Chapter 1 · Lesson 1.6
Risk treatment, the reactor example and the LOPA table
The last step compares the calculated frequency with the target attached to the severity, and deduces the level of risk reduction to be demanded of new barriers.
Frequency targets according to severity (example)
| Severity of the final consequences | Description | Frequency target | Basics |
|---|---|---|---|
| Minor | Impact limited to the local area of the phenomenon | Variable | Depends on the cost of the protections compared with the cost of the damage |
| Serious | Serious injuries or deaths, on or off the site | 10−6 per year | Criterion defined at company level |
| Major | Impact clearly more serious than “serious” | 10−8 per year | Two orders of magnitude lower than “serious” |
These arecompany targets given as an example, not regulatory values: they should not be confused with the probability classes of the 2005 order, nor with the MMR grid. This approach amounts to a semi-quantitative risk matrix.
The values in this example (10−6 and 10−8 per year and per scenario) are clearly lower than the criteria most often encountered: of the order of 10−4 à 10−5 per year and per scenario for a death in company grids inspired by the CCPS or the British HSE. A target of 10−8 requires risk reductions that few barriers can justify. In France, the classes of the 2005 order give another reference point: class E corresponds to less than 10−5 per year. An organisation sets its targets, justifies them and applies them in the same way to all its scenarios.
The LOPA in Polyodos: frequency classes and the MMR grid
Polyodos does not reason with these company targets. It classifies the scenario frequency in five classes (those of the order of 29 September 2005) and the severity in five levels, then reads the cell of the MMR grid of the circular of 10 May 2010. The targets of the course (10−6 and 10−8 per year) remain valid as company criteria; the grid below gives the corresponding regulatory reference.
| Frequency level | Class | Range (per year) |
|---|---|---|
| Common | A | more than 10−2 |
| Probable | B | 10−3 to 10−2 |
| Improbable | C | 10−4 to 10−3 |
| Rare (very improbable) | D | 10−5 to 10−4 |
| Very rare (extremely unlikely) | E | less than 10−5 |
| Severity \ Frequency | E | D | C | B | A |
|---|---|---|---|---|---|
| Disastrous | Partial No (new sites); MMR rank 2 (existing sites) | No rank 1 | No rank 2 | No rank 3 | No rank 4 |
| Catastrophic | MMR rank 1 | MMR rank 2 | No rank 1 | No rank 2 | No rank 3 |
| Major | MMR rank 1 | MMR rank 1 | MMR rank 2 | No rank 1 | No rank 2 |
| Serious | - | - | MMR rank 1 | MMR rank 2 | No rank 1 |
| Moderate | - | - | - | - | MMR rank 1 |
In Polyodos, each LOPA line carries the frequency of the initiating event, the barriers (IPL) with their PFD, and the resulting residual frequency; this frequency is placed in a class from A to E, which gives the cell of the grid. Example: the reactor below (residual frequency 8.8·10−7 per year, class E) with a “major” severity falls in MMR rank 1, whereas the same scenario with a “disastrous” severity would remain “Partial No”. The course target (10−6 per year) is therefore stricter than the MMR cell for a “major” severity: the two readings complement each other but must not be confused.
Two application examples
- Scenario with F = 10−6 per year and “major” severity: the target is 10−8, hence a reduction factor of 100 (two levels). A SIL 2 SIS is added (PFD between 10−3 and 10−2), or two independent SIL 1 barriers.
- Scenario with F = 10−6 per year and “serious” severity: the target of 10−6 is met; the risk is tolerable, no additional measure is needed.
Complete example: reactor with loss of cooling
A stirred reactor is cooled by cooling water. If cooling is lost, the temperature rises, the pressure increases and the reactor may explode. Consequence: explosion with serious effects, “serious” severity (target 10−6 per year).
- Initiating event : cooling failure, frequency 0.1 per year.
- IPL 1 : temperature alarm and manual opening of the spray by the operator, PFD = 0.1.
- IPL 2 : safety valve, PFD = 0.01.
- Barrier to be defined : automatic spray by an SIS (temperature sensor, logic, spray opening valve, supplied by a back-up water tank).
LOPA table
| N° | Description | Severity | Initiating event (frequency) | Alarm + manual action | SIS (automatic spray) | Valve | Final frequency |
|---|---|---|---|---|---|---|---|
| 1 | Reactor explosion | Serious | Cooling failure, 0.1 per year−1 | 0.1 | PFDSIS to be determined | 0.01 | 0.1 × 0.1 × 0.01 × PFDSIS = 10−4 × PFDSIS |
Without SIS, the frequency is 0.1 × 0.1 × 0.01 = 10−4 per year, i.e. 100 times the target of 10−6. The required reduction factor is 10−6 / 10−4 = 10−2 : the SIS must have PFDSIS ≤ 10−2.
A SIL 1 covers [10−2 ; 10−1[ and therefore does not guarantee the target; we retain a SIL 2. With the calculation of the lesson 4-2 of the course on SIS and SIL (λDU = 2 × 10−6 h−1, annual test), PFDavg = 8.76 × 10−3 and the final frequency is 10−4 × 8.76 × 10−3 = 8.8 × 10−7 per year, lower than 10−6 : the target is met, with a small margin (about 12%) to be monitored (tests, maintenance).
LOPA answers three questions: is the residual risk acceptable? Which barrier should be added? Which SIL should be required of it? The independence of the barriers remains the assumption to check first.
Conclusion on the method
- A semi-quantitative, readable and global method that efficiently assesses the barriers in place.
- It makes it possible to quantify the investment in additional barriers and the gain obtained.
- It fits into the risk management approach (ISO 31000).
- Limit: its result depends on the quality of the frequencies and PFDs retained, and on the reality of the independence between the barriers.