The current state of the art in mechanical system design optimization under uncertainty is mainly focused on individual reliability requirements. System-level reliability requirements have not been widely considered, except with frame and truss structures. A newly developed direct decoupling strategy allows this consideration, and also facilitates a modular approach for including different methods for evaluating different reliability constraints. One particular issue here is the availability of sensitivity information, particularly when Monte Carlo methods are used. The applications of various decoupling approaches to realistic problems and comparisons of their relative performances are yet to be done. The extension of RBDO to robust design is quite recent, with several formulations being evaluated. The robustness requirement has been expressed either as single- or bi-objective optimization, and several approaches are being investigated. The inclusion of model uncertainty within RBDO is yet to be investigated. The extension of RBDO methods to multi-disciplinary systems has recently been accomplished through the decoupling strategies. This has also been shown to be feasible for systems where the different disciplinary codes are connected through feedback coupling. The direct decoupling approach allows the use of different reliability analysis methods (such as extended FORM, Markov Chain Monte Carlo simulation etc.) for such coupled systems. The application of RBDO to time-dependent problems has been indirect, with durability requirements expressed through time-independent measures. Direct inclusion of time-dependent reliability in RBDO, with random process treatment of loads and system properties, is yet to be done, and could be computationally prohibitive. Time-dependent problems are particularly relevant to fluid dynamics in aerospace vehicle design. Previous studies with progressive failure were limited to trusses, frames, and composite laminates, where computational effort was not prohibitive. However, this is a challenge in the case of problems where functional evaluations are very timeconsuming. A conservative first failure criterion has been used as a surrogate for ultimate system failure in the case of a composite aircraft wing, but requires further investigation. An important issue in practical application of RBDO is model uncertainty, especially considering the frequent use of response surface models in many RBDO studies. Ongoing research at Vanderbilt University is developing Bayesian methodologies for model validation, model uncertainty quantification, and inclusion of model uncertainty in design optimization. While RBDO methodology development marches ahead, with simple numerical examples as proofs of concept, practical implementation of RBDO still appears to be a distant goal, due to the lack of actual data on the random variables. Statistical data collection, which is expensive and time consuming, appears to be the single most difficult hurdle in the practical implementation of the RBDO methodology, and yet very important in developing confidence in the proposed methods.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Design optimization of reliability and robustness

    Mahadevan,S. / Vanderbilt Univ.,US | Kraftfahrwesen | 2004


    Design Optimization for Reliability and Robustness

    Mahadevan, Sankaran | SAE Technical Papers | 2004


    Design Optimization for Reliability and Robustness

    Mahadevan, S. / Society of Automotive Engineers | British Library Conference Proceedings | 2004


    Design Optimization for Reliability and Robustness

    Mahadevan, S. / Society of Automotive Engineers | British Library Conference Proceedings | 2004


    Optimization of a wheel profile accounting for design robustness

    Markine, V L / Shevtsov, I Y | SAGE Publications | 2011