CRAFT · 经验技艺
Make, observe, refine
Knowledge lived in people, materials and physical prototypes.
Industrial AI · Agent-native simulation · Generative engineering
Engineering becomesa living loop
At a pivotal junction in engineering history, MZ Lab is building the industrial AI, agent-native simulation and generative design systems that will transform how products are imagined, verified and realized.
30+years shaping industrial CAE
2025U.S. National Academy of Engineering
01→∞one pilot, full lifecycle ambition
A PIVOTAL JUNCTION
Engineering is entering its next foundational era
The change now underway may prove as consequential as the arrival of calculus or the computer. Engineering has advanced through successive ways of knowing and creating: craft experience, mathematical abstraction, computational simulation—and now intelligence that can reason, generate and act with us.
MATHEMATICS · 数学建模
Describe and predict
Calculus and analytical models made engineering laws explicit.
SIMULATION · 数值仿真
Compute before building
Digital experiments let us test complex systems before committing matter.
INTELLIGENCE · 智能共创
Generate, reason, act
Physics-grounded AI and agents become active collaborators across the lifecycle.
From verifying human ideas to co-creating designs from first principles
OUR MISSION
Lead the transformation at the wave-front
We are building the scientific foundations, agent-native simulation software, industrial AI systems and generative design methods that move engineering beyond ‘draw first, test later.’ The engineer becomes an active creator—setting intent, encoding values and constraints, and working with algorithms to discover solutions that neither could reach alone.
We do not see AI as an assistant added to engineering. We see it as a new engineering medium.
THE ENGINEERING LOOP
A system that learns across the product lifecycle
The laboratory is organized around one continuous exchange: engineering knowledge defines intent; agents assemble and operate trusted tools; simulation exposes physical truth; generative design creates alternatives; manufacturing returns evidence. Each cycle improves the next.
Understand
Requirements, standards, product knowledge and engineering intent become structured context.
Output: a trusted problem definitionSimulate
Agent-native workflows assemble models, solvers, data and verification into reproducible evidence.
Output: physical evidenceGenerate
Optimization explores topology, geometry, materials, systems and uncertainty—not just one shape.
Output: design candidatesRealize & learn
DFM, costing, tooling, production and quality close the loop with measured outcomes.
Output: delivery and new evidenceTHREE PRIORITIES · ONE SYSTEM
Our research, in deliberate order
The priorities are not separate departments. Industrial AI leads; agent-native simulation provides the dependable computational substrate; generative design turns both into new products.
01
LEAD FOCUS
Industrial AI
We build AI systems that reason with engineering knowledge, operate validated software, preserve evidence, and support accountable decisions across product development.
02
COMPUTATIONAL FOUNDATION
Next-generation simulation software—native to agentic workflows
Instead of wrapping yesterday’s interfaces with a chatbot, we rethink simulation architecture for tool discovery, structured intent, composable solvers, automatic verification and transparent provenance.
03
CREATIVE ENGINE
Simulation-driven generative design
We connect concept generation to verified, manufacturable finalization through topology, shape, sizing and multidisciplinary optimization—under real physics, uncertainty and process constraints.
PROGRESS & ROADMAP
Build the foundation. Prove one engineering loop. Expand with discipline
Injection-mold engineering is our first industrial proving ground. The platform is now in place; current work focuses on controlled mold-flow execution and result interpretation, followed by drawing generation, 3D mold design and DFM—then extension upstream into product design and development.
FOUNDATION ONLINE
Manufacturing AI Platform
A shared agentic foundation coordinates engineering context, specialist tools, workflow state, evidence and human review.
- Shared contextGeometry, process assumptions and project evidence
- Tool orchestrationControlled calls into engineering software
- Reviewable evidenceTraceable outputs and explicit human gates
ENGINEERING AGENTS
From the first working loop to a broader design system
Progress is staged by engineering evidence—not by feature count.
01IN DEVELOPMENT
Mold-flow analysis agent
The first end-to-end engineering loop turns a simulation tool into a controlled, inspectable workflow.
- Moldflow workflow invocation
- Simulation execution and state tracking
- Result extraction, interpretation and review
NEXT
2D drawing agent
Engineering drawing generation and standards-based processing.
NEXT
3D mold-design agent
Design capability developed progressively around representative mold structures.
NEXT
DFM agent
Product-geometry analysis and systematic manufacturability checks.
Current state · Foundation complete; mold-flow workflow under active development; design and DFM capabilities staged next.
INDUSTRIAL AI ROADMAP
From mold agents to manufacturing intelligence
Replicate proven methods, compound real engineering experience and train an industrial LLM grounded in manufacturing—not a generic model looking in from the outside.
STARTING POINT
Begin with mold engineering
Build repeatable agent workflows around a tightly connected engineering domain, then validate them against real projects and expert review.
- DFMManufacturability checks
- Mold design3D engineering workflows
- Mold-flow analysisSimulation and interpretation
- 2D drawingsStandards-based delivery
REPLICATE & EXPAND
Extend across manufacturing
Transfer the working pattern into adjacent processes while preserving each domain's physics, tools and decision logic.
STAMPING
Process to formed part
- Process design
- Forming simulation
- Springback & defect diagnosis
CASTING
Flow to sound casting
- Gating-system design
- Casting simulation
- Defect diagnosis
Then onward to more manufacturing domains.
COMPOUND INTELLIGENCE
Grow an Industrial LLM through engineering work
Every workflow becomes a learning surface: engineering sources enter with provenance, expert actions become reusable trajectories, and validated outcomes strengthen the next decision.
INDUSTRIALLLMManufacturing-grounded
ENGINEERING SOURCES
- Databases & tables
- Process documents
- CAD
- CAE & simulation
- Equipment & IoT
COMPOUNDED ASSETS
- Engineering knowledge
- Industrial data
- Success & failure cases
LEARNED TRAJECTORIES
- Decision trajectories
- Tool-calling trajectories
Understands industrial knowledge
Understands engineering workflows
Knows how to use engineering tools
REAL ENGINEERING WORK
Knowledge + data + decision experience
STRONGER INDUSTRIAL LLM
A continuously improving platform
AGENT-NATIVE BY DESIGN
A new simulation stack for engineering autonomy
Trust comes from architecture: agents must know what tools can do, how evidence was created, when a result is invalid, and where a human must decide.
DECISION LAYER
Industrial agents
Role-specific reasoning, shared project memory, human gates and enterprise coordination.
ORCHESTRATION LAYER
Engineering workflow fabric
Structured intent, tool routing, workflow state, model/data lineage and automatic verification.
PHYSICS LAYER
Open, scalable simulation & optimization
Modern C++ parallel architecture, solver interoperability, Python/Julia access, sensitivities and optimization.
ENGINEERING INTENT
Assess fill balance and warpage risk for the latest geometry. Preserve every assumption and simulation artifact for review.
THE INTERFACE CHANGES
Conversation is only the surface. Evidence is the product
An engineer should be able to ask for an outcome in domain language, inspect the assumptions, open the generated model, challenge the result and reproduce the complete path. That requires deeper software change than a chat overlay.
CONCEPT → MANUFACTURABLE FINALIZATION
Generative design is a process, not a picture
The research challenge is to carry a promising concept through all the difficult transitions: topology to geometry, idealized physics to real loads, continuous material to discrete parts, performance to robustness, and analysis to manufacturing.
01Explore
Topology, architecture, material and system alternatives.
02Verify
Multiphysics, uncertainty, failure, robustness and sensitivity.
03Realize
Geometry reconstruction, process constraints, tooling and production.
RENAULT TRUCKS × ALTAIR
SIMULATION-DRIVEN DESIGN SHOWCASE
“Additive manufacturing frees us from constraints, unleashing and inspiring engineers’ creativity.”
ENGINEERING EVIDENCE
Ideas that left the page
The laboratory’s direction is grounded in a long record of methods translated into aircraft structures, production software and experimentally tested manufacturing workflows.
AIRCRAFT STRUCTURES · AIRBUS A380
Topology optimization at aircraft scale
A widely cited Altair–Airbus application used topology optimization to redesign a family of A380 leading-edge ribs. The historical case reports a 500 kg aircraft weight saving and shows what simulation-driven generative design means in practice: one method, repeated across a real structural system, carried into an aerospace program.
500 kgreported aircraft weight saving
Rib familynot a one-off demonstration
Industrymethod → system → program
FROM OPTIMIZATION TO PRODUCTION
Engineering translation is the real test
Aircraft-scale impact requires more than an elegant optimum: model credibility, design reconstruction, certification constraints, software reliability and production integration all have to work together.
BOEING 787 · OPTIMIZATION CENTER
Optimization became part of the aircraft design process
At Boeing’s 787 Optimization Center, optimization was embedded into the aircraft design workflow rather than treated as a late specialist study. Structural analysis, mathematical search and multidisciplinary program constraints were brought together to explore the design space systematically—an early proof of the simulation-driven engineering model the lab is extending today.
Historical case material: Boeing / Altair presentation.A CONTINUOUS LINE OF WORK
From optimization foundations to industrial software—and now to agentic engineering
Foundational topology and shape optimization
The COC algorithm connected topological, geometrical and generalized shape optimization.
Design control and integrated realization
Minimum-member control and the integration of topology, sizing and shape made optimization more practical.
Industrial CAE at global scale
Leadership across OptiStruct, HyperStudy, Radioss and the wider simulation–digital-twin–AI ecosystem.
Industrial AI as an engineering system
A new platform joins domain knowledge, agentic workflows, simulation, optimization and enterprise delivery.
PEOPLE & OPPORTUNITIES
Scientific depth. Software discipline. Industrial judgment
MZ Lab is led by Professor Ming Zhou, University Chair Professor and Dean of the School of Engineering at EIT, and a member of the U.S. National Academy of Engineering.
MZ Lab is an interdisciplinary team advancing simulation-driven generative engineering across mechanics, applied mathematics, numerical computing, manufacturing, software engineering and artificial intelligence. Researchers and engineers work side by side so that new methods can be tested, implemented and translated into useful engineering systems.
Computational mechanics
Optimization algorithms
Industrial software
Manufacturing simulation
Agentic AI
Technology translation
View Ming Zhou’s profile →
OUR TEAM
Researchers, engineers and students building the next generation of engineering together
Faculty & Researchers
Visiting Scholars
Postdoctoral Researchers
CURRENT SIGNALS
Work in motion
Selected activities connecting simulation, AI, engineering education and industrial practice.
INDUSTRIAL PILOT
Agentic AI for injection-mold engineering moves from platform foundation to a working simulation loop
The current pilot controls mold-flow execution, result extraction and engineering review; drawing generation, 3D mold design and DFM follow next.
PUBLIC DIALOGUE
Engineering technology across eras: a conversation with young researchers
Professor Zhou joined an EIT dialogue on curiosity, open thinking and the enduring questions behind research.
Read at EIT ↗SIMULATION TECHNOLOGY
AI empowerment and digital twins for intelligent high-end equipment design
At the Simulation Technology Application Conference, Professor Zhou addressed the convergence of aerospace CAE, high-performance computing and intelligent design.
Read at EIT ↗BRING THE NEXT ENGINEERING QUESTION
Build the loop with us
We welcome doctoral students, postdoctoral researchers, research faculty, software engineers and industrial partners who want to connect rigorous mechanics with AI-native engineering systems.