPro utilizes AI technology to address the vulnerabilities of traditional analysis methods within the AIAG-VDA framework, assisting users and outlining the future direction of analysis tools. 

1. The Challenge in FMEA: Data Reliability Issues Due to Subjectivity 

• In the AIAG-VDA FMEA 7-Step Process, Step 5, "Risk Analysis," is a critical stage. It involves evaluating Severity (S), Occurrence (O), and Detection (D) based on the previously identified failure network to determine the Action Priority (AP). 
• However, this stage is also where human subjectivity is most prevalent. Even for identical failure modes, scores often vary—one engineer might assign a 7 while another assigns a 5—depending on their disposition, experience, or even their daily condition.
• This inconsistency in evaluation criteria ultimately degrades the reliability of AP results and creates ambiguity in the organization's overall risk management standards. 

<Evaluators assessing the same failure based on differing subjective criteria>

2. Key Solution: AI-Driven Objective SOD Ratings and Rationale Roadmap 

 To solve this subjectivity problem, VisualPro is developing AI capabilities that use Natural Language Processing (NLP) to analyze written failure information. The AI recommends the most appropriate SOD ratings and provides the rationale behind those judgments. 

Here are examples of the features being implemented: 

• Context-Based Severity Judgment: The AI analyzes the nuance of the text described in the "Failure Effect."  For instance, if critical keywords like "threat to life" or "regulatory violation" are detected, it proposes a Severity of 9-10 according to AIAG-VDA standards. Conversely, for issues like "cosmetic damage" or "noise generation," it suggests lower ratings, ensuring evaluation consistency. 
• Occurrence (O) & Detection (D) Proposals Linked to Controls: The AI evaluates the technical maturity of the "Prevention Controls" and "Detection Controls" input by the engineer (e.g., visual inspection vs. automated sensor inspection) and proposes ratings that align with the handbook's criteria. 
• Decision Support via Rationale: Rather than simply providing a score, the system displays the "Why"—mapping the proposed score to the specific criteria in the handbook.
• Engineers can simply review the AI's proposals to make a final decision. This significantly reduces evaluation time, ensures data is backed by evidence, and accumulates this information as a valuable corporate asset.

<Context-based assessment and suggestion of S, O, D ratings>

3. Building a Worry-Free AI Environment: On-Premise & Frontier Models 

 VWAY supports two AI integration methods tailored to our customers' data security policies and requirements: 

• Frontier Model Integration: We offer API integration with the latest high-performance AI models (such as GPT-5, Claude) to deliver peak performance.
• Internal Closed Network (On-Premise) Deployment: For sectors with high data sensitivity (Defense, Automotive, Aerospace), we deploy security-enhanced sLLMs (Small Large Language Models) directly on the customer's internal servers. This supports safe, local AI functionality without any external data exchange.

4. Conclusion: The Future of Safety Analysis with VisualPro

• Supporting diverse safety analysis techniques such as STPA, FMEA, FTA, and TARA, VisualPro is evolving from a mere "tool" into an "Intelligent Partner" for engineers.
• By leveraging AI for repetitive and complex tasks, engineers can save time and streamline procedures, allowing them to focus on the more critical tasks of safety design and verification.
• VWAY is committed to building the most convenient and secure AI analysis environment optimized for your data landscape.

<VisualPro: TUV-SGS Certified Software for ISO 26262 Functional Safety Analysis Support>