In the previous virtual calibration blog, it was established that there are many benefits to incorporating virtual calibration in the development cycle. In this blog, I’ll go into detail about open-loop virtual calibration and explain how it is easily incorporated into existing development processes.
Open-loop virtual calibration is a method of using virtual models to populate calibration tables/maps, serving as a baseline for calibration development. In this method, the virtual model has no external controls. It also serves as a sensitivity analysis for the inputs that are being varied.
The unique benefits of this are:
- Calibration with the lowest barrier to entry
- Enables simulation assisted testing
- Provides insight into operating areas where controls won’t venture
- Understand sensitivity of input variables to output variables
Comparing Old and New Processes
Because open-loop calibration does not include external controls, it requires the user to vary many input variables in order to determine the output variables of interest from the model.
Imagine if a company develops a new diesel engine that will operate at 10,000 ft above sea level in freezing conditions. A calibration must also be developed to ensure that this engine will operate efficiently in the extreme environment.
The current method to develop this off-nominal calibration takes a significant amount of time in a very expensive test cell because of the extreme altitude and cold. This calibration also has to be developed sequentially at each operating point with risk of damage growing as the points approach full load.
Open-loop virtual calibration addresses these pain points and cuts down the time required for the performance team to develop a base calibration. Through parallelization of the virtual engine and aftertreatment model, multiple operating points are run at the same time. The results of the simulation are then used to develop a base calibration and to understand the limits of the powertrain.
The image below explains this alternative workflow, where an operating space is defined through an initial (or final) DOE space for input variables. The engine and aftertreatment model will be run local on 8 or 16 cores at once (most standard desktop computers) or sent to a computer cluster for massive parallel jobs. The outputs of that simulation will then be used to create a response surface model (meta-model) which will be used to better understand the engine and aftertreatment behavior. Therefore, calibration teams are able to develop a base calibration using the model results and reduce time in the test cell.
