1 of 12

Verified Uncertainty Quantification for Infrastructure Safety

Roger Paredes Leonardo Dueñas-Osorio

Dept. of Civil and Environmental Engineering, Rice University

Spring 2022 Virtual Semi-Annual Meeting

April 29, 2022

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

2 of 12

System-level Uncertainty Quantification

Quantitative support for decision analysis

Data and models

  • Hazard
  • Fragility
  • Restoration

Community goals

  • Robustness
  • Rapidity

Uncertainty Quantification

  • Reliability
  • Resilience
  • Sensitivities

Decision Making

  • Structural Retrofits
  • Expand generation capacity
  • Network redundancy

State-of-the-art is largely heuristic

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

3 of 12

System-level Uncertainty Quantification

State-of-the-practice techniques to enable computation

  • Multi-resolution modeling

  • Surrogate modeling

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

4 of 12

Verifying Heuristic Uncertainty Quantification

Example: Failure probability p

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

5 of 12

Verifying Heuristic Uncertainty Quantification

Heuristics in engineering

Heuristic

Solution Verification

Restoration Strategy

(Optimization)

Objective function

Failure Probability

(UQ)

Exhaustive simulation / enumeration

Polynomial-time evaluation

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

6 of 12

Verifying Heuristic Uncertainty Quantification

Heuristics in engineering

Heuristic

Solution Verification

Restoration Strategy

(Optimization)

Objective function

Failure Probability

(UQ)

Exhaustive simulation / enumeration

System Failure Probability

Failure modes A & B

NP-hard problem

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

7 of 12

System-level Uncertainty Quantification

Data and models

  • Hazard
  • Fragility
  • Restoration

Community goals

  • Robustness
  • Rapidity

Uncertainty Quantification

  • Reliability
  • Resilience
  • Sensitivities

Decision Making

  • Structural Retrofits
  • Expand generation capacity
  • Network redundancy

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

8 of 12

System-level Uncertainty Quantification

Data and models

  • Hazard
  • Fragility
  • Restoration

Community goals

  • Robustness
  • Rapidity

Uncertainty Quantification

  • Reliability
  • Resilience
  • Sensitivities

Decision Making

  • Structural Retrofits
  • Expand generation capacity
  • Network redundancy

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

9 of 12

Verified Uncertainty Quantification

Algorithms for system-level UQ today (mainly heuristic, e.g., IN-CORE)

  • : algorithm output
  • ε: tolerance (relative error)
  • δ: chance of exceeding error

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

10 of 12

Exact Uncertainty Quantification Possible?

How tree-like is real infrastructure?

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

11 of 12

Exact Uncertainty Quantification: Seaside

  • Probabilistic Earthquake
    • 100-year (figure)
    • 250-year
    • 1000-year

  • Power distribution
    • # Substations: 1
    • # Utility Poles: 321

Electricity Service Failure Probability

Paredes, R., Dueñas‐Osorio, L. & Hernandez‐Fajardo, I. Decomposition algorithms for system reliability estimation with applications to interdependent lifeline networks. Earthq. Eng. Struct. Dyn. 47, 2581–2600 (2018).

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/

12 of 12

Principled Sampling-based UQ: Seaside

Electricity Service Failure Probability

Paredes, R., Dueñas-Osorio, L., Meel, K. S. & Vardi, M. Y. Principled network reliability approximation: A counting-based approach. Reliab. Eng. Syst. Saf. 191, 106472 (2019).

Stopping-rule with (ε,δ)-approximation guarantees

NIST Center for Risk-Based Community Resilience Planning

roger.paredes@rice.edu | https://paredesroger.github.io/