Improving Remaining-Fatigue-Life Evaluation Using Data Interpretation
This paper presents a methodology that improves fatigue-performance evaluations using model-based data interpretation. The accuracy of stress-range values is essential for quantifying fatigue damage. These values are usually predicted using physics-based models such as those used within finite element analyses. In the modelling process, simplifications are inevitable, thus causing systematic errors in model predictions. Structural health monitoring coupled with model-based data-interpretation approaches have the potential to reduce uncertainties associated with the evaluation of stress-range predictions. Because of the presence of modelling and measurement uncertainties, many models may explain the true structural behaviour. A model falsification approach, which is able to cope with incomplete knowledge of uncertainties, is used to isolate candidate models from an initial population of models. This approach is robust for systematic errors that are correlated spatially. The candidate models that are identified using the model-falsification approach predict stress ranges in structural members, from which the remaining fatigue life is determined. Due to the uncertainty reduction in model predictions during data interpretation, the accuracy of the fatigue prognosis is improved. A steel beam composed of a circular hollow-section truss is studied for illustration. Monitoring data that is interpreted using a model-falsification methodology shows potential for improving evaluations of remaining fatigue life.