From Monte Carlo Accuracy to Physical Validation in Pump Spindles

Abstract

What evidence is needed to connect repeated computational validation with component tests and field-relevant damage histories? This review answers by treating simulation-to-test validation as a property of a sociotechnical workflow rather than a feature that can be read from average accuracy. The focal setting is qualification of machine-learned pump-spindle fatigue estimates, where a fast surrogate can move a design decision far outside the domain where its training simulations were physically credible. Evidence from the assigned publications is synthesized with foundational studies of calibration, distribution shift, causal structure, and responsible deployment. Four requirements follow: preserve the lineage of parameterized geometries, finite-element and multiscale outputs, loading histories, material properties, and validation measurements; measure stability across relevant perturbations; connect confidence to a specific action; and maintain a route for human challenge and correction. The framework distinguishes descriptive performance from decision utility and separates uncertainty about the world from uncertainty created by the model and its evaluator. It also shows why faster inference or richer reasoning is valuable only when it improves a defined decision under a transparent resource budget. The article is a literature review and research agenda, not a report of a newly completed trial.