Real Life Scenarios and Multi-Objective Optimization with xLINK

xLINK’s mature simulation environment includes all GT-SUITE’s powerful features for parametric analysis and optimization. In the previous xLINK blog, we created a full vehicle model by assembling three external subsystem models (Drivetrain and Engine as two FMUs and a soft ECU as a Simulink DLL). xLINK identified the expected inputs, outputs and all exposed parameters of each module and we saw how you can easily connect the exchanged signals. Now let’s try to test xLINK capabilities by promoting the exposed parameters to Case Setup, then group them to replicate real-life scenarios and perform a challenging optimization.

Alaska or Death Valley, how will your vehicle model behave?

How sure can I be that my xLINK vehicle model will not die in Death Valley or freeze in Alaska? These are two extreme real-life climate scenarios that can be easily modeled with xLINK’s Case Setup feature. Altitude, humidity, and ambient temperature and pressure are exposed as FMU parameters of the Engine FMU. Through the right click option, I promote them to Case Setup and set them accordingly:

Keep in mind that the drivetrain model will also operate in the same climate and as a result I replace the default FMU parameter values with the Case Setup parameters I defined previously:

Revisiting the drivetrain FMU, I notice that vehicle mass, tire radius, frontal area and drag coefficient are also exposed as FMU parameters and I will use them to create a group of parameters that will stand for the vehicle type, e.g. an SUV and a Hatchback. In xLINK, several parameters can be grouped together in the Case Setup dialog box to form a Super Parameter. Basically, a Super Parameter is a set of parameters in the form of a drop-down list.

First, it is created through the dedicated toolbar button in Case Setup and then it is populated by moving defined parameters from Main Tab to the Super Parameter Tab. This way Environment Conditions and Vehicle Type groups are defined as Super Parameters and all options are named and populated respectively in order to create the extreme real-life scenarios, as shown below:

The above parameter settings stand for 4 different simulation cases. Now, let’s see how the two vehicles behave in the two extreme climate scenarios in terms of engine power output, while the soft ECU drives them through NEDC. The integrated vehicle model behavior is monitored throughout the 4 different scenarios by using xLINK Monitor templates and in the end of the simulation the results are available for further analysis in GT-POST. By using the Combine Cases option, I create comparative analysis plots and assess the behavior of the model at each simulation case:

From the very first 25s of NEDC, I notice that the SUV demands more Brake Power from the Engine than the Hatchback, while the latter obviously responds faster to fluctuations in desired vehicle speed and engine load. If we zoom into the red box above, the aforementioned observations are straightforward; what is still not easy to understand is the model’s behavior difference between the two different Environment Conditions. However, GT-POST offers the ability to efficiently compare the two different environmental conditions for each Vehicle type by applying the percentage difference function:

So far, we have seen that xLINK empowers users with the ability to efficiently create and evaluate multiple scenarios and variants using advanced parameterization features. At the same time, xLINK simulation results are available in GT-POST, which offers the necessary post-processing functionality to highlight the interactions between the different subsystem models and assess their behavior in different real-life scenarios. It seems that GT’s multi-year experience in system simulation has made xLINK a powerful, all-inclusive solution for system integration.

ECO-SPORT mode: Gear shifting strategy multi-objective optimization with xLINK

Earlier we studied how two vehicle types behave when driving through NEDC in extreme climate conditions. At this point, I would like to test xLINK using the Integrated Design Optimizer (IDO) and perform a multi-objective optimization varying some of the external models exposed parameters. A flat acceleration 0-60 mph (0-100 kmh) is by far one of the most exciting situations while driving a vehicle. Some drivers prefer vehicles which need the shortest amount of time for such an acceleration, while others opt for minimum fuel consumption. But, why not go for both? xLINK can perform such an optimization by varying the target vehicle speed before shifting upwards to a next gear. In other words, I will use the exposed Vehicle parameters of the Drivetrain FMU, in order to optimize the gearbox’s up-shifting strategy for both minimizing the 0-60 time and minimizing cumulative fuel consumption and consequently search for a 2-D Pareto Front of optimal solutions.

Reading the Optimizer’s manual, I understand that the most suitable available algorithm for such an Optimization is the Genetic algorithm, which I set appropriately inside Design Optimizer dialog as shown below:

xLINK Figure 6 v3

Genetic Algorithm population size is set to 34 and number of generations to 33, generating a total of 1122 Iterations. In the lower right corner of the above screenshot, you’ll find optimization objectives, and above that are the optimization factors. Specifically, the chosen factors are the target vehicle speed to up shift from 1st to 2nd gear and all vehicle speed differences going from 2nd to 3rd and so on and so forth. In addition, I set the minimum speed difference before shifting from one gear to another to 10 km/h and the maximum 50km/h by limiting the optimization factors between Lower and Upper Limits as shown above. When IDO is enabled, the optimization run starts by simply hitting the Run button and the IDO User interface opens automatically:

The IDO interface is preloaded with interactive plots that display the values of Factors and Objectives during the simulation. Users can multi-select specific designs on the in-progress 2-D Pareto plot and all related values will be highlighted in the Solutions Table while in all other plots the related plot points will be marked inside red boxes. Furthermore, users can export one or more selected designs from the Solution table and perform further standalone tests in a new xLINK model, which is created automatically. Convergence progress is also shown in a separate plot. Once xLINK IDO finds all optimal solutions that satisfy both objectives, it generates the following 2-D Pareto Front:

xLINK Figure 8 v2

Further analysis of the optimal solutions can be performed by exporting all Pareto front solutions to new xLINK models in order to choose the final one, the Eco-Sport mode. It will not be a problem for xLINK to work further with the IDO and search for other driving modes, e.g. Sport or Eco, by performing two new single objective optimizations, but this is where I close the second xLINK blog. From where I stand, xLINK has been proven to be a complete solution for system integration, testing and optimization. Can’t wait to try xLINK? Would you like to test any application with xLINK or the Integrated Design Optimizer? Contact us!

Written by Pantelis Dimitrakopoulos