Key points
With an ever-increasing focus on reducing the cost of maintenance and the need for safety critical equipment, such as pressure safety valves (PSVs), to function reliably, operators need to ensure maintenance is targeted at the most critical equipment.
Being a system that only operates on demand, the only way to be certain that a PSV will operate is for it to be removed from its normal operating location offshore and subject to an as-received pre-overhaul pop (pre-pop) test, to determine whether it ‘lifts’ at the correct pressure.
The combination of logistics to transport PSVs, pre-pop testing, and refurbishment incurs considerable cost to the operator. In-situ online tests such as ‘trevi-testing’ are cheaper, but they do not identify all failure modes and are only ever an interim solution.
The frequency and hence, associated cost of PSV inspection intervals needs to be optimised to ensure the residual risk from their potential failure is as low as reasonably practicable (ALARP). Determining an optimised maintenance interval for a single PSV can be straightforward; the challenge comes when dealing with the sheer number of PSVs on a typical installation, usually many hundred.
Available methodologies are generally characterised by being very simple, even such that the same methodology could logically be applied with a change of units – weeks to months for example – or are very complex and have enough latitude in the parameters used to provide any result that is desired.
DNV GL has solved this issue by developing a risk-based inspection (RBI) methodology that uniquely combines qualitative and quantitative analysis to allow optimisation of maintenance intervals and improve inspection management.
This multi-layered approach gains the most information possible from the pre-pop data. Presentation and understanding of the results are maximised through a digital, cloud-based interactive dashboard used by maintenance teams, technical authorities and inspection engineers for better decision-making.
Pressure Safety Valve failure rates and consequence analysis
PSV failure modes include blockage due to scale or wax and mechanical drifting of the set-point, both of which are time dependent and hard to predict.
Unexpected process conditions may also lead to failure, for example, water in air systems that leads to corrosion. Given the difficult of predictive analysis, the question then arises of how to account for these failure modes in an RBI methodology.
In carrying out rigorous RBI for two drill rigs, ten offshore platforms, one FPSO and an onshore terminal, half a billion service hours of data with around 15,000 pre-pop test results has been gathered. This is 500 times as much data as is in the industry standard OREDA database.
From this depth and wealth of service history, it is apparent that a large proportion of PSV failures are associated with the properties of the fluid in the system they are protecting rather than the PSV itself.
Quantitative assessment of Pressure Safety Valves
The dependency on the fluid, or service that the PSV sees means that to carry out the quantitative part of the analysis, pre-pop data for PSVs in the same service type is amalgamated.
The quantitative part of the analysis determines a failure rate for PSVs with the same service e.g. air, or production hydrocarbon gas and uses this to derive a risk of a PSV failing to lift on demand.
A limit on the residual risk from the potential for a PSV not to operate on demand has been defined: in this case a potential loss of life of 10-5 per year, which for a typical platform equates to an individual risk of around 10-7 per year. The following equation then defines the maximum inspection interval:
Potential loss of life = Failure rate x Inspection Interval x Demand x Consequence /2
This calculation to define the Inspection Interval is made for each PSV using the failure rate for PSVs seeing the same service: this is typically a few tens of PSVs at each site, though for some utility systems such as hydraulics, it may be more.
The consequences are calculated for each PSV across the range of failure modes from minor mechanical failure to full, ignited rupture and are dependent on:
- Size of equipment protected
- Set pressure
- Fluid phase
- Fluid flammability
- The number of personnel in the area, though this is subject to a defined minimum to avoid a longer maintenance level for an area that may have a low population, which could change at a later date.
The demand rate is fixed depending on the type of PSV (pressure, thermal, or fire) and is based on an average demand rate calculated from detailed fault trees across a number of example systems.
The inspection interval calculated above is the maximum possible interval that keeps the risk below the limit. This upper limit can be longer (say >10 years) than is normally seen due to the low risk nature of the service, or good PSV performance, though it is normally limited to six years (or other operator chosen maximums, but if the history is poor and the service is relatively high hazard, then the quantitative limit can be less than six years and this may limit the maintenance interval.
Thus, for PSVs with a lower risk profile, for example, chemical injection, which has few failures and lower consequences, the quantitative limit is highly unlikely to bite, but for any process system with a higher failure rate, it may limit the maintenance interval.
A failure rate may change over time e.g. if chemical injection processes improve and the potential for PSV fouling decreases. In calculating the failure rate, the methodology considers whether all of the pre-pop data should be used, or whether there is a significant change in performance that needs to be accounted for and a subset of the data used (always using the most recent data).
The Chi-squared test is used to test significance. Figure 2 shows the number of successful and unsuccessful pre-pop tests for high pressure clean liquids over two decades. This reveals a worsening in performance after 2015. Therefore, the failure rate is calculated from 2015 only and this would increase the failure rate used and decrease the maximum maintenance interval.
The premise behind the quantitative part of the methodology is that failures are random. If they are not random and concentrated in a particular subset of PSVs due to, say, wax, it is identified in the qualitative part of the methodology. Also, an additional service type (maybe waxy crude in this example) could be made to give the subset its own higher failure rate.
With the quantitative part of the methodology setting an upper limit, the qualitative element sets the actual limit.
Qualitative assessment or Pressure Safety Valves
The qualitative part of the assessment considers performance data from pre-pop tests of a specific PSV and any that operate under identical process conditions. A number of basic premises cover the qualitative approach.
A PSV failure may be due to a random event that will not be repeated, or a specific cause that must either be resolved, or maintenance intervals reduced to account for it. In most cases, it will not be possible to be certain that it was just a random event that will not be repeated and so a cautious approach to failures must be taken and intervals reduced following a failure.
Similarly, intervals must be increased cautiously. This leads to two basic rules in the first part of the qualitative assessment:
- A failed pre-pop test halves the inspection interval
- Repeated successful prep-pop tests means the interval is increased by 50%.
This is always subject to the maximum quantitative interval that is calculated and a maximum overall interval of six years, dependent on operator discretion.
Due to the complexity of historic data and, for example, the fact that opportune tests are sometimes carried out in turnarounds (process shutdowns), the qualitative part of the assessment has more nuances, but the two rules above define the overall approach.
In an RBI scheme, each PSV could be considered on its own merit, but, given the inevitable gaps in the pre-pop testing records, this would limit the benefit of the RBI.
However, groups of PSVs operate under identical conditions and so can be expected to exhibit similar failure modes. Therefore, the data from a group of PSVs is pooled in the second stage of the qualitative assessment.
Consideration of the pre-pop results from a whole group may mean that the inspection interval following a failure does not need to be halved. For example, for a group with passes at three years and a failure for a single valve at four years, the interval for this valve need only be reduced to three years. Without consideration of the group pre-pop history, the failed valve would have its intervals reduced to two years.
In summary, to set the interval of a PSV, there are three stages each using an increasingly large data set, but with each stage having less influence on the final result:
- Qualitative assessments for:
- The specific PSV under consideration
- PSVs in a group seeing exactly the same conditions (i.e. fluid type and pressure)
- Quantitative assessment for:
- PSVs seeing a similar fluid to define the failure rate and maximum possible interval.
In working with operators, DNV GL has found that application of the combined qualitative and quantitative RBI methodology has focussed inspections in the right area, reduced risk and saved money.
It has also shown that collecting and making use of half a billion-service hours of data need not be an onerous task for oil and gas leaders and decision-makers.
For some assets, analysis has shown that maintenance intervals can be doubled, creating significant annual cost savings by concentrating effort on the worst performing PSVs.
Wider applications of Pressure Safety Valves
The approach can also be expanded to other equipment types, such as:
- Fire and gas detection
- Emergency shutdown valves
- Blowdown valves
- Safety Critical Instrumentation
- Deluge.
For 5,000-plus PSVs, data management is a key issue and a dashboard approach is used to allow access by the operator and DNV GL engineers. This can be used to view PSV failures and the results of the RBI.
The calculated inspection interval, and changes to the intervals and failures, allows a clearer and more comprehensive picture of PSV performance to be developed. Maintenance planning also becomes easier and information required for deferral assessments is available in one place, in seconds rather than hours.
In Figure 3, the example dashboard shows all the available data for 27 pre-pop test results for diesel PSVs across many years.
Conclusion
RBI models for PSVs tend to be highly complex, or overly simplistic. Even with half a billion hours of data, a complex model cannot be justified as the uncertainties in the model swamp the result.
A simple model is open to criticism on many grounds including questioning on what the numerical change in risk is associated with wide-scale PSV inspection interval change. A combined quantitative and qualitative analysis has several distinct advantages, including:
- It considers the behaviour of the specific PSV
- It includes data from a wider set or group of PSVs to maximise the possibility of justifiably extending an interval, or limiting the interval for a set of PSVs after a pre-pop failure is observed
- The change in risk from a change in interval is known and the risk from potential failure of any PSV is limited
- There is sufficient confidence in the results with intervals only being extended cautiously.
Typically, the first application of the RBI scheme can lead to upwards of 20% reduction in maintenance, but figures of 50% can be achieved in some instances. Risk is also reduced by concentrating maintenance efforts on areas of higher risk.
By improving visibility of the PSV’s performance, the total cost of PSV change-out reduces because it is being done less often and there is also the possibility of greater savings if a longer PSV maintenance intervals avoids, or shortens a shutdown.
The system is most beneficial when used on a regular basis to take advantage of the gradually improving performance of the PSVs over the lifetime of the facility. Use of the PSV dashboard, rather than complex interaction with a maintenance management system to find the history of a PSV, is also a highly beneficial output of the process.
