Learn about the trend of hybrid simulation.
Hexagon has submitted this post.
Written by: Shawn You, Shawn Gao, Michael Dembinski (MTS Systems), Hemanth Kolera (Hexagon Manufacturing Intelligence)
There is a trend toward convergence of modelling and simulation. Before the test, engineers use virtual methods to guide the test design, configuration, component sizing and selection. Physical and virtual methods are complementary. Virtual testing methods can be used as a rehearsal tool for physical testing, and physical test data can be used to validate the virtual models. In some instances where the complete physical prototype is not available for testing, the missing substructures are replaced by models allowing a combination of modelled and unmodelled substructures to be tested.
This is called the hybrid simulation approach. The availability of mature modelling methods, data transfer methods and computing power also allow models and tests to be run in parallel when there is substantial complexity in this hybrid testing method. The model can then provide compensation signals to the test in real-time to be more accurate.
At MTS, a suite of different approaches are utilized to support Model-Based System Engineering approaches, including virtual testing, hybrid simulation, model-assisted testing, active component testing and machine learning.

Virtual Testing
The principle of Virtual Testing is straightforward: to use analysis software such as Adams, MSC Nastran and Simulink to simulate physical testing. This serves three purposes:
- Analyze and validate the design.
- Aid the setup of the testing equipment.
- Validate the virtual models.
Virtual testing has a strong presence in the engineering development cycle and is growing fast.
It helps engineers understand the complexities in the actual test. The test specimen, the actuator servo valves, could have significant nonlinearities. The test specimen could have undamped vibration modes, causing stability problems when interacting with the actuators. Significant cross-coupling could exist between the actuators causing unforeseen control problems. The models used in virtual testing should have the ability to represent the system with the right components, sized appropriately and with the right modelled physics. They should have the proper fidelity in capturing the physical effects of interest.
What needs to be modelled in virtual modelling depends on a case-by-case basis; some combination of the controllers, the actuators and the engineering system under test is typically implemented. MTS has been modelling its test systems for more than 30 years and the efforts have resulted in the development of many complicated test systems. At the same time, test rig models have become more accurate as more and more validation tests are performed with time.
MTS uses multiple different tools including Adams to simulate their test rigs. Over the years, integrations between MTS software such as RPC Pro and Flextest 793 software and simulation software such as Adams and Simulink have been created. MTS has provided test rig models to its customers and conducted multiple consulting projects to help its customers set up their virtual test platforms.


Hybrid Simulation
In the hybrid simulation approach, a combination of physical tests and virtual models are used to build the system under test. The full systems can therefore be evaluated before their prototyping. In this way, tests can be conducted early in the development process to save time. This hybrid approach is more suited for an extension to test different or more portions of the system under test.
An example is MTS 4-corner damper system shown below. In this system, the physical substructure is the dampers that are tested by the four damper test rigs. The rest of the vehicle structure is in a real-time vehicle model. This hybrid system is used to conduct full vehicle-level evaluation with only the dampers. Therefore, the damper design can be optimized before the rest of the vehicle is available for testing.
The simulation accuracy of vehicle tests by this hybrid simulation system was confirmed to be sufficient for vehicle development, and the results were confirmed to be true during the vehicle prototype test. The hybrid simulation system is an efficient design tool in addition to CAE. It is fruitful both for mechatronics and mechanical engineers to change unknowns into predictable elements before a full vehicle prototype is available.


Active Component Testing
More and more active systems are used in ground vehicle design. To simulate the active systems correctly, the loads from the proving grounds need to be reproduced, and the ECU signals that control the active systems also need to be reproduced. Since the vehicle condition in the lab is very different from the vehicle condition on the actual testing site, the ECU signals generated in a lab environment are not always valid. Therefore, there is a need for a model running in real time with the vehicle on the proving ground for an accurate simulation of the ECU signals that cannot be measured in the lab or are no longer valid.
This approach is called a RESTBUS simulation. The advantage of the RESTBUS simulation is that it reproduces the behavior of the active system under different conditions. It accurately reproduces the target loads and adapts to the vehicle condition during the durability testing, while also simulating the active component adequately.
In the example shown above, the test article is an active suspension. Traditional RPC software is used to reproduce the spindle load. A real-time model on a simulator provides the ECU input signal to the ECU, which generates a current input to control the active suspension system.


Machine Learning Solutions
MTS uses a neural network-based method, the EDM method, to model non-linear dynamic elements such as dampers and bushings. For a non-linear dynamic element like a damper, the force and the velocity are non-linear, and it has hysteresis which cannot be captured by a traditional spline-based method. The EDM method can not only model the non-linear effect but also the hysteresis whereas the spline model can only model the non-linearities.
To learn more, visit Hexagon.
This article was originally published by Hexagon in Engineering Reality Magazine, Summer 2022.