fmea-reviewer

name	fmea-reviewer
description	AIAG & VDA (1st Edition, 2019) FMEA review, qualitative risk evaluation, and deterministic Action Priority (AP) scoring via quality-mcp.

FMEA Reviewer: AIAG & VDA (2019) Risk & Action Priority Evaluation

Overview

The fmea-reviewer skill guides AI agents in conducting structured, standards-compliant Failure Mode and Effects Analysis (DFMEA, PFMEA, and FMEA-MSR) reviews according to the AIAG & VDA FMEA Handbook (1st Edition, 2019), specifically Section 3.5.9 (Action Priority (AP) for DFMEA and PFMEA).

Historically, FMEA evaluations relied on the classic Risk Priority Number ($\text{RPN} = S \times O \times D$), applying arbitrary thresholds (e.g., $\text{RPN} > 100$) that treated Severity, Occurrence, and Detection with equal arithmetic weight. This legacy practice allowed critical, high-severity failure modes with moderate or low occurrence to be masked by low RPN values. The AIAG & VDA 2019 standard replaces arbitrary RPN cutoffs with the Action Priority (AP) logic table, establishing a systematic, non-linear hierarchy that prioritizes Severity first ($S \to O \to D$).

This skill equips agents to:

• Decompose complex designs and manufacturing processes across the 7-Step FMEA methodology.
• Elicit and critique failure chains (Failure Effect $\to$ Failure Mode $\to$ Failure Cause).
• Distinguish between prevention controls (affecting Occurrence) and detection controls (affecting Detection).
• Delegate all quantitative risk scoring ($S$, $O$, $D$ validation, RPN calculation, and AP table lookup) to the deterministic lookup_fmea_ap tool on quality-mcp.
• Synthesize actionable optimization and containment recommendations based on deterministic AP levels (High, Medium, Low).

When to Use

Activate this skill in the following quality engineering scenarios:

• FMEA Document Audits: Auditing existing Design FMEAs (DFMEA) or Process FMEAs (PFMEA) for completeness, structural coherence, and standards compliance.
• Risk Assessment & Scoring: Evaluating risk levels for newly identified potential failure modes during product design or process engineering.
• Legacy FMEA Migration: Migrating legacy AIAG 4th Edition FMEAs using RPN thresholds to the AIAG & VDA (2019) Action Priority methodology.
• Corrective Action & Optimization Review: Evaluating proposed engineering changes or control improvements to verify post-mitigation risk reduction (initial vs revised AP).
• Control Plan Alignment: Verifying that High and Medium AP items from PFMEAs are properly addressed in pre-launch and production Control Plans.

Input Requirements

To execute an FMEA review, collect the following domain inputs:

• System / Item / Process Step: Component, assembly, or manufacturing operation under review.
• Failure Mode (FM): The specific physical or functional manner in which the item fails to meet intent.
• Failure Effect (FE): Impact of the failure mode on the immediate system, vehicle, customer, or regulatory compliance.
• Severity ($S$): Integer rating from 1 to 10 on the AIAG-VDA Severity scale.
• Failure Cause (FC): Physical or chemical mechanism, human action, or process condition that results in the failure mode.
• Current Prevention Controls (PC): Measures implemented to prevent the cause or reduce its occurrence.
• Occurrence ($O$): Integer rating from 1 to 10 on the AIAG-VDA Occurrence scale.
• Current Detection Controls (DC): Measures implemented to detect the cause or failure mode before product release or shipment.
• Detection ($D$): Integer rating from 1 to 10 on the AIAG-VDA Detection scale.

Prerequisites

• Active quality-mcp server connection (verified via mcp-health or ping tool).

Step-by-Step Methodology

Follow the AIAG & VDA 7-Step FMEA execution framework:

┌────────────────────────────────────────────────────────────────────────┐
│                   AIAG & VDA 7-STEP FMEA METHODOLOGY                   │
├──────────────────────────────────┬─────────────────────────────────────┤
│ 1. Planning & Preparation        │ Define scope, boundaries, 5T basis  │
│ 2. Structure Analysis            │ Map system / process hierarchy      │
│ 3. Function Analysis             │ Define functions, requirements, reqs│
│ 4. Failure Analysis              │ Establish FE → FM → FC causal chains│
│ 5. Risk Analysis                 │ Rate S, O, D; call lookup_fmea_ap   │
│ 6. Optimization                  │ Define actions per AP (High/Med/Low)│
│ 7. Results Documentation         │ Document risks, actions, closures   │
└──────────────────────────────────┴─────────────────────────────────────┘

1. Step 1: Planning and Preparation

• Establish analysis boundary, scope, and team charter.
• Identify whether the analysis is a DFMEA (product function), PFMEA (manufacturing process), or FMEA-MSR (monitoring and system response).
• Establish project timeline, baseline documents (BOM, P-Diagram, Process Flow Diagram), and customer requirements.

2. Step 2: Structure Analysis

• For DFMEA: Decompose system $\to$ subsystem $\to$ component.
• For PFMEA: Decompose process item $\to$ process step $\to$ process work element (4M/6M: Man, Machine, Material, Method, Measurement, Milieu).

3. Step 3: Function Analysis

• Map required functions, measurable performance parameters, and regulatory characteristics for each structural element.
• Associate functions directly to customer expectations and engineering specifications.

4. Step 4: Failure Analysis (The Causal Chain)

• For each function, derive the Failure Chain:
  • Failure Effect (FE): What happens to the end user, downstream station, or vehicle? (Governs Severity $S$).
  • Failure Mode (FM): What is the physical failure or non-conformance?
  • Failure Cause (FC): Why did the failure occur? What is the root mechanism?
• Validate bidirectional causal logic: $$\text{Failure Cause (Why?)} \longrightarrow \text{Failure Mode (How?)} \longrightarrow \text{Failure Effect (Impact)}$$

5. Step 5: Risk Analysis (MCP Engine Execution)

• Identify current Prevention Controls (PC) aimed at eliminating or reducing the likelihood of the Failure Cause.
• Identify current Detection Controls (DC) aimed at identifying the Cause or Mode before release/shipment.
• Assign provisional integer ratings ($1 \le S, O, D \le 10$) using the AIAG-VDA standard rating rubrics:
  • Severity ($S$): 10 = Safety hazard without warning / non-compliance; 1 = No discernible effect.
  • Occurrence ($O$): 10 = Extremely high / failure almost inevitable; 1 = Extremely low / eliminated by design.
  • Detection ($D$): 10 = Almost impossible to detect; 1 = Error-proofed / automated detection and lock-out.
• Invoke lookup_fmea_ap on quality-mcp:
  • Send the integer ratings ${S, O, D}$ to the tool.
  • Receive the deterministic RPN and Action Priority (AP: High, Medium, Low).
  • Never calculate RPN or lookup AP in prompt text.

6. Step 6: Optimization (Prioritized Actions)

Based on the deterministic Action Priority result from lookup_fmea_ap:

• High (H) Action Priority:
  • Mandatory Action: The team must identify appropriate actions to improve prevention and/or detection controls, or document why current controls are acceptable (subject to management sign-off).
  • Primary goal: Eliminate failure cause or reduce Occurrence ($O$) via design/process modification (e.g., poke-yoke, geometric keying, automated parameter interlocks).
• Medium (M) Action Priority:
  • Recommended Action: The team should identify actions to improve prevention or detection controls, or document why current controls are adequate.
  • Focus on cost-effective prevention improvements and upgraded detection capability.
• Low (L) Action Priority:
  • Low Priority: The team could identify actions to improve controls; current controls are generally considered capable and acceptable.
• Post-Action Rescoring:
  • After defining corrective actions, assign revised ratings ($S_{rev}, O_{rev}, D_{rev}$).
  • Call lookup_fmea_ap again to verify that the revised AP level has been successfully downgraded (e.g., High $\to$ Low or Medium $\to$ Low).

7. Step 7: Results Documentation

• Compile the complete FMEA record: structural hierarchy, failure chains, initial S/O/D, initial AP, assigned actions, responsible owners, target completion dates, implementation status, and revised S/O/D with verified final AP.

Tool Invocations

lookup_fmea_ap

• MCP Server: quality-mcp
• Purpose: Deterministic lookup of AIAG & VDA (1st Edition, 2019) Action Priority (AP) level and calculation of classic Risk Priority Number (RPN) per Section 3.5.9.
• Parameters:
  • severity (integer, required): Severity rating on the 1–10 AIAG-VDA scale ($1 \le \text{severity} \le 10$).
  • occurrence (integer, required): Occurrence rating on the 1–10 AIAG-VDA scale ($1 \le \text{occurrence} \le 10$).
  • detection (integer, required): Detection rating on the 1–10 AIAG-VDA scale ($1 \le \text{detection} \le 10$).
• Return Type: dict[str, Any]
• Return Schema:
  • severity (integer): Validated input severity score.
  • occurrence (integer): Validated input occurrence score.
  • detection (integer): Validated input detection score.
  • rpn (integer): Calculated Risk Priority Number ($S \times O \times D$, range 1–1000).
  • action_priority (string): Deterministic Action Priority rating ("High", "Medium", or "Low").

Example Tool Invocation

{
  "severity": 10,
  "occurrence": 4,
  "detection": 4
}

Example Successful Response

{
  "severity": 10,
  "occurrence": 4,
  "detection": 4,
  "rpn": 160,
  "action_priority": "High"
}

Protocol Error Handling

If any rating is outside the 1–10 integer range or is of an invalid type (e.g., string, float, boolean, None), quality-mcp returns a standard MCP protocol error response:

• isError: True
• structuredContent: None
• content: [TextContent(type="text", text="... out of range: must be between 1 and 10 ...")]

When encountering a protocol error, the agent must not fabricate a score; instead, inform the user of the invalid parameter and request a corrected rating within the valid 1–10 scale.

Best Practices

• Strict Invariant: Zero Inline Math in Prompt Context. Never perform manual arithmetic ($S \times O \times D$), estimate Action Priority through prompt heuristics, or simulate lookup matrices in agent context. Always route scoring to lookup_fmea_ap.
• Standards Fidelity (AIAG & VDA 2019): Strictly apply AIAG & VDA 1st Edition (2019) Section 3.5.9 principles. Do not apply obsolete RPN threshold rules (e.g., "only act if RPN > 100").
• Severity Dominance: Respect the non-linear structure of Action Priority where Severity is emphasized first ($S \to O \to D$). Note that $S=10, O=4, D=4$ yields $\text{AP} = \text{High}$ despite a modest RPN of 160, whereas $S=4, O=10, D=10$ yields $\text{AP} = \text{High}$ but with different action imperatives.
• Prevention Over Detection: When recommending corrective actions for High or Medium AP, prioritize prevention controls that reduce Occurrence ($O$) over detection controls that merely improve Detection ($D$).
• Distinguish PC vs DC: Ensure that inspection, testing, and audits are classified as Detection Controls (DC), while design margins, error-proofing (poka-yoke), and process capability enhancements are classified as Prevention Controls (PC).
• Closed-Loop Verification: Always require pre- and post-mitigation scoring. Verify that post-action $S_{rev}, O_{rev}, D_{rev}$ values are sent through lookup_fmea_ap to prove risk reduction before closing an action item.