A foundational element of this risk assessment methodology is that all available information be utilized appropriately. All inputs into the risk assessment can be thought of as evidence since inputs lead to our conclusions about risk levels.
Handling evidence requires consideration of that evidence’s accuracy. This includes considering its age, error potential, relevance, etc, all of which can be under an umbrella of ‘confidence’. For our purposes here, we can equate confidence to ‘age and accuracy’ of the evidence.
Proper handing of evidence will also often require parallel ‘tracks’ when quantifying risk values that are based on multiple pieces of evidence. It will also require resolution of apparent conflicts in evidence.
Using parallel tracks for various sources of evidence reflects the reality that we have multiple sources of information giving us knowledge about the same thing. For instance, corrosion rate can be inferred from soil characteristics, observed metal loss (eg, from ILI), and others. Each source can be thought of as ‘evidence’. The strength of each piece of evidence is related to its age and original accuracy (uncertainty).
Confidence
A question often arises: how do we calculate confidence% for this input? As noted above, all inputs should be thought of as evidence. Some are measurements while others are indirect estimates. All types are adjusted to the desired conservatism level via confidence. Confidence can be a separate input but, if not, each input should nonetheless take confidence into account.
Confidence can be formally estimated by considering age and accuracy of the evidence. Oftentimes, formality is not required and a more subjective estimate of confidence is fine. There can also be a combination of formal and informal guidelines. For instance, maybe confidence in P/S reading as a measure of CP effectiveness is initially 90% (considering meter accuracy, error potential, variability of conditions, etc), and this confidence ‘degrades’ 20% per year. So, a 3-yr old CIS suggests that confidence in CP negating all external corrosion in this area is now 90%-3 x 20% = 30%.
Parallel Tracks: Measurements vs Estimates
Use of coupons, probes, successive ILI, etc to estimate degradation rates, in addition to chemistry-based estimates of corrosion and cracking, is an example of measurements vs estimates. Either may provide the more accurate estimate so both must be considered.
To address multiple information sources potentially informing the same value, there should be parallel ‘tracks’. A good example is quantification based on measurements vs estimates based on known or assumed characteristics. That is, measuring something versus inferring its value via what is known about it.
Risk assessments of time-dependent failure mechanisms rely on both materials science based estimates of possible degradation and actual measurements of degradation, often extrapolated (measurements at one location are modeled to represent values at others). This manifests as parallel analyses paths in the assessment where the most recent and most accurate information plays the larger role in the assessment. See the analogous discussion regarding effective pipe wall estimates.
For instance, external corrosion rates (mpy) may be measured via coupons or successive wall thickness measurements–eg, the second ILI compares findings to the first, to measure changes in anomaly depth over time. Alternatively, corrosion rates may also be estimated based on soil chemistry and pipe material properties.
The recommendation is to include both in the risk assessment. Each track–measurements vs estimates–produces a value for mpy which should be adjusted for confidence. With a fair and consistent application of this adjustment, the more optimistic value can govern. That way, more reliable information can override the less reliable.
Apparent Conflicts in Use of Evidence
When multiple information sources potentially inform estimates of the same variable, there will often be potentially conflicting indications that must be sorted out. Choices in conservatism will be critical in this analyses. As a general rule, new inspection information should reduce risk. This is because a conservative risk assessment always assumes that things are getting worse unless information indicates otherwise–“guilty until proven innocent”. If new inspection findings result in surprises, then the risk assessment failed to use an appropriate level of conservatism. The inspection showed actual conditions to be worse than assumed–a very undesirable situation.
General areas where many disparate pieces of information should be used, include:
-
- MPY degradation rates, potentially based on:
- electrolyte (soil, water, etc) chemistry
- successive ILI measurements
- probes, coupons
- AC induction
- etc
- Available wall thickness, often based on:
- nominal wall
- ILI, especially UT
- in ditch NDE
- inferred by pressure test
- last assessment minus degradations since then
- etc
- MPY degradation rates, potentially based on:
For instance, when degradation mechanisms are not directly observable, the assessment must use mostly indirect evidence to infer damage potential. This is consistent with the historical practice of corrosion control on buried pipelines–estimate potential corrosion rates and plan mitigation accordingly.
Then, any detection of degradation damages or direct measurements of actual degradation rate can be used to calibrate the previous inferential data and/or tune the risk assessment. Where a degradation rate is actually measured or confirmed to ‘not have happened’, the risk assessment should include this information. Caution must be exercised in assigning favorable rates based solely on the non-detection of significant damages at certain times and at limited locations. It is important to note that the potential for some corrosion or cracking damages can be high even when no active damage is detected during a sampling process, especially a random sampling process.
When an inspection detects corrosion or cracking damage, it is logical to assume that damage potential existed at one time and may still exist. When there is actual damage, but risk assessment results do not indicate a significant potential for such damage, then a disconnect seemingly exists between the direct and the indirect evidence. Such conflicts are discussed here as well as in discussions around uncertainty and validation.
Identifying the location of the inconsistency is necessary. The conflict could reflect an overly optimistic assessment of effectiveness of mitigation measures (coatings, CP, etc.) or it could reflect an underestimate of the harshness of the environment. Another possibility is that detected damages do not reflect active mechanisms but only past and now-inactive mechanisms. For instance, replacing anode beds, increasing current output from rectifiers, eliminating interferences, and recoating are all actions that could halt previously active corrosion.
The degradation estimates should always include the best available inspection information. The risk assessment should preferentially use recent direct evidence over previous assumptions, until the conflicts between the two are investigated.
For example, suppose that, using information available prior to an ILI, the assessment concluded a low probability of subsurface corrosion because both coating and CP were estimated to be fully effective. If the ILI recent inspection indicates that some external metal loss has occurred, then the subsurface corrosion assessment would be suspect, pending an investigation. The previous assessment based on indirect evidence should probably be initially overridden by the results of the recent ILI pending an investigation to determine the cause of the damage—how and when the mitigation measures may have failed as well as how the ILI assessment may be inaccurate.
If the risk assessment is modified based upon un-verified ILI results, it can later be improved with results from more detailed examinations. That is, conducting excavation, inspection, and verifications to determine if anomalies are present and represent loss of resistance. If a root cause analysis of the detected damages concludes that active corrosion is not present, the original risk assessment may have been correct. The root cause analysis might demonstrate that corrosion damage is old and future corrosion has been mitigated.
Techniques to assimilate ILI and other direct inspection information into risk estimates are discussed in Inspections/Integrity Assessments and Optimizing Inspection Results.
See also: Age as Evidence and Test of Time as Evidence in Data Integration.