Get Variable-Values from former Timestep

Get Variable-Values from former Timestep

Jan-08-13 16:50:35

Hi there,
as many askers before, I'm relatively new to programming in Modelica but already realy fascinated from the language power.
What I wasn't able to find yet, is how to get values of a variable from former timesteps.
So like I want to have a Heat-Loss calculation with a temperature-dependent value for the convection coefficient. As the fan that should generate the new convection coefficient needs 3 seconds to start up, I'd need the Temperature value from 3 seconds ago to calculate the current convection coefficient.

Does anyone knows how to do something like that?

Thank you very much for every kind of answer and help.

Daniel

trojax: 2 Posts; Thank you [ 0 ]

#3001 Permalink

Re: Get Variable-Values from former Timestep

Jan-08-13 16:52:44

delay()

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3004 Permalink

Re: Get Variable-Values from former Timestep

Jan-08-13 18:13:28

Thank you very very much!

trojax: 2 Posts; Thank you [ 0 ]

#3039 Permalink

Re: Get Variable-Values from former Timestep

Jan-15-13 19:14:27

sjoelund.se wrote:
delay()

sjoelund,

Is there any similar function to get only the value of a given variable at the END of simulation can consider that as fixed. For instance in the example der(x)=2, I would like to get Y(x(endTime)) and fix that value as constant throughout the simulation. Y will be a horizontal plot over the entire simulation period. In the chart, since the final simulation value of X = 10, I would like to get Y as 10, means Y plot will be a horizontal line of value 10.

I do not know if a combination of delay function and something else can solve that

Scoda: 31 Posts; Thank you [ 0 ]

#3040 Permalink

Re: Get Variable-Values from former Timestep

Jan-15-13 19:18:34

For a horizontal line based on stop-time... use some other tool for post-processing the data. OpenModelica plotting functionality is limited as the project is more focused on simulation results. Octave and Matlab can both read the results and I believe they have plotting tools.

Maybe it's possible to hold the plot in OpenModelica and put additional lines afterwards, but I do not know the API that well.

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3041 Permalink

Re: Get Variable-Values from former Timestep

Jan-15-13 21:03:41

sjoelund.se wrote:
For a horizontal line based on stop-time... use some other tool for post-processing the data. OpenModelica plotting functionality is limited as the project is more focused on simulation results. ...

In fact, I used the graphical plot analog to explain the problem. But my point of interest is to to get the the final simulation output of a dynamic system and plug that as constant input to a static model through a connector. Consider the above example, der(x)=2 with y=f(x), and if I connect this model to a static model that requires only the final value of der(x), I observe two things: one erroneous output and the other a failed simulation

1) the first model (component) has to be simulated for a time t, say t=10
2) the second model being static, the simulation time can be anything.

Now, the two models being connected, when I run the composite model for a simulation time of 1, der(x) sends only the value x that corresponds to t=1. Since the second model needs only der(x) value at t=10, the whole result is thus false. Now, when I use a simulation time of t=10, the the simulation fails. Because, at time t=10 the input to model2 is rather variable instead of a constant input through the connector.

I am sure there must be a workaround to handle this issue. In other programming languages, it is easy to grab a variable result in a given time t and pass it to any other operation. With Modelica, it is very strange that there is no such a thing to grab or if any, I am just looking for it to know how to do that.

I do hope that you would get my point.

Scoda: 31 Posts; Thank you [ 0 ]

#3042 Permalink

Re: Get Variable-Values from former Timestep

Jan-15-13 21:20:33

Ah. That's possible to do in OpenModelica. Of course you would need to make it into two different experiments. One to get the final value of x, and one to use it:

Code:

loadString("model M

  Real x(start=0);

equation

  der(x) = 1.0;

end M;");getErrorString();

simulate(M, stopTime=10.0);

y := val(x, 10.0);

loadString("model N

  Real x = " + String(y) + ";

  Real y = sin(x);

end N;");getErrorString();

simulate(N);

val(y, 0.0);

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3045 Permalink

Re: Get Variable-Values from former Timestep

Jan-15-13 22:45:33

Thanks sjoelund. Yes, indeed they are separate experiments. I didn't use the appropriate term. Now where can I put this piece of code? Is it as a *.mos or inline call through the loadString function?
I have got a scripting function for loadString :

Code:


function loadString

  input String data;

  input String filename = "<interactive>";

  input String encoding = "UTF-8";

  output Boolean success;

end loadString;

Let me take the following two experiments each based on the corresponding model below (the model M is for experiment 1 and model N be for experiment 2):

Code:

model M  //from experiment 1

  Real x;

  Real y;

equation

  der(x) = 2;

  y = x;

  Outoing_Connector = y;

end M;

and

Code:

model N  //from experiment 2

   Real y;

equation

    y = Incoming_connector;

end M;

Now I have to create a third experiment, Experiment 3 that connects Experiment 1 and Experiment 2
How can I use your code based on this situation?

Scoda: 31 Posts; Thank you [ 0 ]

#3054 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 10:50:42

The loadString is just a way of loading a string value instead of a file in a mos-script. I realized another way of doing it that possibly suits you better.

Run as omc a.mos. The disadvantage to the approach above is that here the value is a parameter (which means fewer optimizations are possible). The advantage is that you can write the models as usual and only script the override. (Can also be done using -overrideFile if you have many parameters to set)

a.mos:

Code:

loadFile("a.mo"); // experiment 1 & 2

simulate(M); // experiment 1

y := val(y, 1.0); // read the end value from result-file

simulate(N, simflags = "-override y=" + String(y)); // experiment 2, read the parameter value at the start of the simulation

val(z, 0.5); // verify that it worked

a.mo:

Code:

model M  //from experiment 1 

  Real x; 

  Real y; 

equation 

  der(x) = 2; 

  y = x;

end M;



model N  //from experiment 2 

   parameter Real y; // to be read from file

   Real z = sin(y); // f(y)

end N;

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3055 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 10:52:19

Oh, and to add my output from running that code:

Code:

$ omc a.mos

/home/marsj/trunk/build/bin/omc 1.9.0 Beta2 (r14687)

true

record SimulationResult

    resultFile = "/home/marsj/tmp/M_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 500, tolerance = 0.000001, method = 'dassl', fileNamePrefix = 'M', storeInTemp = false, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*', measureTime = false, cflags = '', simflags = ''",

    messages = "",

    timeFrontend = 0.052824347,

    timeBackend = 0.18675881800000002,

    timeSimCode = 0.05071623099999989,

    timeTemplates = 0.130822174,

    timeCompile = 2.005965308,

    timeSimulation = 0.01198327,

    timeTotal = 2.439157141

end SimulationResult;

2.0

record SimulationResult

    resultFile = "/home/marsj/tmp/N_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 500, tolerance = 0.000001, method = 'dassl', fileNamePrefix = 'N', storeInTemp = false, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*', measureTime = false, cflags = '', simflags = '-override y=2'",

    messages = "",

    timeFrontend = 0.039377649,

    timeBackend = 0.018874146,

    timeSimCode = 0.011657719,

    timeTemplates = 0.018043621000000003,

    timeCompile = 0.265671996,

    timeSimulation = 0.010105428000000001,

    timeTotal = 0.36379121200000003

end SimulationResult;

0.9092974268256817

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3057 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 13:52:52

The mos execution went fine. Very helpful. With your last code (to add results to output), is this to be executed with a separate .mos or a commandLine istruction? In addition, I thought of using omc from local directory path.

Code:

$ omc a.mos

/home/marsj/trunk/build/bin/omc 1.9.0 Beta2 (r14687)

Just one more question if you bear with me

(Can also be done using -overrideFile if you have many parameters to set)

Could you provide me an example based on the above M & N models?
For instance, if I were to pass an array of a variable, say y[:] in model M to z[:] = f(y[:], stopTime in model N)

Again, thank you.

Scoda: 31 Posts; Thank you [ 0 ]

#3058 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 14:23:21

Something like:

Code:

loadFile("a.mo"); // experiment 1 & 2 

res := simulate(M); // experiment 1 

writeFile("N.override","");

stopTime := 1.0;

y1 := val(y[1], stopTime);

y2 := val(y[2], stopTime);

writeFile("N.override","y[1]=" + String(y1) + "\n",append=true);

writeFile("N.override","y[2]=" + String(y2) + "\n",append=true);

getErrorString();

readFile("N.override");

simulate(N, simflags = "-overrideFile N.override"); // experiment 2, read the parameter value at the start of the simulation 

val(z[1], 0.5); // verify that it worked

There does not seem to be a better way to script it

sjoelund.se: 1700 Posts; Thank you [ 159 ]

#3059 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 16:42:45

sjoelund, thank you very much. I will try out the last one.

Scoda: 31 Posts; Thank you [ 0 ]

#3062 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 22:11:57

sjoe, I have for the following feedback:

for a single parameter simulation, it works BUT when I change the simulation time to anything above 1, it prints NaN. One parameter with stopTime=1.0 is ok.

for multiple parameters, the second model can't process data from the first model even if simulation time is set to 1.0. If time is above 1.0, the same way it prints NaN. The record keeps a stopTime of 1

Look at the models and the results below:

Code:

model M

 Real x, y1,y2;

 equation

    der(x) = 2;

    y1 = x;

    y2 = 2*x;

end M;



model N

  parameter Real y1,y2;

  Real z1 = sin(y1); // f(y)

  Real z2 = sin(y2); // f(y)

end N;

Results (simulation time 1):

Code:


record SimulationResult

    resultFile = "M_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 50

0, tolerance = 1e-006, method = 'dassl', fileNamePrefix = 'M', storeInTemp = f

alse, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*'

, measureTime = false, cflags = '', simflags = ''",

    messages = "",

    timeFrontend = 0.00632564799369095,

    timeBackend = 0.0052462182423628,

    timeSimCode = 0.00283488596844051,

    timeTemplates = 0.548435656583297,

    timeCompile = 1.49165812579071,

    timeSimulation = 1.64479135634378,

    timeTotal = 3.69934957640679

end SimulationResult;

1.0

2.0

4.0



record SimulationResult

    resultFile = "N_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 50

0, tolerance = 1e-006, method = 'dassl', fileNamePrefix = 'N', storeInTemp = f

alse, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*'

, measureTime = false, cflags = '', simflags = '-overridefile N.override'",

    messages = "",

    timeFrontend = 0.00677093252141672,

    timeBackend = 0.00439515979274317,

    timeSimCode = 0.00358519237318874,

    timeTemplates = 0.0135766343759655,

    timeCompile = 1.53475471420793,

    timeSimulation = 0.343025548347943,

    timeTotal = 1.90614690201291

end SimulationResult;

NaN

NaN

Results (stopeTime=10) but the script doesn't seem to pass this time. Instead, we always see the default simulation time 1.0

Code:


record SimulationResult

    resultFile = "M_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 50

0, tolerance = 1e-006, method = 'dassl', fileNamePrefix = 'M', storeInTemp = f

alse, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*'

, measureTime = false, cflags = '', simflags = ''",

    messages = "",

    timeFrontend = 0.00602576249542665,

    timeBackend = 0.00445403059543273,

    timeSimCode = 0.0028285642715074,

    timeTemplates = 0.0140187580552248,

    timeCompile = 4.64686487293784,

    timeSimulation = 0.345604405590593,

    timeTotal = 5.01983788008215

end SimulationResult;



10.0

NaN

NaN



record SimulationResult

    resultFile = "N_res.mat",

    simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 50

0, tolerance = 1e-006, method = 'dassl', fileNamePrefix = 'N', storeInTemp = f

alse, noClean = false, options = '', outputFormat = 'mat', variableFilter = '.*'

, measureTime = false, cflags = '', simflags = '-overridefile N.override'",

    messages = "",

    timeFrontend = 0.00647934425037713,

    timeBackend = 0.00441214935325089,

    timeSimCode = 0.00388468276539472,

    timeTemplates = 0.0128057824561847,

    timeCompile = 1.77746204413651,

    timeSimulation = 0.63764823391543,

    timeTotal = 2.4427305621648

end SimulationResult;

NaN

NaN

Scoda: 31 Posts; Thank you [ 0 ]

#3063 Permalink

Re: Get Variable-Values from former Timestep

Jan-16-13 22:15:23

Change overridefile to overrideFile, stopeTime to stopTime, I think

sjoelund.se: 1700 Posts; Thank you [ 159 ]