Axiomatic Design · Collective System Design · AI
Axiomatic Design Foundations for AI-Driven Collective System Design
David S. Cochran, Ph.D. and Joseph Smith — Purdue University Fort Wayne · System Design, LLC
Opening
The Challenge — The Valley of Design Death
CSD Foundations
The Flame Model of a System
The Flame Model — Diagnosis to Design
The CSD 12-Step Process
Axiom 1 and Design Predictability
A Single Source of Truth
Common Terminology — ISO/IEC/IEEE 42010
The Consistent Design Artifact Model
AI: Keeper of the Flame
AI Across the Four Layers
RAG AI Architecture
Case Studies — Healthcare and Mechatronics
AI in Engineering Education
The CSD Requirements Ledger
Inside the Ledger — Feature Tour
The Student as System Architect
Close
Thank You · Discussion
Figure numbers (Fig. 1–7) match the companion ICAD paper; slides present them in teaching order, so they may appear out of sequence.
From Design Intent to Physical Realization
Where Students Get Stuck
Students struggle to cross the “valley of design death” — from design intention to realization — with an over-reliance on CAD and simulation tools without an understanding of the issues required to build things. The result is last-minute trial and error.
Our Thesis
The CSD 12-Step process is a “liberating structure” that preserves design intent by distilling it into a Consistent Design Artifact Model (CDAM):
- Merges Axiomatic Design with Systems Engineering
- Starts from solution-neutral requirements
- Produces a machine-readable artifact for RAG-driven AI
At the heart of every great engineering feat lies Design Intent — captured, after Nam P. Suh, as independent Functional Requirements (FRs) — turning “complex,” unpredictable systems into predictable, improvable engineered ones.
A System as an Integrated Whole
Collective System Design (CSD) represents a system not as isolated tasks but as an integrated whole.
Tone establishes collective agreement about what the system must accomplish.
Thinking requires a consistently-used ontology to express design information and artifacts — which is where Axiomatic Design fits.
Structure is derived from the Axiomatic Design Decomposition and the other viewpoints that express the design. Applies to a product design or an enterprise design.
Work generally implements the leaf-level solutions of the Axiomatic Design Decomposition. The physical realization and sustainability of a system results from establishing standard work to define “normal.”
Tone sets Thinking; Thinking sets Structure; Structure is realized as Work. People are not the cause of system failures — poorly designed systems fail people.
Fig. 1 — The Flame Model
The Flame Model of a System
Key Points
- System Design relies on the people within a system being able to work together — which requires a conscious understanding of the mindset and attitudes used to approach each other in making design decisions.
- The benefit of Axiomatic Design is that it forces a team to define a solution (a design decision) to achieve a function of the design.
- Lasting change requires a conscious choice to change existing behaviors in each of the four layers represented by the Flame Model of a system.
Key Points
- A design artifact is any work product that expresses part of the design — an FR, an FRm, a PS, a risk, a verification, or a diagram — in a defined, reusable form.
- Artifacts produced on the descent and the verification on the ascent are bound to the same Consistent Design Artifact Model (CDAM) populated at the bottom.
- The purpose of the CSD 12-Step is to provide a methodology for design that yields predictable results and mitigates unstructured or biased solution selection.
Key Points
- The choice of the PS (Physical Solution) to achieve an FR (Functional Requirement) is a decision made by the designer(s).
- Diagonal = uncoupled (most predictable); triangular = decoupled, predictable if the sequence is respected; full = coupled, with unpredictable cross-talk.
Key Points
- Architecture — the fundamental concepts and properties of a system in its environment, embodied in its elements, relationships, and the principles of its design and evolution.
- Architecture Description — the work product (the set of artifacts) used to express an architecture.
- Axiomatic Design is one viewpoint used to frame a stakeholder’s concerns; 42010 is a liberating structure that lets designers frame concerns with multiple viewpoints.
Key Points
- The goal is to ensure that Axiomatic Design practitioners use consistent terminology across projects.
- That we embrace multiple viewpoints within the engineering lifecycle.
- With the emergence of the most powerful tool ever conceived, that we practice a conscious tone of well-being with each other.
An AI Agent, Grounded on the CDAM
A student or engineer query goes to an AI agent grounded on the CDAM and the primer. (“RAG-based” just means retrieval-augmented — the agent answers from our own model and corpus, not generic training.)
Because the knowledge is structured, the agent reviews and explains rather than guesses:
- Detects coupling and checks FR / PS mapping
- Flags missing measures and interfaces
- Explains, in the terms of the CSD 12-Step process
The agent reviews and flags; the student decides and edits.
Fig. 7 — The AI Agent over the CDAM
Key Points — What the CDAM Gives Students and Designers
- One shared model means less rework and fewer late surprises.
- The design can be checked against the process instantly, 24/7.
- Design knowledge carries from one project (and team) to the next.
One Logic, Two Very Different Domains
The same decoupling logic translates high-level intent into measurable requirements before any physical solution is chosen — whether the system is human-centric or hardware-centric.
| Healthcare — ER Triage | Mechatronics — HVAC PLC | |
|---|---|---|
| Functional Requirement (FR) | Categorize patient acuity | Transmit digital data reliably |
| FR Measure (FRm) | Time to triage ≤ 5 min | Packet delivery ≥ 98% |
| Physical Solution (PS) | Rapid Triage protocol | Narrowband BFSK Modem |
| AI’s role | Design Auditor and Synthetic Implementation Partner across a hospital network | Technical Assistant cross-checking datasheets vs. IEC 61000-4-6 |
Franciscan Health (Q4 2025): decoupled logic ensured triage speed did not compromise clinician-matching accuracy — design patterns scaled across hospitals without intent drift.
Residential HVAC: RAG retrieved the IEC standard and modem datasheets together — minutes instead of hours of manual cross-referencing.
Same CDAM grammar, two domains — the “Thinking” layer stays rigorous across both.
The CDAM, Made Usable for Students
The CSD Requirements Ledger brings the CDAM to life for senior design and SE education — one place where every FR, FRm, PS, risk, and verification is a single source of truth. The AI agent:
- Guides students through the 12 Steps, viewpoint-by-viewpoint
- Audits coupling as the design is entered (reviews, never authors)
- Keeps one shared model across mechanical, electrical and software
- A logic-driven co-pilot, available 24/7
- States what a good FR looks like
- Derives / suggests FRs from the System Boundary and Use Case diagrams
- Points to the solutions and interfaces to build early
The Ledger at Work — Catching a Solution in Disguise
A student enters an “FR”
FR3:
“Use a 12 V Li-ion battery pack”
⚠
🤖 AI Review — in real time
That’s a Physical Solution, not a Functional Requirement — it names the part before the function. An FR states what the system must do, solution-neutrally. Define the function first; the PS follows.
Corrected in the ledger
FR3:Store electrical energy
PS3:12 V Li-ion battery pack ✓
The CSD Requirements Ledger
Our Conscious Approach to AI in Engineering Education
The SE Center is using AI technologies to:
•
Build engineering tools that help focus students’ time and energy on designing and building working systems
•
Provide real-time feedback on the design model from trained AI agents
•
Inhibit the use of AI to do “the thinking” by providing a process that guides decision-making and implementation, yet requires engineers to do the thinking and learning
1Live Progress & Health
FRs achieved, model inconsistencies and constraint compliance — updated live as engineers design.
2Single Source of Truth
Every FR, FRm, PS, risk and verification in one shared model — the ledger.
3AD Decomposition Builder
The FR→PS Axiomatic Design decomposition, built from student-authored inputs.
4Report Builder
Documentation auto-populated from the model — one click to export a PDF.
5Presentation Builder
A live, editable presentation view — generated from the model.
6Project Management
Teams, timeline, to-dos, open questions and team chat — the project, organized.
7Ask the Agent
A CSD Primer–grounded co-pilot, 24/7 — it reviews the design, never authors it.
From Part-Picking Bias to System Architect
•
A solution-neutral start reduces ambiguity in every later stage
•
A single source of truth (CDAM) + common terminology (ISO 42010) defeat the silo effect
•
Axiom 1 keeps designs predictable and improvable
•
A grounded AI agent (RAG-based) audits logic — it does not replace judgment
CSD relies on a CDAM and structured viewpoints to do design; the Requirements Ledger gives a user-friendly AI agent (RAG-based) to implement CSD on that foundation — shifting students from part-picking to proactive System Architects.
