How to run an AI-assisted HAZOP
Use an AI to prepare, run and review a HAZOP without losing the rigour of the method: where the AI helps, preparing the file for a node, writing the prompts, checking and measuring what it proposes, trying it step by step in Polyodos on reactor R33030. Course taught entirely in English.
- 6 chapters
- 28 lessons
- 2 exercises
- 6 quizzes
Graded quizzes, exercises and interactive tools are available in the online course.
Read the first lesson → Create an account Log in to the course
At a glance
- Level
- Advanced
- Estimated time
- about 9.5 h
- Audience
- HAZOP facilitators and participants, process engineers.
- Recommended first
- Industrial risk and dependability, How to run a technical or industrial risk analysis
- Learning outcomes
- Prepare a node file for an AI.
- Write HAZOP prompts and check the suggestions.
- Measure the quality of an assisted HAZOP against a human reference.
- Ensure traceability and accountability of the analysis.
- Assessment
- A graded quiz for each chapter; in a group, the next chapter opens once the lessons are read and the quiz passed (60% by default, threshold set by the trainer).
- Badge
- Digital badge (Open Badges format) issued by the trainer once every quiz of the course is passed and the final exam passed at 70%: 20 random questions, 30 minutes, at most 3 attempts, 24 h between two attempts.
- Updated
Course outline
Lessons marked “free to read” need no account; the other lessons, the exercises, the quizzes and the answers do.
Chapter 1Where AI can help in a HAZOP, and where it cannot4 of 4 lessons free to read
The steps of a HAZOP and their cost, what AI can and cannot do, three modes of use, the golden rule.
- 1.1The steps of a HAZOP and their cost in timefree to read
- 1.2What AI can do, and what it cannot dofree to read
- 1.3Three modes of use: before, during, after the sessionfree to read
- 1.4The golden rule: AI proposes, the team decides and keeps a recordfree to read
Chapter 2Preparing the data4 lessons
What to provide, what not to provide, splitting into nodes, building the node file.
- 2.1What to provide to the AI
- 2.2Confidentiality: what you paste, what you do not paste
- 2.3Split into nodes before asking
- 2.4Practical sheet: the node file “cooling of R33030”
Chapter 3Asking: prompts for HAZOP5 lessons
Anatomy of a prompt, systematic deviations, causes, consequences and barriers kept separate, a reference example, templates.
- 3.1Anatomy of a HAZOP prompt
- 3.2Generating deviations systematically
- 3.3Causes, consequences, barriers: ask separately
- 3.4Giving an example: a reference HAZOP line
- 3.5Reusable templates to copy
Chapter 4Checking and measuring what the AI proposes5 lessons
Verification grid, typical errors, comparison with a human reference, repetition, line-by-line decision.
- 4.1The verification grid for a HAZOP line
- 4.2Typical AI errors in HAZOP
- 4.3Measuring: comparing with a human reference
- 4.4Repeating the request: variability and voting
- 4.5Deciding what to keep, and attributing it to the team
Chapter 5Step by step in Polyodos6 lessons
Configure the assistant, prepare the node, run the proposed HAZOP, check, compare with the human reference, record the findings.
- 5.1Configuring the assistant
- 5.2Preparing the node and the model
- 5.3Running the HAZOP proposed by the AI
- 5.4Checking line by line with the grid
- 5.5Compare with the human reference
- 5.6Summary of the trial findings
Chapter 6Responsibility, traceability, acceptability4 lessons
Who answers for the study, how to trace the use of AI, what makes a team accept the tool, a debate.
- 6.1Who is responsible
- 6.2Tracing the use of AI
- 6.3Conditions for a team to accept the tool
- 6.4Debate: supply the AI’s proposal before the session?