AI-Copilot–Enhanced Design of CMOS-MEMS Thermal Flow Sensors for Smart Energy-efficient Buildings
Project Description
Buildings consume a major fraction of global energy, and HVAC systems are a primary driver. A key bottleneck for smarter, more energy-efficient HVAC control is high-quality, distributed airflow sensing that is low-cost, low-power, and deployable at scale (e.g., vents, ducts, zone inlets, filter monitoring points). CMOS-MEMS thermal flow sensors (hot-film / hot-wire / calorimetric) are attractive because they can be miniaturized, batch-fabricated, and co-integrated with on-chip readout circuits for scalable deployment.
This UROP focuses on the design of CMOS-MEMS thermal flow sensors suitable for building applications, emphasizing physics-grounded modeling + multiphysics simulation + fabrication-aware design. In parallel, students will learn an “AI-copilot” workflow inspired by Gabriel Petersson’s method (high-school student applied AI to learn advanced AI and work for OpenAI without finishing 4-year UG program): project-first + recursive gap filling + teach-back validation. Students will build a working first-pass model early, identify what breaks, patch knowledge gaps efficiently with AI assistance, and validate understanding by explaining back and testing with quantitative plots. The end goal is a design package (model + optimized geometry + test plan) that is ready for discussion toward prototyping or foundry implementation.
This UROP focuses on the design of CMOS-MEMS thermal flow sensors suitable for building applications, emphasizing physics-grounded modeling + multiphysics simulation + fabrication-aware design. In parallel, students will learn an “AI-copilot” workflow inspired by Gabriel Petersson’s method (high-school student applied AI to learn advanced AI and work for OpenAI without finishing 4-year UG program): project-first + recursive gap filling + teach-back validation. Students will build a working first-pass model early, identify what breaks, patch knowledge gaps efficiently with AI assistance, and validate understanding by explaining back and testing with quantitative plots. The end goal is a design package (model + optimized geometry + test plan) that is ready for discussion toward prototyping or foundry implementation.
Supervisor
LEE Yi-Kuen
Quota
4
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
Successful candidates will work in a small team (or individually) on the end-to-end sensor design workflow, from spec definition to modeling, simulation, and test planning.
Key responsibilities may include:
Problem framing & specs: define building-relevant sensing use case (HVAC duct/vent/zone monitoring) and translate it into measurable performance targets (flow range, sensitivity, power, response time, drift).
Project-first modeling: build a compact thermal-flow model (energy balance / thermal resistance network) and generate quantitative predictions (ΔT and power vs flow, sensitivity curves).
Multiphysics simulation & optimization: run COMSOL/ANSYS simulations (conduction–convection coupling) to optimize geometry and placement of heater/thermistors.
CMOS-MEMS feasibility: incorporate process and layout constraints (materials/layers, passivation, release windows, routing implications) and produce a risk/mitigation list.
AI-copilot engineering workflow: use AI to accelerate derivations, code debugging, simulation planning, and literature scanning—while keeping a strict engineering log and validation plots.
Test & calibration plan: propose an experimental setup and calibration protocol relevant to building deployment, including an uncertainty/error budget.
Documentation & presentation: prepare final report and slides, including weekly teach-back summaries and a final design recommendation.
Key responsibilities may include:
Problem framing & specs: define building-relevant sensing use case (HVAC duct/vent/zone monitoring) and translate it into measurable performance targets (flow range, sensitivity, power, response time, drift).
Project-first modeling: build a compact thermal-flow model (energy balance / thermal resistance network) and generate quantitative predictions (ΔT and power vs flow, sensitivity curves).
Multiphysics simulation & optimization: run COMSOL/ANSYS simulations (conduction–convection coupling) to optimize geometry and placement of heater/thermistors.
CMOS-MEMS feasibility: incorporate process and layout constraints (materials/layers, passivation, release windows, routing implications) and produce a risk/mitigation list.
AI-copilot engineering workflow: use AI to accelerate derivations, code debugging, simulation planning, and literature scanning—while keeping a strict engineering log and validation plots.
Test & calibration plan: propose an experimental setup and calibration protocol relevant to building deployment, including an uncertainty/error budget.
Documentation & presentation: prepare final report and slides, including weekly teach-back summaries and a final design recommendation.
Applicant's Learning Objectives
By completing the project, students will be able to:
Technical competencies
Understand operating principles of thermal flow sensing (CTA/CPA, hot-film/hot-wire, calorimetric concepts).
Apply heat transfer + fluid mechanics to build simplified models and interpret scaling laws.
Use multiphysics simulation to validate and improve sensor designs quantitatively.
Understand key CMOS-MEMS design constraints that affect performance and manufacturability (thermal isolation, substrate losses, packaging effects, drift).
Design a calibration + testing workflow and articulate measurement uncertainty sources.
AI-augmented engineering competencies (Gabriel-style)
Practice project-first learning: start from a working MVP model/design, not from long prerequisite study.
Execute recursive gap filling: identify exactly what is unknown, learn it fast, patch the model, and iterate.
Use teach-back validation to ensure real understanding (explain-to-AI / explain-to-human, then verify with plots).
Build a repeatable workflow for using AI as a technical copilot without shallow copy-paste: every AI suggestion must be checked by units, scaling, or simulation/plot evidence.
Technical competencies
Understand operating principles of thermal flow sensing (CTA/CPA, hot-film/hot-wire, calorimetric concepts).
Apply heat transfer + fluid mechanics to build simplified models and interpret scaling laws.
Use multiphysics simulation to validate and improve sensor designs quantitatively.
Understand key CMOS-MEMS design constraints that affect performance and manufacturability (thermal isolation, substrate losses, packaging effects, drift).
Design a calibration + testing workflow and articulate measurement uncertainty sources.
AI-augmented engineering competencies (Gabriel-style)
Practice project-first learning: start from a working MVP model/design, not from long prerequisite study.
Execute recursive gap filling: identify exactly what is unknown, learn it fast, patch the model, and iterate.
Use teach-back validation to ensure real understanding (explain-to-AI / explain-to-human, then verify with plots).
Build a repeatable workflow for using AI as a technical copilot without shallow copy-paste: every AI suggestion must be checked by units, scaling, or simulation/plot evidence.
Complexity of the project
Challenging