PL Safety and Essentials of Risk Assessment

2,249 views
2,094 views

Published on

During the conference ''Buisleidingen Platform'' Kent Muhlbauer (DNV) held a presentation about PL safety and essentials of risk assessment. Go to www.iir.nl/buisleidingen for more information.

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
2,249
On SlideShare
0
From Embeds
0
Number of Embeds
1,683
Actions
Shares
0
Downloads
28
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

PL Safety and Essentials of Risk Assessment

  1. 1. PL Safety and Essentials of Risk AssessmentFollowing the Essential Elements
  2. 2. Background© Det Norske Veritas AS. All rights reserved. 2
  3. 3. © Det Norske Veritas AS. All rights reserved.
  4. 4. Kalamazoo River, 2010 10ft creek CoF: $1B PoF: 1/1000yr Expected Loss: $1M/yr for ~3m of pipeline! $1,000,000,000 spent© Det Norske Veritas AS. All rights reserved.
  5. 5. Reality Check  RM is not new; requires RA  Risk-based decision-making is complex - Because the real world is complex, measuring risk is complex - RM is even more complex than RA  Dealing with the complexity is worthwhile - increases understanding - shows full range of options; many opportunities to impact risk - cheaper than prescriptive ‘solutions’ - Improves decision-making - the key to improving PL safety© Det Norske Veritas AS. All rights reserved.
  6. 6. Reality Check, Part Two If you put tomfoolery into a computer, nothing comes out of it but tomfoolery. But this tomfoolery, having passed through a very expensive machine, is somehow ennobled and no-one dares criticize it. - Pierre Gallois The Illusion of Knowledge© Det Norske Veritas AS. All rights reserved.
  7. 7. How good is PL RA now? Pockets of strengths But Weaknesses are apparent - A complicated topic--incomplete understanding is common - Lack of standardized RA approaches - Errors in some industry standards - New demands by regulators and other stakeholders - Myths and misconceptions© Det Norske Veritas AS. All rights reserved. 7
  8. 8. PL Risk Modeling Confusion ASME B31.8s •Subject Matter Experts Qualitative •Relative Assessments Quantitative •Scenario Assessments Semi-quantitative •Probabilistic Assessments Probabilistic Deterministic© Det Norske Veritas AS. All rights reserved.
  9. 9. Regulations: IMP Objectives vs RA Techniques Objectives (a) prioritization of pipelines/segments for scheduling integrity assessments and mitigating action (b) assessment of the benefits derived from mitigating action (c) determination of the most effective mitigation measures for the identified threats (d) assessment of the integrity impact from modified inspection intervals (e) assessment of the use of or need for alternative inspection methodologies (f) more effective resource allocation Techniques • Subject Matter Experts • Relative Assessments • Scenario Assessments • Probabilistic Assessments Numbers Needed •Failure rate estimates for each threat on each PL segment •Mitigation effectiveness for each contemplated measure •Time to Failure (TTF) estimates (time-dep threats)© Det Norske Veritas AS. All rights reserved.
  10. 10. “Threat Interaction” Confusion  Time Dependent Threats  Stable Threats Potential Weaknesses - Corrosion  Manufacturing-related flaws in - External - Pipe body - Internal - Pipe seam - Cracking - Fatigue  Welding / Fabrication-caused flaws in - EAC - Girth welds - Fabrication welds  Time Independent Threats - Wrinkled / buckled bend - Third party - Threads / couplings - Geohazards  Defects present in equipment - Incorrect operations - Gaskets, O-rings - Sabotage - Control / relief devices - Seals, packing - Other equipment© Det Norske Veritas AS. All rights reserved. 10
  11. 11. Myths: Data Availability vs Modeling Rigor Myth:  Some RA models are better able to accommodate low data availability Reality:  Strong data + strong model = accurate results  Weak data + strong model = uncertain results  Weak data + weak model = meaningless results 11© Det Norske Veritas AS. All rights reserved.
  12. 12. Myth: QRA / PRA Requirements Myth:  QRA requires vast amounts of incident histories Reality:  QRA ‘requires’ no more data than other techniques  All assessments work better with better information  Footnotes: - Some classical QRA does over-emphasize history - Excessive reliance on history is an error in any methodology 12© Det Norske Veritas AS. All rights reserved.
  13. 13. © Det Norske Veritas AS. All rights reserved.
  14. 14. © Det Norske Veritas AS. All rights reserved.
  15. 15. © Det Norske Veritas AS. All rights reserved.
  16. 16. Removing the Weaknesses—The New Look of PL RA PoF (len adjusted) 140 120 100 Frequency 80 60 PoF (unitized) 40 12.0% 20 10.0% 8.0% 0 6.0% 0. 0 1 0. 01 0. 04 0. 7 0. 13 0. 16 0. 19 0. 22 0. 25 0. 28 1 e 00 03 or 00 0 0 0 0 0 0 0 0 M 00 00 00 00 00 00 00 00 00 00 4.0% 0. Bin 2.0% 0.0% 0 20000 40000 60000 80000 100000 120000 140000 station© Det Norske Veritas AS. All rights reserved.
  17. 17. Essential Elements© Det Norske Veritas AS. All rights reserved. 17
  18. 18. Essential Elements  The Essential Elements are meant to - Minimize the weaknesses in current PL RA practice - Be common sense ingredients that make risk assessment meaningful, objective, and acceptable to all stakeholders - Be concise yet flexible, allowing tailored solutions to situation-specific concerns - Lead to smarter risk assessment  The elements are meant to supplement, not replace, guidance, recommended practice, and regulations already in place© Det Norske Veritas AS. All rights reserved. 18
  19. 19. The Essential Elements Measurements in Verifiable Units Proper Probability of Failure Assessment Characterization of Potential Consequences Full Integration of Pipeline Knowledge Sufficient Granularity Bias Management Profiles of Risk Proper Aggregation© Det Norske Veritas AS. All rights reserved. 19
  20. 20. Measure in Verifiable Units Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  Must include a definition of “Failure”  Must produce verifiable estimates of PoF and CoF in commonly used measurement units  PoF must capture effects of length and time  Must be free from intermediate schemes (scoring, point assignments, etc) “Measure in verifiable units” keeps the process transparent by expressing risk elements in understandable terms that can be calibrated to reality© Det Norske Veritas AS. All rights reserved.
  21. 21. Measure in Verifiable Units Risk Estimates as Measurements Risk = Frequency of consequence - Temporally - Spatially TTF = time to failure •Incidents per mile-year remaining life estimates in years •fatalities per mile-year •dollars per km-decade conseq prob© Det Norske Veritas AS. All rights reserved.
  22. 22. Measure in Verifiable Units Why measurements instead of scores? Direct measurements are efficient - Less subjective - Anchored in ‘real world’, able to capture real world phenomena - Defensible, auditable, transparent - Avoids need for standardization - Avoids erosion of score definitions - Allows calculation of costs and benefits - Supports better decisions© Det Norske Veritas AS. All rights reserved.
  23. 23. Probability of Failure Grounded in Engineering Principles Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  All plausible failure mechanisms must be included in the assessment of PoF  Each failure mechanism must have the following elements independently measured: - Exposure - Mitigation - Resistance  For each time dependent failure mechanism, a theoretical remaining life estimate must be produced© Det Norske Veritas AS. All rights reserved.
  24. 24. Probability of Failure Grounded in Engineering Proper PoF Characterization Principles Independent Evaluations of:  Exposure: likelihood and aggressiveness of a failure mechanism reaching the pipe when no mitigation applied (ATTACK)  Mitigation: prevents or reduces likelihood or intensity of the exposure reaching the pipe (DEFENSE)  Resistance: ability to resist failure given presence of exposure (SURVIVABILITY )© Det Norske Veritas AS. All rights reserved.
  25. 25. Probability of Failure Grounded in Engineering Simple and Robust Relationships are Available Principles Probability of Failure (PoF) = Exposure x (1 - Mitigation) x (1 – Resistance) Probability of Damage TTF = exposure * (1 – mitigation) / resistance PoF = f (TTF)© Det Norske Veritas AS. All rights reserved.
  26. 26. Probability of Failure Grounded in Engineering Estimating Threat Exposure Principles Events per mile-year for time independent mechanism - third party - incorrect operations - weather & land movements MPY for degradation mechanisms - Ext corr - Int corr - Cracking (EAC / fatigue)© Det Norske Veritas AS. All rights reserved.
  27. 27. Estimating Mitigation Measure Effectiveness Strong, single measure Or Accumulation of lesser measures Cathodic Public Maint Pigging protection Patrol Education system Coating Depth of Casing Training & Chem Inhibition system cover Competency Exposure Damage Slide 27© Det Norske Veritas AS. All rights reserved.
  28. 28. Probability of Failure Grounded Estimating Resistance in Engineering Principles  Pipe spec (original)  Required pipe strength - Normal internal pressure  Historical issues - Normal external loadings - Low toughness - Hard spots - Seam type - Manufacturing  Pipe spec (current) - ILI measurements - Calcs from pressure test - Visual inspections - Effect of estimated degradations© Det Norske Veritas AS. All rights reserved.
  29. 29. Probability of Failure Grounded in Engineering Best Estimate of Pipe Wall Today Principles Adjust measurements for age and accuracy Measurement error Degradation Since Meas Current Estimate Press Test 1992 8 mpy x 21 yrs = 168 mils (inferred) +/- 5% ILI 2005 8 mpy x 8 yrs = 64 mils +/- 15%© Det Norske Veritas AS. All rights reserved.
  30. 30. PoF: Critical Aspects© Det Norske Veritas AS. All rights reserved.
  31. 31. Fully Characterize Consequence of Failure Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  Must identify and acknowledge the full range of possible consequence scenarios  Must consider ‘most probable’ and ‘worst case’ scenarios HCA Spill path PL Hazard Zone© Det Norske Veritas AS. All rights reserved.
  32. 32. Fully Characterize Consequence Consequence Estimation Overview of Failure Sequence of Analysis 1. Chance of failure (threat models) Probability 2. Chance of failure hole size 3. Spill size (considering leak detection and reaction scenarios) - Volume Out 2 4 5 6 4. Chance of ignition Product Hole size Hole size Ignition scenario Ignition Distance thermal from source hazard Contaminati on hazard Total probability of hazard probability probability (ft) - Immediate immediate ignition 5% (ft) 0 zone (ft) 400 zone (ft) 0 400 zone 0.2% - Delayed rupture 4% delayed ignition no ignition 10% 85% 600 600 500 0 400 900 1100 1500 0.4% 3.4% - None immediate ignition 2% 0 200 0 200 0.3% oil medium 16% delayed ignition 5% 200 300 200 500 0.8% no ignition 93% 200 0 500 700 14.9% 5. Spill dispersion immediate ignition 1% 0 50 0 50 0.8% delayed ignition 2% 80 100 0 180 1.6% - Pipeline/product characteristics small 80% no ignition 97% 80 0 80 160 77.6% - Topography (if liquid release) - Meteorology (if gaseous release) 6. Hazard area size and probability (for each scenario) 7. Chance of receptor(s) being in hazard area (counts, density, or area) 8. Chance of various damage states to various receptor (including consequence mitigation) 9. Calculate Expected Loss (Prob x Consequence $)© Det Norske Veritas AS. All rights reserved. 32
  33. 33. Integrate Pipeline Knowledge Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  The assessment must include complete, appropriate, and transparent use of all available information  ‘Appropriate’ when model uses info as would an SME© Det Norske Veritas AS. All rights reserved.
  34. 34. Integrate Pipeline External Corrosion Model Knowledge EC POF (prob/mile-yr) EC TTF (Years –assuming a per mile basis ) Available Pipe Wall (in) Growth Rate (mpy) Estimate x AdjustmentEstimate (in) Adjustments (%) Estimate (mpy) Total mpy x (1-Mitigation) Measured (mpy)Max based on: Cumulative:1. NOP Environment (mpy) Mitigation (%) based on Active 1. Joint Type Direct measurements2. Hydrotest Sum Corrosion Locations 2. Reinforcements adjusted by3. NDE/ILI 3. Manuf & Const Confidence2&3 adjusted for 4. Pipe Typempy growth since 1. Above/Below Ground 5. Toughness CP Gaps (Prob of gaps/ft)measurement 2. Atmospheric CGR (mpy) 6. Flaws Sum of gaps/mi converted to probability External Coating Holiday Rate 7. External Loads3. Electrical Isolation (%) 8. Spans 4. Soil based CGR (mpy) 1. Corrosivity 2. Moisture Content CP Effectiveness CP Interference Estimated (defects/mi) 3. MIC 1. Defects/mi adjusted by 5. Mitigated AC Induced confidence CGR (mpy) Measured Gaps /mi Locations/mi: Measured (defects/mi) 1. CP Readings 1. DC Sources 1. Defects/mi adjusted by adjusted by (mitigated) confidence and age confidence 2. Coating Estimated Gaps/mi Shielding 1. Distance from 3. Casing test station Shielding 2. PL Age 3. Criteria 4. Rectifier out of service history© Det Norske Veritas AS. All rights reserved. 34
  35. 35. Incorporate Sufficient Granularity Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  Risk assessment must divide the pipeline into segments where risks are unchanging  Compromises involving the use of averages or extremes can significantly weaken the analysis and are to be avoided© Det Norske Veritas AS. All rights reserved.
  36. 36. Incorporate Sufficient Dynamic Segmentation Granularity Due to the numerous and constantly-varying factors effecting the risk to the pipeline, proper analysis will require at least 10 segments per mile (not uncommon to have thousands of segments per mile) Steel Pipe wall 0.320” Pipe wall 0.500” 1995 1961 Landslide Threat Population Class 3© Det Norske Veritas AS. All rights reserved.
  37. 37. Control the Bias Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  Risk assessment must state the level of conservatism employed in all of its components  Assessment must be free of inappropriate bias that tends to force incorrect conclusions© Det Norske Veritas AS. All rights reserved.
  38. 38. Control the Bias Certainty “Absolute certAinty is the privilege of fools And fAnAtics.”© Det Norske Veritas AS. All rights reserved.
  39. 39. Control the Understanding Conservatism and Uncertainty Bias A way to measure and communicate conservatism in risk estimates - PXX - P50 - P90 - P99.9 Useful in conveying intended level of conservatism© Det Norske Veritas AS. All rights reserved.
  40. 40. The Role of Historical Incidents Control the Bias Problems:  Historical data usefulness in current situation  Small amount of data in rare-event situations  Representative population  Behavior of the individual vs population weightings© Det Norske Veritas AS. All rights reserved.
  41. 41. Profile the Risk Reality Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  The risk assessment must be performed at all points along the pipeline  Must produce a continuous profile of changing risks along the entire pipeline  Profile must reflect the changing characteristics of the pipe and its surroundings© Det Norske Veritas AS. All rights reserved.
  42. 42. Risk Profile: Passing the ‘Map Point’ Test Profile the Risk Reality© Det Norske Veritas AS. All rights reserved.
  43. 43. Profile the Profile to Characterize Risk Risk Reality Scenario 1 100 km oil pipeline widespread coating failure river parallel remote Scenario 2 50 km gas pipeline 2 shallow cover locations high population density high pressure, large diameter© Det Norske Veritas AS. All rights reserved.
  44. 44. Profile the Risk Reality Risk Characterization Scenario 2 Scenario 1 50 km gas pipeline 100 km oil pipeline 2 shallow cover locations widespread coating failure high population density river parallel high pressure, large diameter remote location EL EL km km Very different risk profiles© Det Norske Veritas AS. All rights reserved.
  45. 45. System Profiles Compared© Det Norske Veritas AS. All rights reserved.
  46. 46. Proper Aggregation Probability of Measure in Fully Failure Integrate Incorporate Characterize Profile the Risk Unmask Verifiable Grounded in Consequence of Pipeline Sufficient Control the Bias Aggregation Engineering Knowledge Granularity Reality Units Failure Principles  Proper process for aggregation of the risks from multiple pipeline segments must be included  Summarization of the risks from multiple segments must avoid simple statistics or weighted statistics that mask the actual risks© Det Norske Veritas AS. All rights reserved.
  47. 47. Unmask Aggregation Aggregating Risks for Collection of Pipe Segments PoF total = PoF1 + PoF2 + PoF3 + PoF4 + …. PoFn PoF total = 137% . . . ? Simple sum only works when values are very low. Len-weighted avg masks ‘weak link in chain’© Det Norske Veritas AS. All rights reserved.
  48. 48. Unmask Aggregating Risks Aggregation PoF total = Avg(PoF1, PoF2, …PoFn) Avg PoF = Avg PoF But PoF PoF ≠ KM KM© Det Norske Veritas AS. All rights reserved.
  49. 49. Unmask Aggregating Risks Aggregation PoF total = Max(PoF1, PoF2, …PoFn) Max PoF = Max PoF But PoF PoF ≠ KM KM© Det Norske Veritas AS. All rights reserved.
  50. 50. Aggregating Risks PoF total = Max(PoF1, PoF2, …PoFn) Max PoF = Max PoF But PoF PoF ≠ KM KM© Det Norske Veritas AS. All rights reserved.
  51. 51. Unmask Aggregation Aggregating Failure Probabilities Overall PoF is prob failure by [(thd pty) OR (corr) OR (geohaz)…] PoS = 1 - PoF Overall PoS is prob surviving [(thd pty) AND (corr) AND (geohaz)….] So… PoF overall = 1-[(1-pfthdpty) x (1-pfcorr) x (1-pfgeohaz) x (1-pfincops)]© Det Norske Veritas AS. All rights reserved.
  52. 52. The Essential Elements Measurements in Verifiable Units Proper Probability of Failure Assessment Characterization of Potential Consequences Full Integration of Pipeline Knowledge Sufficient Granularity Bias Management Profiles of Risk Proper Aggregation© Det Norske Veritas AS. All rights reserved. 52
  53. 53. The New Look of PL RA PoF (len adjusted) 140 120 100 Frequency 80 60 PoF (unitized) 40 12.0% 20 10.0% 8.0% 0 6.0% 0. 0 1 0. 01 0. 04 0. 7 0. 13 0. 16 0. 19 0. 22 0. 25 0. 28 1 e 00 03 or 00 0 0 0 0 0 0 0 0 M 00 00 00 00 00 00 00 00 00 00 4.0% 0. Bin 2.0% 0.0% 0 20000 40000 60000 80000 100000 120000 140000 station© Det Norske Veritas AS. All rights reserved.
  54. 54. Intended Outcomes© Det Norske Veritas AS. All rights reserved. 54
  55. 55. Intended Outcomes of Application of EE’s  Efficient and transparent risk modeling  Accurate, verifiable, and complete results  Improved understanding of actual risk  Risk-based input to guide integrity decision-making: true risk management  Optimized resource allocation leading to higher levels of public safety  Appropriate level of standardization facilitating smoother regulatory audits - Does not stifle creativity - Does not dictate all aspects of the process - Avoids need for (high-overhead) prescriptive documentation  Expectations of regulators, the public, and operators fulfilled© Det Norske Veritas AS. All rights reserved. 55
  56. 56. If you don’t have a number, you don’t have a fact, you have an opinion.© Det Norske Veritas AS. All rights reserved.
  57. 57. Safeguarding life, property and the environment www.dnv.com Kent Muhlbauer wkm@pipelinerisk.com© Det Norske Veritas AS. All rights reserved. 57

×