Co-simulation or full export?

When a full export is feasible it is always better than aco-simulation. Indeed having the entire system integrated by one solver makes the analysis of the system easier. For example using the linear analysis facility in Simcenter Amesim is more accurate if Simcenter Amesim knows all states of the system.

During a full export to Adams, Simcenter Amesim is used as a function evaluator by Adams. All derivatives and outputs are passed to Adams whenever it asks for them. The Simcenter Amesim solver is not used and values in Adams may be discontinuous and may make the solver fail.

A co-simulation can be easier to set-up, especially if the models are complex. During a co-simulation both solvers exchange information at a fixed interval. When Adams receives a signal, this will stay constant during the whole sample interval, until a new value is received. The same thing happens in Simcenter Amesim. You will have to set this communication interval with care since different values can lead to different results.

We want to use as large an interval as possible to get a quick run and as small as possible for accurate values! Sampled environments are a difficult subject and there are books written on the stability of them. How do we efficiently set up a communication interval for discrete mode?

Keeping in mind that introducing a communication interval is more or less equivalent to introducing a delay in the system, the following study can be done.

We consider a linear actuator modeled within Simcenter Amesim and coupled with an Adams model. In the figure you will notice that some variables are exchanged. The torque is here always zero.

Figure 15: Simcenter Amesim sketch of a linear actuator coupled with Adams

An equivalent system is given above. A simple calculation will give the global inertia which must be the one of the Adams structures on which the Simcenter Amesim force acts. You can also include some friction if there is any in the Adams model.

In order to test the validity of a chosen communication interval we want to construct a similar system in Simcenter Amesim only.

Figure 16: Simcenter Amesim sketch with a delay loop.

The sample period of the discrete delay blocks is set to the communication interval we are testing.

In addition you must also construct the same system in the same sketch without the discrete blocks. With this sketch you can then compare results for both systems, with and without discrete blocks depending on the delay=communication interval you have chosen.

Figure 17: Simcenter Amesim sketch without discrete blocks.

Run simulations with different delays (communication interval) and check that results on both systems are equivalent. Choose the highest communication interval that preserves the quality of the results.

With this methodology you are able to establish an optimal communication interval.