Transforming Engineering and with powerful AI and Uncertainty Quantification Software

Upcoming Webinars

Discover

What We Do

SmartUQ is a modern AI software for engineering and physical sciences. Our Machine Learning, and Uncertainty Quantification tools are optimized for engineering applications including simulation, digital twins, testing, and manufacturing. With industry-leading model accuracy and user-friendly GUIs and APIs, it's possible to handle the toughest challenges and easily solve everyday problems.

Featured Customers

Why SmartUQ

SmartUQ's combination of unique sampling capabilities, powerful machine learning tools, and easy to use analytics help our customers solve previously unsolvable problems:

“SmartUQ has the best prediction accuracy among all tools I have ever used.”

– Technical Fellow at a Fortune 100 Aerospace company

“Our Uncertainty Quantification discipline now uses SmartUQ as its central tool and with it we have helped save millions of dollars and thousands of hours of work.”

– Statistician at a Fortune 100 Jet Engine company

“SmartUQ's adaptive design can significantly reduce the number of required simulations [a 72% reduction] and lead to much higher model accuracy [96% reduction in reference prediction error]”

– A Fortune 500 Semiconductor Company

Upcoming Events

    AI and Machine Learning for COMSOL Simulations Webinar Series Part 2: Fast, Accurate, Flexible Surrogate Models
    The success of surrogate modeling requires fast training speed and high prediction accuracy. Without speed, training a model can become infeasible as the scale and complexity of the problem increases. Without high accuracy a ML model’s predictions will have too much uncertainty to be usable. This 2nd of 4 webinars, will cover how SmartUQ addresses the need for speed and accuracy with its best in class ML models. Further discussed will be how SmartUQ’s accuracy and speed advantages are augmented by a flexible approach featuring many unique ML models, specifically designed to handle cases common to engineering simulation.

    AI and Machine Learning for COMSOL Simulations Webinar Series Part 3: Model Calibration
    COMSOL simulations and surrogate models of COMSOL simulations are of course only of benefit if the results they produce agree well with physical data, for example in the form of test or experimental data. This 3rd of 4 webinars will cover SmartUQ’s tools for model calibration including unique statistical calibration approaches which address the role that both parameter uncertainty and modeling assumptions play in the disagreement between simulation results and physical data.

    AI and Machine Learning for ANSYS Simulations Webinar Series Part 3: Model Calibration
    ANSYS simulations and surrogate models of ANSYS simulations are of course only of benefit if the results they produce agree well with physical data, for example in the form of test or experimental data. This 3rd of 4 webinars will cover SmartUQ’s tools for model calibration including unique statistical calibration approaches which address the role that both parameter uncertainty and modeling assumptions play in the disagreement between simulation results and physical data.

    AI and Machine Learning for COMSOL Simulations Webinar Series Part 4: Sensitivity Analysis
    In this last of 4 webinars the use of SmartUQ’s sensitivity analysis tools for gaining insights into COMSOL simulations will be discussed. Demonstrations of sensitivity analysis in SmartUQ will illustrate applications including assessing how robust the design or process being modeled is to uncertainty and which COMSOL model inputs are the greatest drivers of uncertainty in the outputs.

News

17
Sept

Resources

Discover how SmartUQ drives innovative solutions

Who are We?

SmartUQ offers Industry-Scale Uncertainty Quantification

Predictive Analytics

SmartUQ: Predictive Analytics for Engineering and Physical Sciences

What is Uncertainty Quantification?

Manage your risk with SmartUQ

Why Uncertainty Quantification?

Gain the competitive edge with your analytics

SmartUQ 10.1 Now Available!

Download Free Trial

Sample Use Cases

Fortune 100 Jet Engine OEM

Best In Class Gaussian Process Predictive Models for Jet Engine Design

Turbine engine

Challenge

Turbine engines, like other complex systems, are composed of many subsystems featuring a wide variety of physics and extreme behavior. From a simulation and analysis perspective, this means there are many input dimensions and the system suffers from the curse of dimensionality: i.e., it requires an exponential increase in sampling to cover the design space for the same level of resolution.

Solution

With existing tools, the Jet Engine OEM couldn't scale up their engine performance exploration and characterization efforts without an exponential increase in simulation resources. Particularly challenging was high fidelity CFD simulation of transient thermal events.

Results

SmartUQ developed faster and more efficient Design of Experiment and Emulation/ML modeling tools resulting in Best in Class Gaussian Process Modeling tools and several novel model types. These new tools made sampling and simulation requirements manageable while maintaining or improving model accuracy.

Learn More

Fortune 500 Heavy Duty Engine OEM

Combustion Model Calibration Project

Automotive engine

Challenge

The OEM was using traditional approaches to calibrate individual components and full engine models with large numbers of calibration parameters. This process takes a large number of simulation runs and a large amount of engineering time. Despite the effort, this process can result in poor model fit and doesn’t produce model form error information.

Solution

SmartUQ ran a proof of concept demonstrating advance statistical and Bayesian calibration techniques for a cylinder combustion model. This also involved the construction of a predictive model and construction of discrepancy maps.

Results

SmartUQ succeeded in generating accurate calibration parameters and discrepancy maps using a fraction of the simulation runs used with prior methods. This reduced the computation time substantially and allowed model form errors to be investigated. The success of this project lead to purchase and ongoing efforts towards full engine model calibration.

Learn More