We invite you to stop by booth #7 at NAFEMS CAASE 2018; meet experts in engineering analytics and uncertainty quantification, see demonstrations, and explore how SmartUQ can improve your analysis.
June 5 - 4:35 PM to 6:20 PM EDT - Room: 21
Presented by Dr. Mark Andrew, SmartUQ Technology Steward
Experienced practitioners who construct complex simulation models of critical systems know that replicating real-world performance is challenging due to uncertainties in found in simulation and physical tests. This course will discuss the types of uncertainties in the context of representing them with a design of experiments, constructing surrogate models and finally applying analytical methods to understand how sources of uncertainty impact replicating reality. This course will discuss the broad applications these probabilistic techniques have in analyzing numerous forms of engineering systems including Digital Thread/Digital Twins.
June 6 - 10:50 AM to 12:35 PM EDT - Room: 22
Presented by Dr. Mark Andrew, SmartUQ Technology Steward
Analyses of complex data streams using basic Uncertainty Quantification (UQ) methods do not always yield the results sought. The emergence of complex data from modern computer simulations has spawned new UQ methods for extracting meaningful information. Complex simulations can be time-consuming to run, have a temporal, functional, or transient response, contain calibration parameters, or involve inputs of aleatoric and epistemic uncertainties. This course discusses the challenges of complex data from such simulations and emerging technologies like Digital Thread – Digital Twin, the new UQ methods, and the results they yield.
June 6 - 4:05 PM to 6:00 PM EDT - Room: 22
Presented by Dr. Mark Andrew, SmartUQ Technology Steward
As simulation models increase in complexity, so does our inability to understand the impact uncertainties have on the simulation results. It is possible for simulations to find better agreement with reality by calibrating the simulation model to physical data. Statistical calibration tunes the model to be as realistic as possible and quantifies the uncertainty in the simulation model. This tutorial will use a catapult to sequentially walk through the model calibration process and determine the model form uncertainty.
June 6 - 1:50 PM to 2:25 PM EDT - Room: 24
Presented by Kevin O'Flaherty, SmartUQ Application Engineer
This presentation illustrates both conceptual and practical applications of using Uncertainty Quantification (UQ) techniques to perform probabilistic analyses. The application of UQ techniques to the output from engineering analyses using model-based approaches is essential to providing critical decision-quality information at key decision points in a aerospace system’s life cycle. Approaches will be presented for the continued collection and application of UQ knowledge over each stage of a generalized life cycle framework covering system design, manufacture, and sustainment. The use of this approach allows engineers to quantify and reduce uncertainties systematically and provides decision makers with probabilistic assessments of performance, risk, and costs which are essential to critical decisions. As an illustration, a series of probabilistic analyses performed as part of the initial design of a turbine blade will be used to demonstrate the utility of UQ in identifying program risks and improving design quality. The application of UQ concepts to life cycle management will be addressed, highlighting the benefits to decision makers of having actionable engineering information throughout a system’s life cycle.
June 7 - 10:20 AM to 10:55 PM EDT - Room: 24
Presented by Kevin O'Flaherty, SmartUQ Application Engineer
Polynomial Chaos Expansion (gPCE) was developed using applied mathematics and does not require knowledge of a simulation’s physics. Thus, gPCE may be used across disparate industries and is applicable to both individual component and system level simulations. A shortcoming of gPCE is that the assumption that simulation results can be collected precisely on a set of carefully chosen input configurations. Because of this assumption, gPCE encounters problems when any of the input configurations fail to produce valid simulation results. We propose a statistical framework to mitigate the cost of missing values. This framework decreases the additional simulation runs necessary to yield accurate UQ results in the event that simulation failure makes gPCE methods unreliable. Several examples are used to demonstrate the proposed frame work and its utility including a simple beam deflection model and several non-linear test functions.
June 7 - 3:40 PM to 4:15 PM EDT - Room: 24
Presented by Dr. Mark Andrew, SmartUQ Technology Steward
Statistical calibration can reduce design cycle time and costs by ensuring that simulations are as close to reality as possible and by quantifying how close that really is. It optimizes tuning of model parameters to improve simulation accuracy, and estimates any remaining discrepancy which is useful for model diagnosis and validation. Because model discrepancy is assumed to exist in this framework, it enables robust calibration even for inaccurate models. The presentation will introduce the concepts and advantages of statistical calibration and model validation. Using an air foil CFD model case study, this presentation will walk through the step‐by‐step process of performing statistical calibration and quantification of the uncertainty of the final calibrated model. An analysis will be performed to compare results from a calibrated model and an uncalibrated model. The analysis will include optimization and sensitivity analysis. The results will illustrate the importance of calibrating a model before drawing design conclusions.
June 7 - 1:35 PM to 2:10 PM EDT - Room: 26B
Presented by Kevin O'Flaherty, SmartUQ Application Engineer
The process and results of using engineering analytics methods will be presented for three unique automotive applications in order to demonstrate the utility of performing advanced analytical techniques in a variety of scenarios. The engineering analytics solutions for these applications have been successfully tested on real industry challenges. The applications to be demonstrated are:
