J85 Simulation

[Antonio Hadade]

Dear Sirs,

I am now simulating J85 engine using T-MATS with the use of published maps (document posted in another topic). Currently, I need to linearize the model for designing controllers. A nice tool called TTECTrA has been mentioned, but it seems to work for turbofans and linearized models (by region). I would like to ask you a few questions:

- Is TTECTrA suitable also for jet engines and non-linear models such as T-MATS?

- Jeff, did you have a nice time in JPC? Is it possible to have access to your article on how to make a T-MATS model out of J85 data or could you explain how have you done it? I have also tried to simulate J85 by scaling the T-MATS example maps, but the RPM value gets different for lower values (for instance, for Wf = 0.1, it should lie around 11000 to 12000 rpm but it turns to be around 14500 rpm), despite the other variables seem to have appropriate values (Rotation would be important since its used in the control laws). The reason I have done this is that the example map is more complete (compressor corrected speed goes beyond 1.0) and also to evaluate the simulation behavior with the example maps.

- Some variables, like P3, have an unexpected behavior when you zoom in the scope: It behaves, at the overall, like in first order dynamics, but if you zoom in, it is like it has sawtooths added to its dynamics. This could make it more difficult for linearization, and this is one more reason why I have tryed to scale the example maps (this bihavior is observed also in that case). Is it possible to avoid those little added sawthooths somehow?

- Have any of you tried to run T-MATS using real time with external hardware (Arduino, for instance)?

Thank you very much!

Antonio

[grc.rhc.dev]

Antonio, 

I am familiar with TTECTrA but have not directly used it. From what I understand it should be usable with your model to create a quick simple controller.

- As for the paper on the J85 modeling effort.  It is currently located at http://arc.aiaa.org/doi/10.2514/6.2016-4527 ,  if you don't have access to this,It is also nearing release as a TM that will be posted to the NASA STI server.

- As for the high rotor speeds at idle.  First, Idle speeds are typically difficult to converge at and data gathered at low speeds from models is sometimes difficult to match due to the high level of non-linearity while operating in those regions. That being said,  very high levels of speed error at those speeds may be due to improperly defined map values (operation is moving off the defined region of the map or the values just don't line up with reality) or the system may not be converging.  In the paper presented at JPC I noted that at low speed values the TMATS generic maps were quite different then the J85 maps I found in literature and I scale the generic maps differently at each speed value. This may help your issue as well. 

- It is hard to say why you are getting jagged behavior in your P3 signal, but my first thought would be that your convergence limit (in the solver) is too high.  If this value is set too high then values may seem to jump randomly as the convergence variables drift around.

- Yes, T-MATS has been run on hardware.  If you have the simulink coder, code can be generated for T-MATS that can be loaded on external hardware.  I have not tried to use them with arduino, but I would assume it was possible if you were familiar with building and running simulink code from that device. 

-Jeff

[Antonio Hadade]

Jeff,

Thank you very much for the link. I was able to recently get access to your work through ITA. Its a very interesting work and really useful in case you only have a portion of the actual engine maps, which may happen in many cases. 

About the jagged behaviour of P3, this problem has been solved by tightening the convergence limit as suggested (changing from 0.01 to 0.001, for instance, will solve the issue. The problem was that an actual number was written in the solver's dialog block and one needs to change it for the corresponding variable's name in order to externally change this parameter). I am currently working on linearisation of the engine model at points of interest with further design of controllers. 

About the external hardware, what I have in mind is slightly different from saving T-MATS in an external board. I would like to still be able to use MATLAB's nice features, such as scopes, control GUIs and SIMULINK flexibility, but controlling the model in real time through one or more external boards, which would require a simulation step in accordance to the desired step time. The model seems to run quickly enough and MATLAB has many possibilities for external communication (even an Arduino package). Has anything like that also been done? 

Thanks

Antonio

[grc.rhc.dev]

Antonio,  

Glad you found the paper interesting. 

As for using T-MATS with MATLAB and external hardware. I have not specifically done this type of thing, so I don't really know if it is compatible. However, I would think it should work if you can use S-functions. Timing shouldn't typically be an issue unless your model is very large, or you have a lot of convergence issues. Most of the models I have built run faster than real time unless it stops converging.

-Jeff