Stage 3: Risk assessment and safe systems · topic 11 of 50

HAZID, HAZOP, FMEA and bow-tie basics

Structured hazard studies for process sites: what-if, HAZOP, FMEA, fault trees and bow-ties. You will be able to pick a method, run a small study and say which laws require one.

  • Optional
  • Intermediate
  • about 12 hours

Part of the EHS officer roadmap: how to become one and what to learn

Written by the GreenHired editorial teamSources last checked 20 November 2026How we research roadmaps

Scope comes before method. Fix the scope, get drawings that match what is built, and put two people in the room: a facilitator who knows the method and an operator who knows the plant. Then pick the lightest method that fits the process, run it, and plan the follow-up before the first meeting, because a list of findings that nobody closes changes nothing.

A what-if study on a solvent day tank

A day tank holds a flammable solvent, filled from a drum by a transfer pump and warmed by a steam coil. This is an illustrative walk-through for learning, not advice for a real site.

A what-if suits a system this small. Work along the process and record each question with its cause, consequence, existing safeguard and action:

  • What if the pump keeps running after the tank is full? Cause: the operator is called away. Consequence: overflow into the bund and solvent vapour in the room. Safeguard today: the bund and a person watching. Action: a high-level alarm, with an owner and a date.
  • What if the steam valve sticks open? Cause: valve fault. Consequence: the solvent overheats and gives off more vapour. Safeguard today: none found. Action: a temperature trip, with an owner and a date.

Read the first row again. The only safeguard is a person, so ask what happens when that person is not there. That second question is where a study earns its time.

When the plant grows, so does the method

If the tank were one of several vessels joined by pipework, a what-if would start to miss things. That is the case HSE’s staged HAZOP is built for: a coarse study on the flowsheet, then a full one on the frozen piping and instrument diagram. HSE also warns that a HAZOP gives no likelihood. Where an action does not remove a hazard, a further analysis should judge whether the remaining risk is acceptable. Which regimes make a study compulsory is covered in process safety and major-hazard regimes.

Checking the barriers on a bow-tie

A bow-tie helps you check the barriers on the same tank. Put “tank overflows” in the middle. Preventive barriers on the left might be the level alarm and a pump trip; the bund is a mitigating barrier on the right. If the alarm and the trip both read the same level sensor, you have one barrier, not two. Mark that on the diagram.

Keeping the actions alive

When the plant is modified later, the study is out of date, which is why management of change sits next to this topic, and the work itself is controlled day to day through permits and isolation. Give every action an owner and a date, and review the list until each one is closed.

What you will learn

  • The named methods regulators accept for process hazard analysis.
  • How a HAZOP study works in stages and what it cannot tell you.
  • What barriers and layers of protection are, and how a bow-tie shows them.
  • What a process hazard analysis must cover, who must do it and how often it is revalidated.
  • Which laws trigger a study: US PSM and RMP, Great Britain COMAH, and Indian hazardous-process rules.

The parts of this topic

What-if, checklists, FMEA and fault treesabout 3 hours

What-if questions and checklists suit simple processes. FMEA looks at how each component can fail, and a fault tree works back from a bad outcome.

Picking a method because it is familiar, not because it fits, is the usual slip. OSHA lists what-if, checklist, what-if/checklist, HAZOP, FMEA and fault tree analysis as accepted methods and says the choice must suit the complexity of the process. A what-if asks questions step by step, such as 'what if the pump stops?'. A checklist covers known hazards of familiar equipment. FMEA lists each component's failure modes and their effects. A fault tree starts from an unwanted event and works back to causes. No free primary guide to FMEA or fault trees is listed here.

HAZOPabout 3 hours

A hazard and operability study uses a team and guide words to examine each part of a design for deviations. It finds hazards but does not estimate how likely they are.

HSE describes HAZOP as done in stages: identify major hazards and check key data, run a coarse study on the flowsheet, run a full study on the frozen piping and instrument diagram, check that the actions are implemented, and audit after a few months of operation. Written procedures should set the level of study needed. HSE adds that a HAZOP finds hazards but gives no likelihood, so further analysis should judge whether the remaining risk is acceptable. The workshop is the start of the actions, not the end of the work.

Barriers, bow-tie and LOPAabout 3 hours

Barrier thinking asks which independent layers stand between a hazard and harm. A bow-tie draws them, and LOPA checks whether they are enough.

Layers can be design, alarms, trips and interlocks, relief, procedures and emergency response. Two barriers that rely on the same sensor or person count as one. A bow-tie puts the hazard in the middle, with causes and preventive barriers on one side and consequences and mitigating barriers on the other. LOPA, layer of protection analysis, is a semi-quantitative check on whether the layers are enough. HSE's control systems document covers alarms, trips, interlocks and integrity levels for safety-related systems. No free primary guide to bow-tie or LOPA is listed here, so treat this as an overview.

Process hazard analysis in practiceabout 3 hours

Under US PSM a process hazard analysis uses accepted methods, covers set topics, is done by a qualified team, is revalidated every five years and drives a documented action list.

A good hazard analysis workshop with open actions has not yet made the plant safer. 29 CFR 1910.119 says the analysis uses one or more listed methods and addresses the process hazards, earlier incidents with catastrophic potential, engineering and administrative controls, what happens when controls fail, facility siting, human factors and a qualitative evaluation of health and safety effects. A team with engineering and process-operating expertise does it, including one person who knows the process and one who knows the method. It is revalidated at least every five years and kept for the life of the process.

When a study is requiredabout 3 hours

Laws trigger studies by chemical, quantity or type of site, so check the thresholds for your site and region.

OSHA's PSM guidance says the standard applies to processes with listed toxic and reactive chemicals at threshold quantities and to flammable liquids or gases of 10,000 pounds or more at one location, with some exceptions. EPA's RMP rule requires facilities above a threshold quantity of a regulated substance to run a program that includes hazard assessment. Great Britain's COMAH Regulations 2015 cover major-hazard sites. India's OSH Code lists industries with hazardous processes in its First Schedule. Thresholds as read on 20 November 2026: check the current text, and do not assume a small site is exempt.

Practice task

Take a simple system, such as filling and heating a day tank. Run a small what-if study with at least ten questions, record each cause, consequence, existing safeguard and recommended action in a table, assign owners and dates, and add a one-paragraph note saying which method you would choose for a larger plant and why.

Skills you build

Process hazard analysisHAZOP basicsBarrier thinkingAction tracking

Resources for this topic

Every source here is free. Each link opens the provider's own site in a new tab.

Back to the whole roadmap

Live jobs

Live Environmental, health and safety (EHS) officer jobs

Roles whose title matches this career path.

All Environmental Health & Safety jobs

Spot a problem with this topic?

Tell us what is wrong, missing or unclear here. Your note is attached to this topic.

Notes are sent from this page with JavaScript on. Without it, write to us on the contact page.