| |
| |
Preface | |
| |
| |
Acknowledgements | |
| |
| |
Introduction and reading guide | |
| |
| |
Notation | |
| |
| |
Acronyms and abbreviations | |
| |
| |
| |
Applications and practices of modelling, risk and uncertainty | |
| |
| |
| |
Protection against natural risk | |
| |
| |
| |
The popular 'initiator/frequency approach' | |
| |
| |
| |
Recent developments towards an 'extended frequency approach' | |
| |
| |
| |
Engineering design, safety and structural reliability analysis (SRA) | |
| |
| |
| |
The domain of structural reliability | |
| |
| |
| |
Deterministic safety margins and partial safety factors | |
| |
| |
| |
Probabilistic structural reliability analysis | |
| |
| |
| |
Links and differences with natural risk studies | |
| |
| |
| |
Industrial safety, system reliability and probabilistic risk assessment (PRA) | |
| |
| |
| |
The context of systems analysis | |
| |
| |
| |
Links and differences with structural reliability analysis | |
| |
| |
| |
The case of elaborate PRA (multi-state, dynamic) | |
| |
| |
| |
Integrated probabilistic risk assessment (IPRA) | |
| |
| |
| |
Modelling under uncertainty in metrology, environmental/sanitary assessment and numerical analysis | |
| |
| |
| |
Uncertainty and sensitivity analysis (UASA) | |
| |
| |
| |
Specificities in metrology/industrial quality control | |
| |
| |
| |
Specificities in environmental/health impact assessment | |
| |
| |
| |
Numerical code qualification (NCQ), calibration and data assimilation | |
| |
| |
| |
Forecast and time-based modelling in weather, operations research, economics or finance | |
| |
| |
| |
Conclusion: The scope for generic modelling under risk and uncertainty | |
| |
| |
| |
Similar and dissimilar features in modelling, risk and uncertainty studies | |
| |
| |
| |
Limitations and challenges motivating a unified framework | |
| |
| |
References | |
| |
| |
| |
A generic modelling framework | |
| |
| |
| |
The system under uncertainty | |
| |
| |
| |
Decisional quantities and goals of modelling under risk and uncertainty | |
| |
| |
| |
The key concept of risk measure or quantity of interest | |
| |
| |
| |
Salient goals of risk/uncertainty studies and decision-making | |
| |
| |
| |
Modelling under uncertainty: Building separate system and uncertainty models | |
| |
| |
| |
The need to go beyond direct statistics | |
| |
| |
| |
Basic system models | |
| |
| |
| |
Building a direct uncertainty model on variable inputs | |
| |
| |
| |
Developing the underlying epistemic/aleatory structure | |
| |
| |
| |
Summary | |
| |
| |
| |
Modelling under uncertainty - the general case | |
| |
| |
| |
Phenomenological models under uncertainty and residual model error | |
| |
| |
| |
The model building process | |
| |
| |
| |
Combining system and uncertainty models into an integrated statistical estimation problem | |
| |
| |
| |
The combination of system and uncertainty models: A key information choice | |
| |
| |
| |
The predictive model combining system and uncertainty components | |
| |
| |
| |
Combining probabilistic and deterministic settings | |
| |
| |
| |
Preliminary comments about the interpretations of probabilistic uncertainty models | |
| |
| |
| |
Mixed deterministic-probabilistic contexts | |
| |
| |
| |
Computing an appropriate risk measure or quantity of interest and associated sensitivity indices | |
| |
| |
| |
Standard risk measures or q.i. (single-probabilistic) | |
| |
| |
| |
A fundamental case: The conditional expected utility | |
| |
| |
| |
Relationship between risk measures, uncertainty model and actions | |
| |
| |
| |
Double probabilistic risk measures | |
| |
| |
| |
The delicate issue of propagation/numerical uncertainty' | |
| |
| |
| |
Importance ranking and sensitivity analysis | |
| |
| |
| |
Summary: Main steps of the studies and later issues | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
A generic tutorial example: Natural risk in an industrial installation | |
| |
| |
| |
Phenomenology and motivation of the example | |
| |
| |
| |
The hydro component | |
| |
| |
| |
The system's reliability component | |
| |
| |
| |
The economic component | |
| |
| |
| |
Uncertain inputs, data and expertise available | |
| |
| |
| |
A short introduction to gradual illustrative modelling steps | |
| |
| |
| |
Step one: Natural risk standard statistics | |
| |
| |
| |
Step two: Mixing statistics and a QRA model | |
| |
| |
| |
Step three: Uncertainty treatment of a physical/engineering model (SRA) | |
| |
| |
| |
Step four: Mixing SRA and QRA | |
| |
| |
| |
Step five: Level-2 uncertainty study on mixed SRA-QRA model | |
| |
| |
| |
Step six: Calibration of the hydro component and updating of risk measure | |
| |
| |
| |
Step seven: Economic assessment and optimisation under risk and/or uncertainty | |
| |
| |
| |
Summary of the example | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Understanding natures of uncertainty, risk margins and time bases for probabilistic decision-making | |
| |
| |
| |
Natures of uncertainty: Theoretical debates and practical implementation | |
| |
| |
| |
Defining uncertainty - ambiguity about the reference | |
| |
| |
| |
Risk vs. uncertainty - an impractical distinction | |
| |
| |
| |
The aleatory/epistemic distinction and the issue of reducibility | |
| |
| |
| |
Variability or uncertainty - the need for careful system specification | |
| |
| |
| |
Other distinctions | |
| |
| |
| |
Understanding the impact on margins of deterministic vs. probabilistic formulations | |
| |
| |
| |
Understanding probabilistic averaging, dependence issues and deterministic maximisation and in the linear case | |
| |
| |
| |
Understanding safety factors and quantiles in the monotonous case | |
| |
| |
| |
Probability limitations, paradoxes of the maximal entropy principle | |
| |
| |
| |
Deterministic settings and interval computation � uses and limitations | |
| |
| |
| |
Conclusive comments on the use of probabilistic and deterministic risk measures | |
| |
| |
| |
Handling time-cumulated risk measures through frequencies and probabilities | |
| |
| |
| |
The underlying time basis of the state of the system | |
| |
| |
| |
Understanding frequency vs. probability | |
| |
| |
| |
Fundamental risk measures defined over a period of interest | |
| |
| |
| |
Handling a time process and associated simplifications | |
| |
| |
| |
Modelling rare events through extreme value theory | |
| |
| |
| |
Choosing an adequate risk measure - decision-theory aspects | |
| |
| |
| |
The salient goal involved | |
| |
| |
| |
Theoretical debate and interpretations about the risk measure when selecting between risky alternatives (or controlling compliance with a risk target) | |
| |
| |
| |
The choice of financial risk measures | |
| |
| |
| |
The challenges associated with using double-probabilistic or conditional probabilistic risk measures | |
| |
| |
| |
Summary recommendations | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Direct statistical estimation techniques | |
| |
| |
| |
The general issue | |
| |
| |
| |
Introducing estimation techniques on independent samples | |
| |
| |
| |
Estimation basics | |
| |
| |
| |
Goodness-of-fit and model selection techniques | |
| |
| |
| |
A non-parametric method: Kernel modelling | |
| |
| |
| |
Estimating physical variables in the flood example | |
| |
| |
| |
Discrete events and time-based statistical models (frequencies, reliability models, time series) | |
| |
| |
| |
Encoding phenomenological knowledge and physical constraints inside the choice of input distributions | |
| |
| |
| |
Modelling dependence | |
| |
| |
| |
Linear correlations | |
| |
| |
| |
Rank correlations | |
| |
| |
| |
Copula model | |
| |
| |
| |
Multi-dimensional non-parametric modelling | |
| |
| |
| |
Physical dependence modelling and concluding comments | |
| |
| |
| |
Controlling epistemic uncertainty through classical or Bayesian estimators | |
| |
| |
| |
Epistemic uncertainty in the classical approach | |
| |
| |
| |
Classical approach for Gaussian uncertainty models (small samples) | |
| |
| |
| |
Asymptotic covariance for large samples | |
| |
| |
| |
Bootstrap and resampling techniques | |
| |
| |
| |
Bayesian-physical settings (small samples with expert judgement) | |
| |
| |
| |
Understanding rare probabilities and extreme value statistical modelling | |
| |
| |
| |
The issue of extrapolating beyond data � advantages and limitations of the extreme value theory | |
| |
| |
| |
The significance of extremely low probabilities | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Combined model estimation through inverse techniques | |
| |
| |
| |
Introducing inverse techniques | |
| |
| |
| |
Handling calibration data | |
| |
| |
| |
Motivations for inverse modelling and associated literature | |
| |
| |
| |
Key distinctions between the algorithms: The representation of time and uncertainty | |
| |
| |
| |
One-dimensional introduction of the gradual inverse algorithms | |
| |
| |
| |
Direct least square calibration with two alternative interpretations | |
| |
| |
| |
Bayesian updating, identification and calibration | |
| |
| |
| |
An alternative identification model with intrinsic uncertainty | |
| |
| |
| |
Comparison of the algorithms | |
| |
| |
| |
Illustrations in the flood example | |
| |
| |
| |
The general structure of inverse algorithms: Residuals, identifiability, estimators, sensitivity and epistemic uncertainty | |
| |
| |
| |
The general estimation problem | |
| |
| |
| |
Relationship between observational data and predictive outputs for decision-making | |
| |
| |
| |
Common features to the distributions and estimation problems associated to the general structure | |
| |
| |
| |
Handling residuals and the issue of model uncertainty | |
| |
| |
| |
Additional comments on the model-building process | |
| |
| |
| |
Identifiability | |
| |
| |
| |
Importance factors and estimation accuracy | |
| |
| |
| |
Specificities for parameter identification, calibration or data assimilation algorithms | |
| |
| |
| |
The BLUE algorithm for linear Gaussian parameter identification | |
| |
| |
| |
An extension with unknown variance: Multidimensional model calibration | |
| |
| |
| |
Generalisations to non-linear calibration | |
| |
| |
| |
Bayesian multidimensional model updating | |
| |
| |
| |
Dynamic data assimilation | |
| |
| |
| |
Intrinsic variability identification | |
| |
| |
| |
A general formulation | |
| |
| |
| |
Linearised Gaussian case | |
| |
| |
| |
Non-linear Gaussian extensions | |
| |
| |
| |
Moment methods | |
| |
| |
| |
Recent algorithms and research fields | |
| |
| |
| |
Conclusion: The modelling process and open statistical and computing challenges | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Computational methods for risk and uncertainty propagation | |
| |
| |
| |
Classifying the risk measure computational issues | |
| |
| |
| |
Risk measures in relation to conditional and combined uncertainty distributions | |
| |
| |
| |
Expectation-based single probabilistic risk measures | |
| |
| |
| |
Simplified integration of sub-parts with discrete inputs | |
| |
| |
| |
Non-expectation based single probabilistic risk measures | |
| |
| |
| |
Other risk measures (double probabilistic, mixed deterministic-probabilistic) | |
| |
| |
| |
The generic Monte-Carlo simulation method and associated error control | |
| |
| |
| |
Undertaking Monte-Carlo simulation on a computer | |
| |
| |
| |
Dual interpretation and probabilistic properties of Monte-Carlo simulation | |
| |
| |
| |
Control of propagation uncertainty: Asymptotic results | |
| |
| |
| |
Control of propagation uncertainty: Robust results for quantiles (Wilks formula) | |
| |
| |
| |
Sampling double-probabilistic risk measures | |
| |
| |
| |
Sampling mixed deterministic-probabilistic measures | |
| |
| |
| |
Classical alternatives to direct Monte-Carlo sampling | |
| |
| |
| |
Overview of the computation alternatives to MCS | |
| |
| |
| |
Taylor approximation (linear or polynomial system models) | |
| |
| |
| |
Numerical integration | |
| |
| |
| |
Accelerated sampling (or variance reduction) | |
| |
| |
| |
Reliability methods (FORM-SORM and derived methods) | |
| |
| |
| |
Polynomial chaos and stochastic developments | |
| |
| |
| |
Response surface or meta-models | |
| |
| |
| |
Monotony, regularity and robust risk measure computation | |
| |
| |
| |
Simple examples of monotonous behaviours | |
| |
| |
| |
Direct consequences of monotony for computing the risk measure | |
| |
| |
| |
Robust computation of exceedance probability in the monotonous case | |
| |
| |
| |
Use of other forms of system model regularity | |
| |
| |
| |
Sensitivity analysis and importance ranking | |
| |
| |
| |
Elementary indices and importance measures and then-equivalence in linear system models | |
| |
| |
| |
Sobol sensitivity indices | |
| |
| |
| |
Specificities of Boolean input/output events � importance measures in risk assessment | |
| |
| |
| |
Concluding remarks and further research | |
| |
| |
| |
Numerical challenges, distributed computing and use of direct or adjoint differentiation of codes | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Optimising under uncertainty: Economics and computational challenges | |
| |
| |
| |
Getting the costs inside risk modelling - from engineering economics to financial modelling | |
| |
| |
| |
Moving to costs as output variables of interest � elementary engineering economics | |
| |
| |
| |
Costs of uncertainty and the value of information | |
| |
| |
| |
The expected utility approach for risk aversion | |
| |
| |
| |
Non-linear transformations | |
| |
| |
| |
Robust design and alternatives mixing cost expectation and variance inside the optimisation procedure | |
| |
| |
| |
The role of time - cash flows and associated risk measures | |
| |
| |
| |
Costs over a time period - the cash flow model | |
| |
| |
| |
The issue of discounting | |
| |
| |
| |
Valuing time flexibility of decision-making and stochastic optimisation | |
| |
| |
| |
Computational challenges associated to optimisation | |
| |
| |
| |
Static optimisation (utility-based) | |
| |
| |
| |
Stochastic dynamic programming | |
| |
| |
| |
Computation and robustness challenges | |
| |
| |
| |
The promise of high performance computing | |
| |
| |
| |
The computational load of risk and uncertainty modelling | |
| |
| |
| |
The potential of high-performance computing | |
| |
| |
Exercises | |
| |
| |
References | |
| |
| |
| |
Conclusion: Perspectives of modelling in the context of risk and uncertainty and further research | |
| |
| |
| |
Open scientific challenges | |
| |
| |
| |
Challenges involved by the dissemination of advanced modelling in the context of risk and uncertainty | |
| |
| |
References | |
| |
| |
| |
Annexes | |
| |
| |
| |
Annex 1 - refresher on probabilities and statistical modelling of uncertainty | |
| |
| |
| |
Modelling through a random variable | |
| |
| |
| |
The impact of data and the estimation uncertainty | |
| |
| |
| |
Continuous probabilistic distributions | |
| |
| |
| |
Dependence and stationarity | |
| |
| |
| |
Non-statistical approach of probabilistic modelling | |
| |
| |
| |
Annex 2 - comments about the probabilistic foundations of the uncertainty models | |
| |
| |
| |
The overall space of system states and the output space | |
| |
| |
| |
Correspondence to the Kaplan/Garrick risk analysis triplets | |
| |
| |
| |
The model and model input space | |
| |
| |
| |
Estimating the uncertainty model through direct data | |
| |
| |
| |
Model calibration and estimation through indirect data and inversion techniques | |
| |
| |
| |
Annex 3 - introductory reflections on the sources of macroscopic uncertainty | |
| |
| |
| |
Annex 4 - details about the pedagogical example | |
| |
| |
| |
Data samples | |
| |
| |
| |
Reference probabilistic model for the hydro component | |
| |
| |
| |
Systems reliability component - expert information on elementary failure probabilities | |
| |
| |
| |
Economic component - cost functions and probabilistic model | |
| |
| |
| |
Detailed results on various steps | |
| |
| |
| |
Annex 5 - detailed mathematical demonstrations | |
| |
| |
| |
Basic results about vector random variables and matrices | |
| |
| |
| |
Differentiation results and solutions of quadratic likelihood maximisation | |
| |
| |
| |
Proof of the Wilks formula | |
| |
| |
| |
Complements on the definition and chaining of monotony | |
| |
| |
| |
Proofs on level-2 quantiles of monotonous system models | |
| |
| |
| |
Proofs on the estimator of adaptive Monte-Carlo under monotony (section 7.4.3) | |
| |
| |
References | |
| |
| |
Epilogue | |
| |
| |
Index | |