CACHE
Computer Aids for Chemical Engineering
You are here
Machine Learning and Artificial Intelligence
AI and Machine Learning for Chemical Engineering Education
Use this page to plan a course, add an assignment, learn a skill, or find engineering data. Most links include courses, notebooks, assignments, engineering examples, or documented data.
Find what you need
- I'm teaching
- I'm learning
- I need Data-Driven Engineering resources
- I need GenAI, LLM, or Agentic Engineering resources
- I need a case study or dataset
- I need assessment and responsible AI guidance
Resource key
- Best use shows where a resource fits.
- Audience identifies instructors, learners, or both.
- Level shows the background knowledge needed.
- Time gives an estimated workload.
- Scale describes course scope, model size, or setup effort.
- Runs where distinguishes browser, desktop, local server, and hosted services.
- Data location shows whether code and prompts stay on a device or go to a service.
- Access covers cost, login, and reuse terms.
- Watch for flags limits before assigning a resource.
For instructors
Choose what fits your course.
- One class or assignment. Start with a notebook or one of the assignment ideas below.
- A short module. Teach one method, one chemical engineering example, model checking, and engineering interpretation.
- A full course. Adapt a current course model. State how students may use AI and follow campus policy.
Course models and teaching materials
| Resource | Audience and level | What is included | Best use | Access and limits |
|---|---|---|---|---|
| Data Science and Machine Learning in Chemical Engineering Carnegie Mellon, Spring 2026 |
Instructor and learner Intermediate |
Syllabus, lectures, class exercises, 13 homework sets, and a course project | Full course or a multiweek module | Open |
| Machine Learning for Engineers APMonitor and BYU |
Instructor and learner Introductory to intermediate |
Python notebooks, projects, data engineering, case studies, and Agentic Engineering lessons | Engineering-first ML course | Open. Code is MIT licensed. |
| Machine Learning in Chemical Engineering | Instructor and learner Intermediate |
Jupyter Book, notebooks, and exercises on properties, VLE, CSTR models, control, PCA, and energy data | Ready-to-use chemical engineering assignments | Open. Source code is Apache 2.0. |
| Advanced Mathematics and Computation for Chemical Engineers CBE 660, University of Wisconsin-Madison |
Instructor and learner Advanced |
Python and Jupyter tutorials with Colab links on neural networks, graph neural networks, topological data analysis, and t-SNE for solvent data | Advanced mathematics, data analysis, and ML modules | Public repository. No license file is posted. |
| Statistics for Chemical Engineers Book companion materials, 2025 |
Instructor and learner Intermediate |
Slides, MATLAB and Python scripts, worked examples, and data files for exercises | Statistics, uncertainty, estimation, and data-to-decision assignments | Companion repository is public. The book is sold separately. No repository license is posted. MATLAB and Python results can differ because of routines and randomization. |
| AI for Chemical Engineers University at Buffalo |
Instructor and learner Intermediate to advanced |
Slides and Jupyter tutorials on fault detection, materials, optimization, generative models, and an LLM agent | Upper-level or graduate modules | Materials are CC BY 4.0. Code is Apache 2.0. |
| Artificial Intelligence in Chemical Engineering Columbia University |
Instructor Graduate |
Course outline, prerequisites, topics, and grading plan | Hybrid first-principles and data-driven course design | Open course page |
| ThinkAI Chemical Engineering curriculum map Purdue University |
Instructor and curriculum committee Undergraduate |
Four levels of AI integration across core courses, laboratories, safety, controls, and capstone design | Department-wide curriculum planning | Open |
Ideas for core chemical engineering courses
| Course area | Student work | Suggested resource |
|---|---|---|
| Material and energy balances | Clean sensor data, check units, and compare two regression models | Energy Efficiency data |
| Thermodynamics | Predict a property, then check units, uncertainty, and extrapolation | NIST ThermoML |
| Transport and reaction engineering | Compare a data-only model with a physics-informed model | UW Physics-Informed ML or IDAES examples |
| Process dynamics and control | Build a soft sensor or fault detector. Split data by run or time. | Tennessee Eastman or PRONTO |
| Process design and manufacturing | Use prediction to support an operating decision | Dow soft-sensor assignment or IDAES |
| Materials and molecular engineering | Predict properties or compare molecular representations | Deep Learning for Molecules and Materials, ORD, or Matbench |
| Safety and professional practice | Write a risk register and an independent checking plan | Responsible and reliable AI |
Ready-to-adapt assignments
| Assignment | Time | Source | What students submit |
|---|---|---|---|
| Property prediction | One or two classes Introductory |
Energy Efficiency or a small ThermoML sample | A notebook, parity plot, residual plot, error measure, and statement of where the model should not be used |
| Chemical-process soft sensor | Two to three weeks Intermediate |
Dow Chemical process assignment | A baseline, validation plan, interpretable model, and operator recommendation |
| Fault diagnosis | Two to four weeks Intermediate |
SECOM, Tennessee Eastman, or PRONTO | Results for false alarms, missed faults, detection delay, and class imbalance |
| Hybrid model comparison | Two weeks Intermediate |
CSTR or VLE modules or IDAES | Accuracy, sample needs, conservation checks, and extrapolation results for two models |
| LLM checking lab | One week Intermediate |
Teaching with GenAI | The saved exchange plus checks of units, balances, equations, sources, code, and results |
| Agentic Engineering capstone | Three to six weeks Advanced |
Agentic Engineering resources | A working workflow, action log, test cases, failure review, cost record, plots, and technical report |
For learners
| Step | Start here | What to do | Time and result |
|---|---|---|---|
| 1. Learn ML basics | Google Machine Learning Crash Course scikit-learn MOOC |
Study regression, classification, data splitting, model selection, and evaluation | 15 to 30 hours Complete the exercises |
| 2. Apply ML to engineering | Machine Learning for Engineers Machine Learning in Chemical Engineering |
Reproduce one example. Change the data, features, or model and explain the result. | 20 to 40 hours One checked notebook |
| 3. Add physics and dynamics | Data-Driven Engineering | Compare a data-only model with physical knowledge, constraints, or a first-principles model | One course module Model comparison |
| 4. Learn how LLMs and agents work | GenAI, LLMs, and Agentic Engineering | Start with prompting and checking. Add retrieval or tools only when the task needs them. | 15 to 40 hours A tested workflow |
| 5. Complete a project | Assignment ideas Engineering data |
Choose one engineering question and one dataset | Two to six weeks A notebook that runs, engineering interpretation, known limits, and an AI-use record |
Data-Driven Engineering
| Resource | Level | What is included | Best use | Access and limits |
|---|---|---|---|---|
| Data-Driven Engineering APMonitor and BYU |
Introductory to intermediate | Sensor data, cleaning, statistics, visualization, time series, uncertainty, edge computing, and a group project | Early undergraduate course or lab | Open. Code is MIT licensed. |
| Data-Driven Science and Engineering Brunton and Kutz |
Advanced undergraduate or graduate | Videos, exercises, Python and MATLAB code, dynamics, control, and reduced-order models | Dynamics, controls, and model reduction | Companion materials are open. The textbook is sold separately. |
| Physics-Informed Machine Learning University of Washington |
Upper undergraduate or graduate | Slides and code for heat transfer, reaction discovery, additive manufacturing, and adhesive bonding | Physics-informed lesson or project | Open to view. No reuse license is posted. |
| Physics-Based Deep Learning | Advanced | Digital book and notebooks on physical constraints, differentiable simulation, surrogates, generative methods, and reinforcement learning | Scientific ML reading and labs | Open. Some work benefits from a GPU. |
| IDAES surrogate-modeling examples US Department of Energy and NETL |
Advanced | PySMO, Keras and OMLT, autothermal reforming, supercritical CO2, and flowsheet optimization | Process systems and surrogate optimization | Open source. Some solvers and ALAMO have separate terms. |
| PySINDy examples | Advanced | Notebooks for finding governing equations from dynamic data | Dynamics, control, and equation discovery | Open source |
| Deep Learning for Molecules and Materials | Advanced undergraduate or graduate | Interactive text and Colab notebooks on representations, graph neural networks, equivariance, and generative models | Molecular and materials courses | CC BY-NC 3.0 |
Generative AI and LLMs
LLMs can sound confident and still be wrong. Students should check sources, units, physical constraints, code, and calculations.
Teaching with GenAI
| Resource | Best use | Time | Access and notes |
|---|---|---|---|
| Incorporating Generative AI in the Chemical Engineering Classroom Rebecca K. Lindsey for CACHE |
Assignment rules, student reflection, checking work, and chemical engineering examples | 30 to 60 minutes | Open |
| Using LLMs in Scientific Research Carnegie Mellon |
Research use, literature work, data analysis, coding, and repeatable work | One class or short module | Open |
| Guidance for Using Generative AI in Education Cornell Engineering |
Course rules, privacy, accessibility, and academic integrity | 30 to 60 minutes | Open. Adapt it to local policy. |
Learn how LLMs work
| Resource | Level | What it covers | Limits |
|---|---|---|---|
| Hugging Face LLM Course | Intermediate | Transformers, tokenizers, data, fine-tuning, demos, and reasoning models | An account and GPU help with later exercises |
| Full Stack LLM Bootcamp 2023 archive |
Intermediate | Foundations, prompting, LLM operations, user experience, and agents | Software examples are dated. Check current versions. |
| Stanford CS336, Language Modeling from Scratch Spring 2026 |
Graduate | Tokenization, model and optimizer code, systems, scaling, data, evaluation, and alignment | Strong ML, PyTorch, and systems skills are needed. Some assignments need a powerful GPU or cloud computing. |
Agentic Engineering
An agentic workflow lets a model choose and run steps such as reading data, running code, checking a result, or drafting a report. Begin with a workflow that students can inspect. Give the model more freedom only after tests show that it improves the result.
| Resource | Level | What is included | Best use | Limits |
|---|---|---|---|---|
| APMonitor Agentic Engineering Workplan, coding, coordination, visualization, and technical communication |
Intermediate | Five lessons, prompt builders, and a project workflow | Engineering capstone or team project | Students need basic Python and LLM experience |
| Hugging Face Agents Course | Intermediate | Function calling, agentic retrieval, frameworks, observability, evaluation, and a benchmark project | Hands-on agent course | An account helps with hosted exercises and certification |
| Building Effective Agents | Intermediate | Chaining, routing, parallel work, orchestration, and evaluator loops | Short architecture reading | Vendor-authored. Pair it with independent testing. |
| Berkeley RDI Agentic AI MOOC Fall 2025 archive |
Graduate | Planning, infrastructure, evaluation, multi-agent systems, science, deployment, safety, and security | Graduate readings or instructor preparation | The live course has ended |
| Inspect AI UK AI Security Institute |
Intermediate to advanced | Datasets, solvers, scoring, tool use, multi-turn tasks, sandboxes, and limits | Repeatable agent and LLM tests | Model calls may cost money. Docker is needed for sandboxed tasks. |
Engineering case studies and datasets
Pick a dataset that fits the engineering question and learning goal. Check its license, citation request, privacy rules, and file size before assigning it. The UCI Machine Learning Repository has more options.
| Dataset | Engineering use and task | Level | Scale | Access and limits |
|---|---|---|---|---|
| Dow Chemical process assignment | Soft-sensor regression, feature selection, and process interpretation | Intermediate | Course-sized | Open course page with a complete assignment |
| PRONTO heterogeneous benchmark | Process monitoring, fault detection, and mixed data types | Advanced | About 1.7 GB | Open download. Large files and several data types need more setup. |
| Tennessee Eastman process simulation | PCA, classification, anomaly detection, and time-series checking | Intermediate to advanced | 500 runs, 52 variables, 20 fault types | Open repository record. The data are simulated. Split by run and time. |
| UCI SECOM | Manufacturing classification, missing data, feature selection, and class imbalance | Intermediate | 1,567 rows and 591 sensor features | CC BY 4.0. Many missing values and few failure cases. |
| NASA C-MAPSS | Remaining useful life, predictive maintenance, and domain shift | Advanced | Several simulated run sets | Public download. The catalog does not state a dataset license. |
| UCI Energy Efficiency | Regression, model comparison, and residual plots | Introductory | 768 rows, 8 inputs, 2 targets | CC BY 4.0. Small and simple. It is not a process dataset. |
| NIST ThermoML Archive | Thermophysical properties, XML data work, units, uncertainty, and source tracking | Intermediate to advanced | Large XML archive | Public archive. Availability and reuse terms depend on the publisher. |
| Open Reaction Database | Reaction prediction, experiment planning, data schemas, and source tracking | Intermediate to advanced | More than 2 million reactions | Data are CC BY-SA 4.0. Software is Apache 2.0. Data preparation takes more work than a CSV file. |
| Matbench | Materials regression, classification, and fair model comparison | Intermediate to advanced | 13 tasks with about 300 to 132,000 samples | Open benchmark. Check the terms of each source dataset. |
| ChemBench | LLM testing, error review, and tool versus no-tool comparison | Intermediate to advanced | More than 2,700 questions | Open documentation and dataset. Use it for evaluation, not training. |
Responsible and reliable AI
- Data. Record the source, license, units, missing values, sampling, changes, and measurement uncertainty.
- Model checking. Match the split to the real use. Keep related times, equipment, and batches out of both training and test data.
- Engineering checks. Test balances, units, bounds, limiting cases, and behavior outside the training range.
- Reliability. Report uncertainty, failure cases, drift, and when the model should stop or ask for help.
- GenAI. Check calculations, code, quotations, sources, and claims. A confident answer is not proof.
- Agents. Limit tools and permissions. Protect private data. Log actions. Set cost and time limits. Require a person to approve any action that affects safety, grades, data, or equipment.
- People and policy. Address safety, bias, accessibility, privacy, intellectual property, security, environmental cost, and campus rules.
Guides for instructors
| Resource | Best use | Notes |
|---|---|---|
| NIST AI Risk Management Framework, Generative AI Profile | Risk register, test plan, source record, and incident exercise | Policy framework, not a hands-on course |
| UNESCO Guidance for Generative AI in Education and Research | Department or campus policy discussion | Broad guidance that needs local examples |
| MIT OpenCourseWare, Ethics for Engineers and Artificial Intelligence | Ethics or professional practice module | Course materials are from Spring 2020 |
| Inspect AI | Fixed test sets and clear scoring for LLM or agent labs | Review model-based scores instead of accepting them automatically |
Student AI-use record
Ask students to report the tool and model, date used, task, prompts or a short interaction log, output used, changes made, sources checked, independent tests, and known limits. Students remain responsible for the engineering work they submit.
More tools and examples
| Resource | Best use | Access and notes |
|---|---|---|
| Data Analysis in Chemical Engineering Georgia Tech |
Notebooks for regression, classification, data management, exploratory analysis, and feature engineering | Open repository |
| Streamlit | Turn a Python model or analysis into a small interactive app | Open source |
| NIST, Artificial Intelligence for Chemical Manufacturing | Industry context for measurement, standards, process monitoring, and control | Open |
| Dow's Citizen Data Science Program 2025 |
Industry case study on data literacy and company-wide training | Open article |
Suggest a resource or correction
Use this form to suggest a resource or report a broken link.
When suggesting a resource, include its audience, prerequisites, topic, expected time, access needs, license, and teaching materials.
Send feedback to Dr. John Hedengren at Brigham Young University, Dr. Martha Grover at Georgia Institute of Technology, or Dr. Victor Zavala at the University of Wisconsin-Madison.
Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer