SmartUQ is a powerful Machine Learning (ML) software tool optimally designed for science and engineering applications. By providing powerful tools and highly accurate ML models with user-friendly GUIs and APIs, SmartUQ makes it easy to perform predictive modeling, optimized sampling, uncertainty quantification, and model calibration. From Fortune 500 manufacturers to startups and engineering consulting firms, SmartUQ's best in class predictive modeling accuracy helps our customers go beyond analysis to bring uncertainty into the decision-making process.
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.
Industries Served: Automotive, Aerospace & Defense, Turbomachinery, Heavy Equipment, Medical Device, Semiconductors, Energy, Oil & Gas, HVAC
Tools and Application Areas:
-Acceleration of simulation efforts, Uncertainty Analysis, Testing and evaluation planning;
-Optimization under uncertainty, Robust design, Model calibartion and validation;
-Embedded models, virtual sensors, Root cause analysis, Manufacturing analytics;
-Digital twin analytics, Predictive Maintenance, Quality Control, Process Optimization.
DOEs are typically used to collect new data from a system. In many cases, sufficient data has already been collected. Often in these scenarios, the data collected has been accumulated over long periods of time, and there is enough data that analysis is simply intractable. For example, health monitoring data from sensors on fielded components may capture live data continuously over the entire operating life of the component. SmartUQ’s data sampling tools can divide the data to mimic a space-filling DOE consisting of subsets of the full data set. Unlike DOEs which are developed before data collection, data sampling like subsampling and sliced sampling takes existing input-output data pairs and selects the points that will represent the design space well.
Designs of Experiments
SmartUQ provides a number of breakthrough data sampling techniques and a comprehensive library of advanced DOE generators for both simulation and physical experiments. Invented by thinking outside the box, our technologies ensure accuracy and minimize the number of data points required to generate uncertainty quantification and analytics results. Several of our more popular tools include subsampling for Big Data applications and Adaptive Design, which maximizes sampling efficiency by using already gathered data to select additional data points.
Game-changing emulation technology allows SmartUQ to fit accurate emulators in record-setting time. These extremely fast analytical models can predict the behavior of complex black-box computational and physical systems. Using emulators enables extremely fast uncertainty propagation, sensitivity analysis, design space exploration, statistical optimization, statistical calibration, and inverse analysis. No more expensive Monte Carlo sampling and no more waiting hours for analytics calculations.
SmartUQ’s technology can handle categorical and continuous inputs, systems with multiple and functional outputs, high dimensional systems, and big data, opening new doors for accelerating uncertainty quantification and analytics.
Simulation accuracy continues to improve but it is still necessary to ground simulations with test data to ensure that they accurately represent the real world. Our statistical calibration tool quickly and automatically determines model calibration parameters given limited simulation and test data. It also provides model discrepancy measurements to help identify opportunities for improvements and to provide metrics for model validation. By increasing model accuracy and accelerating model validation, statistical calibration can decrease the time and number of tests required to understand complex systems, shortening the design cycle.
Rapidly determine the sensitivity of outputs with respect to inputs across the entire design space. This is useful when determining sensitivity of part geometries, instrumentation accuracy, and regulatory compliance with respect to manufacturing tolerances, environmental conditions, and wear levels. Sensitivity analysis shows which factors have a relatively low or high impact, allowing engineers to focus design effort and resources where they are needed most.
Propagation of uncertainty lets users predict the probability distributions of system outputs resulting from distributions of uncertain or variable system inputs. Almost all systems have some input uncertainty usually from inputs like physical measurements, manufactured dimensions, material properties, environmental condition, and applied forces. Propagation of uncertainty helps engineers determine whether the system outputs will meet requirements, what the extreme probabilities really are, and which inputs have the most effect on the output distributions. All this means better initial designs, faster development, and simplified trouble shooting.
SmartUQ can be used to conduct statistical optimization. This novel approach combines adaptive sampling techniques and analytical models providing improved performance on complex problems relative to search based methods. Statistical optimization also allows very rapid search area reduction with multiple objectives and very large numbers of input parameters. Even better, the required system evaluations may be determined using adaptive design, recycled from earlier data sets, or run in parallel batches for large clock-time savings and shortened testing cycles.
In general, SmartUQ is very different in both purpose and capabilities from basic statistics packages.
SmartUQ is a predictive analytics engine focused on building an accurate prediction surface for all what-if input scenarios and quantifying various uncertainties in simulation and test. Once you have built with SmartUQ an accurate high-speed predictive model or emulator, it can be used to perform sensitivity analysis, uncertainty analysis, exploration of the entire design space, and optimization.
For uncertainty quantification, SmartUQ can significantly outperform Monte Carlo methods with the same sample size. SmartUQ also features statistical and Bayesian calibration which can significantly improve the accuracy of a simulation model using a small set of physical test points.
SmartUQ was invented to solve complex UQ problems. Designed from the ground up to enable practicing engineers to make use of advanced statistics, SmartUQ has a user-friendly interface and a powerful Python API for automation and integration with other tools.
SmartUQ has also focused development on advanced tools leading to a set of unique features including:
(1) The ability to handle UQ problems with multivariate, multi-fidelity, functional or spatial outputs, or with categorical and continuous inputs;
(2) Advanced Design of Experiments tools including several methods of adaptive design which choose new sampling points based on previously sampled design points;
(3) A variety of choices for statistical and Bayesian calibration;
(4) Technology for using scanned 3d surfaces as inputs for other UQ tools; and,
(5) The ability to handle problems with many more dimensions, larger sample sizes, and more complex structure than any previous tools. This is a crucial part of the success with customers in large engineering companies. None of these capabilities appear in any basic statistics packages.
Are you convinced, or do you wish to convince yourself? You can request a evaluation license at firstname.lastname@example.org. Do you have technical or other in-depth questions related to SmartUQ, you can contact email@example.com.
You are invited to our free webinar:
Artificial Intelligence and Machine Learning for Automotive Applications
December 7, 2022 2:00 PM ET
Improving automotive design and manufacturing processes requires understanding and accounting for uncertainties. For example, there will be uncertainty in the properties of the materials used and manufacturing process for any component. Even for a perfect process that produced identical components, the performance of each will vary depending on uncertainties associated with its use. For example, the fatigue life of a component could vary based on the vehicle model it is installed in and road conditions.
Determining optimal design configurations or manufacturing processes under such uncertainties is difficult and can require substantial time using test data, experiments, and physics-based simulations (e.g. CFD and FEA). Also, it is time consuming to sort through large amounts of manufacturing data to identify the most useful and relevant information.
The solution is to first train an AI or machine learning model using data from the design or manufacturing process collected by an intelligent sampling plan. Once trained, the model can rapidly make accurate predictions for all what-if scenarios. With the roadblock of computational cost removed, many otherwise infeasible analyses may be conducted to improve the design or process.
Join us for this webinar to learn how AI and machine learning models can be used to enhance automotive design and manufacturing applications.
Smit Consult offers fast delivery thanks to direct contact with software manufacturers.
We deliver exceptional costumer service and support based on technical knowledge.
Consultancy & Courses
Smit Consult is your consultant and delivers various training courses in statistics.