Theory¶
Operational Resilience Engineering is organized around three linked but distinct analytical domains:
Vulnerability → Resilience → Control
The sequence is analytical, not chronological. Vulnerability characterizes susceptibility to undesirable change and the perspectives used to evaluate it. Resilience characterizes capacities and design choices that preserve, extend, recover, or improve acceptable performance under disturbance. Control characterizes the processes through which a system senses, anticipates, acts, and learns as conditions change.
The present vulnerability formulation is developed in Towards a Theory of Quantitative Vulnerability Analysis for Engineering Systems, available as an arXiv preprint [EA26].
Vulnerability¶
ORE defines the vulnerability of an engineering system in terms of its susceptibility to undesirable change. A quantitative vulnerability analysis begins with a vulnerability space:
where:
\(S\) is a set of scenarios or system states describing what can go wrong;
\(L : S \rightarrow [0,1]\) represents information about the likelihood of each scenario; and
\(C : S \rightarrow \mathbb{R}_{\geq 0}\) represents information about the adverse consequence associated with each scenario.
The tuple \(\mathcal{V} = (S,L,C)\) is not itself a risk measure. It contains vulnerability data that must be interpreted before it can support a quantitative evaluation or an engineering decision.
Data, perspectives, and decisions¶
The central claim of the framework is:
Data → Perspective → Decisions
An interpretation maps likelihood and consequence data into a mathematical form that supports a particular class of vulnerability measures. A vulnerability perspective is a specific combination of likelihood and consequence interpretations. Perspective is therefore not merely a verbal description of an analyst’s viewpoint; it has formal mathematical content.
Taking a perspective is necessary for quantitative decision-making, but it also narrows the information brought forward for evaluation. No single perspective is expected to reveal every relevant vulnerability of a complex engineering system.
Four vulnerability perspectives¶
The current theory identifies four foundational perspectives on a vulnerability space. They differ according to whether likelihood is interpreted as probability or possibility, and whether consequence is interpreted as a numeric value or as membership in a category or performance set.
Likelihood interpretation |
Consequence interpretation |
Perspective |
Representative measures |
|---|---|---|---|
Probability \(P(S)\) |
Numeric or scaled value \(G(S)\) |
Risk \(\mathrm{Risk} = (P,G)\) |
Expected loss; expected consequence; value-at-risk; conditional value-at-risk; expected unserved energy |
Probability \(P(S)\) |
Set-value or category \(\Gamma(S)\) |
Reliability \(\mathrm{Reliability} = (P,\Gamma)\) |
Probability of failure; availability; mean time to failure; loss-of-load probability; interruption indices |
Possibility \(\Pi(S)\) |
Numeric or scaled value \(G(S)\) |
Adversary \(\mathrm{Adversary} = (\Pi,G)\) |
Network interdiction; maximum load shed; worst-case contingency performance; failure-budget analysis |
Possibility \(\Pi(S)\) |
Set-value or category \(\Gamma(S)\) |
Safety \(\mathrm{Safety} = (\Pi,\Gamma)\) |
Security criteria; reachability analysis; reserve margins; minimum contingency budget to violation |
Here, \(P\) maps likelihood information to a probability distribution and \(\Pi\) maps it to a possibility distribution. \(G\) maps consequence information to a scalar value, whereas \(\Gamma\) maps it to a category, group, or performance classification.
For a binary operational boundary \(\tau\), a consequence interpretation can be written:
This distinction matters because each perspective retains, aggregates, and orders vulnerability data differently. Therefore, different perspectives can favor different engineering designs even when they are evaluated using the same scenarios and the same underlying system data.
Risk is not reliability¶
Risk and reliability are not interchangeable measures of the same system property. In the current formulation, risk applies probability to a scalar-valued consequence interpretation:
Reliability applies probability to a categorical or thresholded consequence interpretation:
For example, a common expected-consequence risk measure is:
while a threshold-based unreliability measure is:
where \(F(\tau) = \{s \in S : C(s) \geq \tau\}\) is the set of non-operational states.
The first measure preserves relative consequence magnitudes. The second groups all states beyond an operational boundary into a common category. Consequently, risk and reliability can rank vulnerabilities differently and can recommend different designs. Neither should generally be assumed to be a special case, substitute, or synonym for the other.
Adversary and safety perspectives¶
Probability is not the only coherent interpretation of likelihood information. A possibility interpretation changes how likelihoods are normalized, combined, and compared. Possibility-based perspectives are especially relevant when the task concerns plausible worst cases, adversarial action, security criteria, or reachability of unacceptable states.
The adversary perspective is:
and emphasizes a scalar-valued consequence among plausible scenarios. A simple worst-case measure is:
The safety perspective is:
and combines possibility-based likelihood with a categorical interpretation of whether system performance remains within an acceptable boundary. A representative safety measure is:
The theory therefore distinguishes risk, reliability, adversary, and safety as co-constituted perspectives on a common vulnerability space. It does not treat one as the parent category of the others.
Implications for decisions¶
A vulnerability measure does more than assign a numerical score. It determines which features of system behavior are prioritized in a design decision.
For example, a risk perspective may favor an intervention that reduces a rare but catastrophic consequence. A reliability perspective may instead favor an intervention that reduces the more frequent crossing of an operational threshold. An adversary perspective can prioritize the most consequential plausible attack or disruption, while a safety perspective can prioritize the possibility of reaching an unacceptable state.
A design that improves one perspective need not improve another. When this occurs, the problem is not resolved simply by treating all measures as synonyms, nor necessarily by combining them into a single aggregate score. Rather, the decision-maker must identify and justify the tradeoffs among distinct vulnerability perspectives.
Resilience¶
ORE defines resilience as the ability of a system to manage undesirable change. Resilient design concerns the selection of engineering designs, actions, policies, or configurations intended to preserve or improve that ability.
Vulnerability and resilience are distinct but mutually constitutive concepts. Vulnerability analysis evaluates susceptibility to undesirable change. Resilient design evaluates what can be changed in the engineering system and what those changes are intended to achieve.
Design perspectives¶
The developing framework distinguishes four perspectives on resilient design:
Robustness concerns maintaining intended operation by making operational states more stable.
Rebound concerns recovery by making non-operational states less persistent.
Extensibility concerns changing available capacity or operational boundaries so that a system can continue providing function under stressed or disrupted conditions.
Evolvability concerns functional redesign and new capabilities that improve future operation, including opportunities arising during non-operation, repair, or reconfiguration.
These perspectives distinguish changes in state stability from changes in operational boundaries and system function. They also make the interpretation of system performance explicit: identifying a system as operational or non-operational is part of the analysis, not simply an assumption that every performance measure resolves automatically.
Robustness and rebound remain important engineering design perspectives. Extensibility and evolvability address additional ways of managing undesirable change that cannot be captured adequately by resistance and recovery alone.
Relationship to established resilience research¶
The framework is informed by Woods’s four concepts of resilience: robustness, rebound, graceful extensibility, and sustained adaptability [Woo15, Woo18].
ORE is a developing formal interpretation rather than a reproduction of Woods’s terminology or theory. Its objective is to connect established resilience concepts to explicit engineering design questions, including what performance is desirable, what decisions are feasible, and what kinds of system change those decisions produce.
The work is also motivated by limitations of using rebound or system-performance curves as stand-alone models for engineering decisions [DEP26, ESA25]. Describing performance loss and recovery is not equivalent to specifying which design intervention should be selected.
Relationship to vulnerability analysis¶
A vulnerability perspective and a resilience design perspective answer different questions. The former determines how susceptibility to undesirable change is evaluated; the latter determines what kinds of engineering change are being considered.
ORE therefore does not present risk and resilience as competing alternatives. Vulnerability measures can inform resilient design without making risk, reliability, adversary, or safety synonymous with robustness, rebound, extensibility, or evolvability.
Development status¶
A companion manuscript on quantitative resilient design is in preparation. This page provides a conceptual introduction only. Formal definitions, decision models, worked examples, and research findings will be released with the public preprint.
Control¶
Control concerns the continuing processes that relate observed conditions and vulnerability information to resilient action. In ORE, control follows the identification of relevant vulnerabilities and the specification of resilience capacities.
The developing control framework will connect vulnerability and resilience to four resilience-engineering processes:
Sensing: detecting system performance, disturbances, constraints, and conditions near operational boundaries.
Anticipating: identifying possible future states, demands, threats, opportunities, and boundary challenges.
Acting: selecting and implementing feasible interventions, designs, or operational strategies.
Learning: revising models, operational boundaries, strategies, and intervention choices from both successful and unsuccessful performance.
The control formulation will ultimately specify how sensing and interpretation identify vulnerability; how anticipation and action invoke resilience capacities; and how learning changes the future decision model.
Sensing, anticipating, acting, and learning¶
The developing ORE control formulation treats resilient performance as a continuing process of sensing conditions, anticipating future challenges, acting through feasible interventions, and learning from the consequences of action.
This process-oriented account builds on resilience engineering work that links human, organizational, and technical capacities in national infrastructure resilience [TESF19]. It also responds to the central challenge of surprise in critical infrastructure: the objective is not to assume that all disruptive conditions can be predicted or eliminated, but to prepare systems and their operators to recognize, respond to, and learn from conditions that exceed existing expectations [ADES22, SAP+25].
The vocabulary of resilience remains highly context-dependent across critical infrastructure research. ORE therefore makes its use of vulnerability, resilience, state stability, operational boundary, and control processes explicit rather than assuming that these terms are interchangeable [MHS+23].
The SAAL processes do not presume that surprise can be eliminated. Instead, they organize how an engineering system and its operators can detect changing conditions, anticipate plausible futures, select and implement feasible actions, and revise the system’s future decision model through learning [ADES22, SAP+25].
Research status¶
The vulnerability space theory and its four perspectives are described in Towards a Theory of Quantitative Vulnerability Analysis for Engineering Systems, available as arXiv:2610.02424 [EA26]. The paper is a public preprint; future peer-review and publication status will be recorded on Artifacts.
The resilience and control sections state the intended architecture of Operational Resilience Engineering. Their formal notation, propositions, decision models, and engineering examples remain under active development and will be released as versioned technical notes, preprints, and publications.