A Performance Based Method for Information Acquisition in Engineering Design under Multi-Parameter Uncertainty

Abstract Uncertainty pertaining to multiple parameters is a critical issue in designing complex systems. Whether or not to acquire more information to reduce uncertainty, and how to acquire information are the meta-level decisions to be made. Key challenges in making such decisions are that there are multiple information sources to choose from, and the cost of information as well as its effects on the overall design utility are different. To address these challenges, a performance-based stepwise information acquisition method is proposed. In the proposed method, the utility-based compromise Decision Support Problem construct is used to formulate design decisions to maximize the overall utility. For meta-level decisions, a performance index is developed for selecting the most appropriate information in each acquisition trial. The index is an integration of the improvement potential of the overall utility, the sensitivity of each ranged parameter, and the cost of the acquired information. Advantages of this proposed method are: 1) sensitivity-efficiency ensures that acquired information is invested on the critical parameters which avoids ineffective information acquisition; 2) cost-efficiency ensures that every acquisition is cost-efficient which avoids budget overruns. The efficacy of this method is demonstrated using the design of a hot rod rolling process. It is shown in the results that the performance-based method leads to an 8-45% larger drop of improvement potential compared to the random method.