In petroleum operations, HSE management is largely about keeping barriers in place between hazards and harm. This guide covers the tools used to identify hazards (HAZID, HAZOP), to describe and manage barriers (bow-ties and performance standards), and to measure performance (lagging and leading indicators), with synthetic examples. It also explains why low personal injury rates say little about the risk of a major accident.

1. Why barrier thinking: Piper Alpha
On 6 July 1988 the Piper Alpha platform in the UK North Sea was destroyed and 167 people died. Lord Cullen's public inquiry (1990) traced the initial leak to a condensate pump whose pressure safety valve had been removed for maintenance and replaced by an inadequately tightened blind flange; the permit-to-work system and shift handover failed to stop the pump from being restarted. Escalation followed through fire walls not designed for explosions, continued supply of hydrocarbons from connected platforms and the failure of gas risers. The inquiry led to the UK Offshore Installations (Safety Case) Regulations 1992: operators must demonstrate that major accident hazards have been identified and that risks are controlled, rather than only follow prescriptive rules.
2. Hazard identification
- HAZID: early, at concept stage, to identify major hazards, layout and siting issues from plot plans and process descriptions.
- HAZOP (IEC 61882): during detailed design, a team applies guide words (no, more, less, reverse, other than...) to parameters (flow, pressure, temperature, level, composition) for each node of the P&IDs, and records causes, consequences, safeguards and actions.
- What-if and checklist reviews: for smaller changes under management of change.
- Bow-tie analysis: to describe and manage the barriers for each major accident event during operation.
HAZOP example (synthetic): separator inlet, "more pressure". Causes: outlet control valve failing closed, a pressure transmitter failing low so the controller closes the outlet, a blocked downstream line. Consequence: overpressure of the separator and possible loss of containment. Safeguards to check: the control loop, an independent high-high pressure trip that closes the inlet, and a pressure safety valve sized for the blocked-outlet case. The team does not conclude that risk is acceptable simply because two safeguards exist: it checks that they are independent of the cause and of each other, and that the trip's required reliability has been set by a layer of protection analysis or SIL assessment (IEC 61511).
"More temperature" from an external fire is a different scenario: a vessel engulfed in fire can rupture below the relief valve set pressure because the steel loses strength, so fire protection relies on detection, emergency shutdown, depressurization (API Standard 521), fireproofing and deluge, not on the relief valve alone. Similarly, a downhole safety valve closes on loss of control-line pressure, usually commanded by the emergency shutdown system, not on low flow.
3. Bow-ties and barriers
A bow-tie places the top event (for example loss of containment of hydrocarbons from a separator) in the centre, threats on the left with preventive barriers, and consequences on the right with mitigating barriers. Synthetic example:
- Threat: overpressure. Barriers: high-high pressure trip; pressure safety valve.
- Threat: internal corrosion. Barrier: an integrity management system combining inspection, corrosion monitoring and assessment. The corrosion allowance is a design feature, and inhibitor injection is a control that keeps the barrier healthy.
- Threat: dropped objects. Barriers: lifting controls and impact protection.
- Consequence: fire or explosion. Barriers: gas detection, emergency shutdown and blowdown, ignition source control (equipment certified for hazardous areas, automatic isolation of non-essential electrical equipment), fire and blast protection, escape and evacuation.
- Consequence: toxic exposure (sour service). Barriers: H2S detection and alarms, breathing apparatus and escape routes.
A useful barrier must be able to stop the sequence on its own, be independent of the threat and of other barriers, and be auditable. Many items placed on bow-ties (training, procedures, monitoring) are better treated as controls that keep barriers effective. Industry guidance such as CCPS/Energy Institute Bow Ties in Risk Management (2018) explains these distinctions.
4. Performance standards
Each safety-critical element needs a performance standard stating its function, the criteria it must meet and how this is verified. Synthetic example for a pressure safety valve: opens at the set pressure within the tolerance of the applicable code; relief capacity covers the governing relief case in the current relief study (updated after any process change); no seat leakage; inlet and outlet lines unobstructed; any block valves secured open under a controlled system; tested and overhauled at the interval defined by the inspection programme. A barrier that fails a criterion is reported as impaired, and the operating risk is reassessed with compensating measures until it is restored.
Expected-value arguments can support maintenance budgets, as long as their inputs are stated. With an assumed demand on the valve of 0.05 per year, a probability of failure on demand of 0.10 without maintenance and a consequence cost of $45 million, the risk avoided by keeping the valve reliable is 0.05 × 0.10 × 45 million = $225,000 per year, before counting harm to people, which money does not measure.
5. Accident pyramids: useful, but not for major hazards
Heinrich (1931) and later Frank Bird (1969, from about 1.75 million reported accidents) described ratios between accidents of different severity; Bird's ratio was 1 serious injury to 10 minor injuries, 30 property damage accidents and 600 incidents with no visible injury or damage. These ratios vary widely between industries and activities, and they mix events with different causes.
The main lesson from major accidents such as the 2005 Texas City refinery explosion (CSB investigation; Baker Panel report, 2007) is that a good personal safety record does not show that process hazards are controlled. Process safety needs its own indicators, such as loss-of-containment events classified with API RP 754 or IOGP Report 456.
6. Lagging indicators
TRIR (OSHA basis) = recordable cases × 200,000 / hours worked; LTIF (IOGP basis) = lost time injuries × 1,000,000 / hours worked
Synthetic platform: on average 120 people on board working 12-hour shifts all year, so 120 × 12 × 365 = 525,600 hours. Six recordable cases (2 lost time injuries, 3 medical treatment cases, 1 restricted work case) give TRIR = 6 × 200,000 / 525,600 = 2.28 (or 11.4 per million hours), and LTIF = 2 × 1,000,000 / 525,600 = 3.81. Because one more injury would change LTIF by 1.9, small sites should use rolling averages: with 1 and 0 lost time injuries in the two previous years, the three-year LTIF is 3 × 1,000,000 / 1,576,800 = 1.90. Always state the hour base, since the OSHA and IOGP conventions differ by a factor of five. IOGP publishes annual industry figures for comparison.
7. Leading indicators
Leading indicators track whether barriers and the systems that maintain them are working. Synthetic monthly dashboard:
| Indicator | Target | Result |
|---|---|---|
| Safety-critical maintenance completed on time | 95% | 118 / 127 = 92.9% |
| Permits found compliant in audit | 90% | 22 / 28 = 78.6% |
| Near-miss reports per person per month | 0.5 | 23 / 120 = 0.19 |
| Impaired safety-critical elements open more than 30 days | 0 | 2 |
Indicators should point to action, not to targets that can be gamed: a quota of near-miss reports can generate trivial reports, so the quality of reports and of follow-up actions matters more than the count.
8. Safety culture
Westrum's typology describes organizations as pathological (information hidden, messengers punished), bureaucratic (information flows through channels) or generative (information actively sought, failures lead to inquiry). Safety climate surveys measure perceptions such as management commitment, fair treatment after incidents and the ability to stop work; they help locate problems, but scores from different questionnaires are not comparable, and results should be discussed with the workforce rather than reduced to a single number.
References
- Lord Cullen, The Public Inquiry into the Piper Alpha Disaster (HMSO, 1990)
- IEC 61882 (HAZOP)
- IEC 61511 (safety instrumented systems)
- API Standard 521
- API RP 754
- IOGP Report 456 and annual safety performance indicators
- CCPS and Energy Institute, Bow Ties in Risk Management (2018)
- F.E. Bird and G.L. Germain, Practical Loss Control Leadership
- US CSB report on the 2005 BP Texas City explosion
- R. Westrum, "A typology of organisational cultures", BMJ Quality & Safety (2004)
Related reading and community
Questions or field experience to share? Join the PetroSmart Telegram group, watch worked examples on the PetroSmart Academy YouTube channel and follow us on Facebook.