Collaborative Research: Model-Based Multidisciplinary Dynamic Decisions in Design

Project: Research project

Description

The complexity of engineered systems such as combat aircraft has reached a tipping point that challenges existing design and analysis methods. Aerospace systems such as the F-22 and F-35, featuring highly integrated airframe components, electromechanical subsystems, avionics, and software, often exhibit “emergent behaviors” that can go undiscovered until late in development. Empirical and low-order physics-based design methods, applied in traditional paradigms and guided by expert opinion, are inadequate for capturing these complex behaviors; however, high-fidelity analyses and experimental tests cannot be applied in many situations in early design phases because adequate elaboration of the system geometry and composition is lacking. Consequently, a rigorous approach is needed for systematically fusing information from multiple sources and of multiple fidelities in hierarchical and/or multidisciplinary system models.
Building upon multidisciplinary expertise in computational mathematics, Bayesian statistical inference, multifidelity modeling, decision theory, multidisciplinary optimization, and aerospace engineering, this team proposes to create a mathematical framework that views design of a complex system as an information-seeking and knowledge-generation process that can be modeled as a stochastic discrete time dynamical system. In this paradigm, decisions are associated not only with the choice of attributes of the system but also the choice of subsequent actions in the design process, such as which source of information to use for the next analysis. These actions should be chosen to balance the needs of uncertainty reduction, optimization of system performance, and limits on the cost and duration of the design process.
Intellectual Merit: The novelty of our approach lies in the integration of fundamental principles in information theory and decision science. An overarching Bayesian information-integration framework is established to fuse heterogeneous information from multifidelity simulations, experiments, and expert opinion by extending spatial random process (SRP) modeling for multidisciplinary problems with complex data structures and high-dimensional physics-based reduced order models (ROMs). Approaches based on multidisciplinary global sensitivity analysis (MGSA) are also established for managing the couplings and complexity inherent in fusing information across fidelities, disciplines, and levels of a system model. The proposed dynamic decision making framework provides a uniform accounting for aleatory and epistemic uncertainties, and decision functions grounded in expected utility theory are formulated to guide subsequent design actions. By optimizing not only the system, but also the information seeking process, our research will revolutionize current approaches to decision making in engineering design.
Broader Impact: Although the framework is generic, we will demonstrate and validate our approach on two complex combat aircraft design testbeds that consider close couplings between propulsion system performance, thermal management, aerodynamics, and structural dynamics. In addition to multifidelity analysis models, we will also incorporate historical data and expert opinion, as well as experimental data via an “iron bird” propulsion/thermal test rig. These testbed problems will provide tools and insights that directly benefit DoD’s capability for designing next-generation military aircraft. Beyond aerospace engineering, our research will provide a new paradigm for understanding model/data complexity that will enr
StatusFinished
Effective start/end date9/1/158/31/19

Funding

  • National Science Foundation (CMMI-1537641   )

Fingerprint

Aerospace engineering
Testbeds
Propulsion
Physics
Decision making
Aircraft
Military aircraft
Engineering research
Airframes
Decision theory
Information theory
Structural dynamics
Avionics
Birds
Electric fuses
Random processes
Temperature control
Sensitivity analysis
Data structures
Large scale systems