Looking for a way to develop actuator displacement signals on an Adams rig capable of reproducing measured responses on a test (physical) vehicle. Any suggestions for tools or processes?
Looking for a way to develop actuator displacement signals on an Adams rig capable of reproducing measured responses on a test (physical) vehicle. Any suggestions for tools or processes?
You can use Adams Car ride plugin (in Adams Car) for this.
Using the Adams Car Ride four-post test rig for four-wheeled Adams Car vehicle models you can simulate a vehicle traveling over a rough road or simulate a vehicle on a real four-post shaker test machine. You can play displacement or force RPC III file data into the test rig, make your own bumps with table-lookup functions and drive over them, or create and drive over a road-profile surface using a mathematical model for generating road roughness. In the time domain, the four-post test rig also supports sinusoidal sweeps (displacement, velocity, acceleration, or force) and arbitrary Adams Solver functions.
I did not phrase my question well enough. I am looking for a way to define the actuator command signals to make the model respond a certain way.
I ended up speaking to MTS, who develops the RPC Pro software, which does just this for physical testrigs. They also have a way to hook it up to an Adams model to do what I need.
Are you looking for file formats to be imported into ADAMS ?
Ore are you looking for a way to define an actuator that can be "controlled" by a file input to move/excite something in the vehicle ?
Both is included in the above example.
Check the subsystem Wheel_Actuators. They include point force/torque actuators and even a motion actuator for the steering.
If you want to reproduce an existing testrig, it may be smarter to buy that from MTS, but if it's just "I want to move something by reading a file", you should try on your own.
I have a virtual four-post testrig on which a section of a truck frame is mounted. A cab is mounted on the frame with an air suspension. I'm looking for a controller scheme that can figure out what the actuator inputs need to be in order to reproduce accelerations measured on the cab during an on-road test (with a full vehicle).
That is not an easy task. As you say, MTS has a professional software that has been developed over decades that does this and it is not cheap.
And if I remember correctly (over 25 years since I worked with it), it works with an iteratively process, start with sending in some random signal and measure the response and in that way can build up a (non-linear) MIMO transfer function from inputs to outputs. Inverting that transfer function, you then have the control system that you are asking for, where you can provide the desired response of the vehicle and it will apply the inputs that will provide those outputs.
But I never heard about anyone doing it all in Adams. When I worked with it, we used the MTS software on a rig, recorded the actuator motions, which is now a quasi-road, and then applied that in the Adams model. But that only works for very similar vehicles (i.e. you can't use the response of a RAM1500 to get road signals and then apply it to a Fiat 500).
Yes, that is how it works. I've used the actuator displacement signals developed in the lab before and was very happy with correlation against the lab results. Unfortunately, correlation of lab against vehicle isn't always great due to physical limitations of the rig (no powertrain for example) which can be overcome in simulation.
I've been spending a lot of time lately developing a reinforced learning connection between Adams and Bonsai (Microsoft's reinforced learning AI).
This actually could be a very good example to apply that technique to. Let me know of you are interested to learn more. My email is jslat@hexagon.com
The only problem I can see is that it will take quite some time to do the training and the trained neural network will likely only work for vehicles in a similar size and weight range. But training can be sped up by running multiple parallel runs. We are looking for early adopter projects and could give you the chance to try this out for free (or close to).
