Adaptive Decision-Making for Infrastructure Renewal Portfolios

Abstract

Infrastructure renewal is becoming a major challenge as networks age while expectations for safety, availability and service continuity continue to rise. In the Dutch context, this renewal wave pressures assets and portfolio resources, including budgets, engineering capacity, access windows, market capacity and preparation time. Although asset owners use inspections, monitoring, assessments, risk analyses, prognosis studies and portfolio information, these sources do not automatically show when an intervention direction should be reconsidered, when evidence is needed, or when preparation should begin.
This thesis examines how Dynamic Adaptive Policy Pathways can be operationalised as a decision architecture for infrastructure renewal portfolios. It develops a two-level architecture connecting asset-level evidence interpretation with portfolio-level steering and commitment. At asset level, monitored signposts are interpreted through staged trigger postures. Where uncertainty remains decision-relevant, evidence maturation can be initiated through inspection, recalculation, assessment or option development. Gate 1 reviews whether the current intervention state should be retained, deferred or updated. At portfolio level, the asset outcome is transferred through an asset-to-portfolio handover that makes urgency, uncertainty, preparation needs, dependencies and delivery constraints explicit. Assets can be prepared independently, coordinated, sequenced, held, bundled or reconfigured before Gate 2 tests whether the configuration is mature enough for commitment.
The architecture was applied to ten Rijkswaterstaat tand-nok bridges and six portfolio configurations. Tand-nok bridges are relevant because they involve hidden deterioration, restricted inspectability, structural-safety uncertainty, uneven evidence and long preparation periods. The application showed that single inspection findings, scores or structural concerns cannot directly determine intervention states. Signals must be interpreted in relation to vulnerability, persistence, evidence maturity, inspectability, structural-safety uncertainty and preparation time. At portfolio level, technical or geographical similarity alone was insufficient to justify execution bundling; assets may benefit from shared investigation or preparation while requiring different intervention routes or sequencing.
The product and process evaluation supported the architecture’s coherence and usability, while identifying the need for clearer guidance on de-escalation, handover, coordinated preparation, execution bundling, Gate 2 maturity and decision recording. The final architecture is an analytically applied and refined prototype. Its contribution is not to remove uncertainty or predict an exact intervention year, but to make uncertainty governable by clarifying what may remain uncertain, what requires evidence, when preparation should begin and when an intervention state or configuration is mature enough to change or commit.