Simulation Governance — An Organizational CAE Quality Management Framework

Exactly. In a nutshell, it's the transformation from CAE dependent on individual craftsmanship to CAE with organizationally managed quality. NAFEMS published a document titled "Simulation Governance & Quality Management" in 2019, systematizing its framework.

There are three main pillars.

• Who does it — Classifying analysts' capabilities (competency) into levels and defining the types of analysis permitted at each level.
• How to check — Setting up review gates at key points in the analysis process to verify quality step by step.
• How to record — Centrally managing models, input data, results, and review records to ensure traceability.

For example, in an automotive company, it means defining as a system, not as a personal custom, that only experienced engineers can be in charge of crash safety analysis.

Exactly. Personal judgment ceases to function the moment that person transfers or leaves the company. The essence of governance is creating a "system that maintains quality even when people change."

NAFEMS's Simulation Governance framework has the following 5-layer structure.

The higher layers focus more on "why to do it," and the lower layers focus more on "how to do it." Governance only functions when all five layers are in place.

Good point. In fact, NAFEMS's framework is designed to align with ISO 9001 and AS9100 (aerospace quality). It's like adding "simulation-specific requirements" to an existing quality management system.

Exactly, NAFEMS recommends a risk-based approach. Analysis risk is classified along two axes.

• Consequence (Impact of Results): How much would safety, cost, or schedule be affected if the analysis results were wrong?
• Uncertainty: How much uncertainty exists in the analysis model itself (new materials, complex physics, lack of experimental data, etc.)

For example, "strength confirmation for a minor shape change of an existing part" is classified as low risk, while "failure analysis of a new composite aircraft wing" is classified as highest risk. The strictness of review, required analyst level, and number of verification items are varied according to the risk level.

Exactly. In practice, an efficient pattern is "80% of analyses are handled with a simplified process for low-medium risk, and resources are concentrated on the remaining 20% of high-risk analyses."

Yes. NAFEMS defines simulation maturity in 5 stages. This applies the concept of CMMI (Capability Maturity Model Integration).

That's true in reality. According to one survey, about 60% of global manufacturers remain at Level 1~2. However, many companies in aerospace (Airbus, Boeing, etc.) and automotive OEMs (BMW, Toyota, etc.) have achieved Level 3 or higher. The important thing is "not aiming for Level 5 in one leap." The transition from Level 2 to Level 3 has the greatest impact.

It's more of a tool to visualize "what level of analysis can be entrusted to this person" rather than ranking. Based on NAFEMS's Professional Simulation Engineer (PSE) system, many companies define 4 competency levels.

Exactly. For high-risk analysis like crash safety, a typical combination is that a Level 3 or higher analyst is in charge, and a Level 4 expert reviews it. Conversely, for a simple strength check of an existing part, it's sufficient for a Level 1 person to use a template and a Level 2 person to check it.

Good question. A competency matrix must always be defined together with a career path. Specifically:

• Training required for promotion to each level (in-house training, NAFEMS seminars, e-learning, etc.)
• Required type and number of practical experiences (e.g., Level 3 requires at least 5 independent analysis experiences across different analysis types).
• Evaluation method (technical interview by higher-level personnel, portfolio review, etc.)
• Recertification cycle (typically every 2-3 years)

Human resource development and governance are two sides of the same coin.

Good image. In typical CAE governance, 4 gates are set up in the analysis process.

That's exactly the point. Without Gate 0, cases where analysts start on their own and then later realize "this analysis was actually insufficient" never end. Spending just 10 minutes on Gate 0 at the beginning dramatically reduces rework effort. In practice, this is said to be the gate with the highest ROI.

Let me introduce a typical Gate 1 checklist.

• Geometry Simplification: List of omitted features (fillets, holes, chamfers, etc.) and reason for omission.
• Mesh Quality: Aspect Ratio < 5, Skewness < 0.8, Jacobian > 0.3 satisfied for all elements.
• Material Model: Source of material data used (material test report number or handbook name), consideration of temperature/speed dependency.
• Boundary Conditions: Whether constraints are not over-constrained (for static analysis, whether reaction forces are physically reasonable).
• Load Conditions: Comprehensiveness of load cases, basis for safety factor.
• Contact Definition: Selection of contact surfaces, basis for friction coefficient, check of initial gap.

Standardizing this as a template in Excel or an SPDM tool makes reviews run efficiently.

Rolling it out company-wide from the start is a recipe for failure. The golden rule is to start small with a pilot project. The recommended implementation steps are as follows.

1. Current State Assessment (1-2 months): Take inventory of the current analysis process, self-assess maturity level. Uncover judgment criteria that have become "tacit knowledge."
2. Pilot Selection (2 weeks): Select one product line or analysis team as the pilot. A mid-sized team (5-10 people) where success is easier to achieve is ideal.
3. Governance Document Preparation (2-3 months): Create drafts of policy, process flow, checklists, and competency matrix.
4. Pilot Operation (3-6 months): Trial governance on a real project. Collect field feedback and revise documents.
5. Company-wide Rollout (6 months~): Report pilot results and KPI data to management and expand to other departments gradually.

An effective method is "past trouble analysis." Gather analysis mistakes, rework, and defect outflow cases from the past 3-5 years and analyze "why they couldn't be prevented." Then, points like "if there had been a checkpoint here, it could have been prevented" or "only veterans knew this judgment" naturally become apparent. Trouble cases are a treasure trove of tacit knowledge.

Half correct and half insufficient. SPDM stands for Simulation Process and Data Management. The key point is that it manages not only data but also processes. Specifically:

• Data Management: Version control and access control for CAD models, FE meshes, input files, result files, and reports.