Closed Loop Particle Forecasting Platform for Decision Support and System Prognostics.
T2019-029 A computational platform for forecasting the state of evolving systems and processes.
Particle methods are a class of computational algorithms used to estimate potential outcomes of a system or process. Although popular for their simplicity and scalability, the use of fixed sized "particle ensembles" renders the simulations unable to provide performance guarantees in quantifying system uncertainty. Thus, there is no way of knowing how accurate the generated forecast is. Since the state-of-the-art algorithm cannot guarantee achieving the desired level of accuracy, users often "over-compute" by using larger ensembles in the hopes that it works over the entire duration of forecasting. For moderately to highly complex systems, this is likely to be too computationally burdensome. Moreover, over-computing still does not provide any guarantees of accuracy.
This invention provides system forecasts with guaranteed performance while also providing the flexibility of using the smallest possible ensemble to achieve the user's desired accuracy. The particle ensemble is dynamic and adaptable, making the algorithm more computationally efficient than the state of the art. As a result, it delivers system forecasts with a guaranteed estimation accuracy of quantities of interest over the entire duration of the forecast, thereby enabling more robust decision making and system prognostics.
The forecasts derived by this platform are intended for use in a wide range of industries, such as:
- prognostics for electric car batteries & aerospace propulsion;
- decision making support for wind farms & structural design; and
- reliability assessments in chemical and nuclear processing centers
- Improved computational efficiency
- More accurate system forecasting
- Robust decision-making capabilities for the user