Webinars

This webinar covers a statistical framework for imperfect data in engineering applications in a one-hour technical format including examples and a software demonstration. For a shorter, management style briefing a 30-minute version focused on the overall framework and its value without the detailed examples or demonstration is also being offered. Engineering decisions are rarely made with perfect information. Early in development, data may be sparse and the important parameters may not yet be known. Later, the cost of high-fidelity simulations limits the number of scenarios and parameters that can be investigated leading to imperfect exploration of the full input or design space. Models may also struggle to agree with physical data and even when an accurate model is developed, apparent optimums may not remain optimal in the presence of real world uncertainty. This webinar will discuss a statistical framework making use of techniques such as modern design of experiments (DOE), Gaussian process models, Adaptive DOEs, and statistical calibration to address imperfect data. These approaches can help quantify predictive uncertainty, determine where data can be most and least trusted, identify where additional data is most needed, determine which parameters are most influential, and reveal and correct the disagreements between models and physical data. The capabilities covered in this webinar have already been implemented in SmartUQ version 11.1. SmartUQ was originally invented for a Fortune 100 jet engine company and is now used by Fortune 500 companies across many industries.