Create a real-time control in DAQ Matlab

Question

han xiao 2018-7-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/412405-create-a-real-time-control-in-daq-matlab

评论： hurrch 2019-10-19

close_loop_design_in_real_time.m

I'm using the Data Acquisition Toolbox for acquiring and generating data with High-Speed Dynamic Signal Analyzers (Modell DT9847 from data translation).

Production link: https://www.mccdaq.com/Products/Sound-Vibration-DAQ/DT9847

I have a Mini SmartShaker™ with Intergrated Power Amplifier (MODEL K2007E01) connecting to output channel as acutator;

Production link: http://www.synotech.de/produkte/datenblatt/?untergruppe=mod_shaker&h=TMS&m=K2007E01

and one Dynamic Force Sensor (MODEL PCB-288D01) connecting to input channel 0&1 for collecting data.

Production link: http://www.synotech.de/produkte/datenblatt/?untergruppe=mod_impedanz&h=PCB&m=288D01

Now I have built up an open loop which containing one output channel and two input channels. The acquisition and generation rates set to

s.Rate = 100000;
%%Construct the output data 
%sweep sine wave source    
t = linspace(0, 60, 6000000)';
outputSignal1= chirp(t, 100, t(end), 500)*1/10;
%fixed frequency sine wave source
data = linspace(0,pi*100,s.Rate)';
outputSignal2 = sin(data)*1/10 ;
%%Queue the output data  
    % Use the |queueOutputData| command to queue the data.
    %first queue 60sec of sweep signal(60sec), after that queue fixed frequency signal.
queueOutputData(s,outputSignal1);
queueOutputData(s,outputSignal2);
%Listener fires every 1/10 sec
s.NotifyWhenDataAvailableExceeds = 10000;  
s.IsContinuous = true;
s.startBackground;

Then I got the result by calling two listeners;

dataAvailableListener = addlistener(s,'DataAvailable',...
    @(src,event)plot(event.TimeStamps, event.Data));
lh = addlistener(s,'DataAvailable',@(src, event)logData(src, event, fid1));
function logData(src, evt, fid)
data = [evt.TimeStamps, evt.Data]' ; 
fwrite(fid,data,'double');
end

Figure 1 Plot data during the acquisition (dataAvailableListener )

Figure 2 Plot data of the whole measurement from log file (lh)

In the first 60sec, the result is from the sweep signal outputsignal1. In the 60-70sec the result is from the outputsignal2 with fixed frequency.

You can see from figure 2, the amplitude of the sweep signal changing with fluctuations when the signal frequency increasing from 100Hz to 500Hz.

Now I would like to build a control loop, which could influence the output signal(amplitude) by setting a trigger for the input channel.

hk = addlistener(s,'DataAvailable',...
      @inputReceived);
function inputReceived(src,event)
if any(event.Data(:,2) > 0.4)
 disp('Too high')
%%control algorithm%%
   end
  if min(event.Data(:,2) < 0.4)
      disp('Too low')
%%control algorithm%%
   end
end

I would like to ask you about the possibility to realize the control loop or do you have any experience on building a control loop in matlab? I'm not a professtional in matlab, so could you give me some tools or algorithm to achieve the goal in real-time control? Thank you!

3 个评论
显示 1更早的评论隐藏 1更早的评论

han xiao 2018-7-31

在 MATLAB Online 中打开

I have figured out that we couldn't build a standard real-time control loop with the Data Acquisition Toolbox in Matlab. The device itself is working with buffers which are the data queued into the output channel. So in my understanding, the real-time control loop is unrealizable because we couldn't influence the data queued into the buffer. Therefore, I decided to create a controlled signal source and then queue the new signal afterwards. Based on what I've created in the workspace, the original sweep sine wave is defined as a 6000000 scans buffer with sample interval of 0.00001. The frequency increase from 200 Hz to 500 Hz, amplitude factor is 1/10.

t = linspace(0, 60, 6000000)';
outputSignal1= chirp(t, 200, t(end), 500)*1/10;

Set the acquisition and generation rates

s.Rate = 100000;

Then I assumed a very basic control algorithm to control the amplitude of sweep. I introduced a variable(factor) to define the desired amplitude factor at each timestamps. for example, if we got the amplitude is 0.5, and we set a reference value equal to 0.4, then the factor should be F=0.4/0.5=0.8; Our code is defined as, (assume the reference value is equal to 0.4)

factor= 0.4/ max(evt.Data(:,2))

Then I saved the factor into a listener

lh = addlistener(s,'DataAvailable',@(src, event)inputReceived);
function inputReceived(src,event)
factor= 0.4/ max(evt.Data(:,2))
            Timestamp = Timestamp + 1
          if Timestamp > 60
              disp('finished')
              Timestamp = 1;
          end
  end
And the listener is fired once per second.

Figure 1: Factor of each timestamps shows in command window. After we acquired all factors for the 6000000 scans in the buffer, then we could create new signal in another workspace.

%% %Create a sweep signal, from 200 Hz to 500 Hz, duration is 60 sec. % queueoutputdata per second, each queue equal to 100000 scans; % Sample interval=0.00001=(x2-x1)/(n-1). % x2=1, x1=o; % increment equal to 5 Hz per second; % 60 intervals in total; %%

t = linspace(0, 1, 100000)';
outputSignal1 = chirp(t, 200, t(end), 205)*1/10* 1.3364;
queueOutputData(s,outputSignal1);
outputSignal2 = chirp(t, 205, t(end), 210)*1/10*1.3381;
queueOutputData(s,outputSignal2);
outputSignal3 = chirp(t, 210, t(end), 215)*1/10*1.3302;
queueOutputData(s,outputSignal3);
.
.
.
outputSignal58 = chirp(t, 485, t(end), 490)*1/10*1.3249;
queueOutputData(s,outputSignal58);
outputSignal59 = chirp(t, 490, t(end), 495)*1/10*1.3188;
queueOutputData(s,outputSignal59);
outputSignal60 = chirp(t, 495, t(end), 500)*1/10*1.3133;
queueOutputData(s,outputSignal60);

I just simply copy and paste every factor to each timestamps. There are 60 new outputsignals in total. If somebody could give me a better idea, like save the factor to a file for example .csv and then array all factors to the new signals, I would really recommend it. The following screenshots are the comparision of results;

But you can see from the second picture, there are several fluctuations evenly distributed between each buffers. We can see more clear when we zoom in to the picture.

I'm very confused about where does the noise comes from...It seems that the noise appeared in every begins or ends of each buffer. If someone could help me, please leave your comment in below. I would really appreciate it!

And one more question about my control design, what is an exact name for my loop? Using factors in the controlled signal source.. is this single-point acquisition and control? Please comment below this, thank you!

Kyle Heintz 2018-8-1

在 MATLAB Online 中打开

Each call to the "chirp" function creates a vector which starts at the value of 1 at time of 0. If you inspect the last values in the "outputSignal1" vector, they are different than the beginning values of "outputSignal2". The queued data ultimately has jumps in values at these points which is likely causing the spikes of noise you are seeing at the ends of each buffer.

A better way to handle this would be to call "chirp" only once for the entire range of time, or develop your own frequency sweep function. You could then extract only the points you want to be queued to the output channel, for example:

queueOutputData(s, outputSignalFull(1:1e6))

han xiao 2018-8-6

Good morning, based on my last request on the noise in DAQ system, I've found out a solution to it.

Now I've been clear, that the noise came from the definition of t = linspace(0, 1, 100000)' of chirp function. Refer to the introduction of "linspace"(Generate linearly spaced vector),

y = linspace(x1,x2,n) generates n points. The spacing between the points is (x2-x1)/(n-1). (from 1. https://ww2.mathworks.cn/help/matlab/ref/linspace.html?lang=en )

we must choose two points x1 and x2 with an interval bigger or equal than 2, which means, we have to select two random points with x2-x1≥2.

For example, t = linspace(0, 2, 100000)' or t = linspace(1, 3, 100000)' etc.

In order to not influene the sample interval of sweep sine wave, we have to modify the total amount of scans from each buffer. Based on the theory of configuration, we have to enlarge the total number of scans.

Sample interval=0.00001=(x2-x1)/(n-1). x2=2,x1=0,n=200000;

If the the acquisition and generation rates is set to s.Rate = 100000; Each buffer will run for 2 seconds ( double the duration of time).

Now we change the acquisition and generation rates to s.Rate = 200000,

According to the QuickDAQ sweep function generation ,If we set the acquisition time equal to 60 sec, the sample interval will be 0,000005 Sec and the number of scans will be 12000000.

s.Rate = 200000;

t = linspace(0, 60, 12000000)';

outputSignal1= chirp(t, 200, t(end), 500)*1/10;

If we divided the chirp signal into 60 equivalence intervals, and we don't change the sample interval at the same time, we have to define 400000 scans per block;

t = linspace(0, 2, 400000)';

Sample interval=0.000005=(x2-x1)/(n-1)=(2-0)/(400000-1). These can lead to a double duration of the acquisition. Each block runs 2 seconds. If we create 60 new signals, the total duration will be 120 seconds.

The figure shows above represented 60 equivalence interval blocks multiplied with respective factors.

Until now, I've realized the basic control loop by adding factors to a group of new signals in workspace, but it worked a little bit complicated. I have to copy each factor in command window to the corresponding signal maunally,because the listener is called once per second, so each second the listener will generate a new factor in command window. I created a new .m file"sweep_signal_source_200000Rate" to save the new signals and then copy to my workspace"close_loop_design_in_real_time" later on.These two files are in my attachment.

Do you know a algorithm in Matlab, which can automatically copy and paste each factor sequentially? This may save a lot of time and reduce the chance of making mistakes. If you have an idea how to deal with it, I'll be very grateful!

Thank you very much in advance!

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Kyle Heintz 2018-7-30

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/412405-create-a-real-time-control-in-daq-matlab#answer_330972

MATLAB and the Data Acquisition Toolbox operate in a non real-time environment, and for this reason true deterministic real-time control is not possible.

If your control loop is not required to be true real-time, you may be able to program the control logic in a loop in MATLAB. It appears that you have already started building the framework for this control loop in MATLAB code. Refer to the following MATLAB Answers post for more information on this workflow:

https://www.mathworks.com/matlabcentral/answers/97354-does-the-data-acquisition-toolbox-support-real-time-closed-loop-control

If you are familiar with Simulink or willing to learn, our products Simulink Real-Time and Simulink Desktop Real-Time are designed for this type of application.

2 个评论
显示无隐藏无

han xiao 2018-7-31

Thank you for your answer, but unfortunately I left the same question in that workflow. I tried with the simulink but it doesn't support my hardware. I have updated new situation in my question, please have a look at it! :D

hurrch 2019-10-19

mathwork DAQ help.pdf

Good day

sorry i have to sneak into this conversation but i really need your expertise Sire.

I am a final year student of an institution in Africa and I have issues with my chosen project cause I do not have prior knowledge of coding or programming.

Please, I really need help with this project of mine as it will be due in 1 weeks’ time (1/11/2019), I don’t mind being assigned to someone good with the matlab DAQ toolbox for an extra fee. This is the crux of my project.

I am to write a code that will read data from two NI 9237 modules (6 channels in total). the modules are to read data from a force balance (strain due to force). The final result should be different plots for different force values against time. I will explain with a diagram and a flow chart, please do correct the flow chart if it is wrong. Some where in the code, there should be a scaling coefficient that will convert strain to force as needed for the equations. The code should plot the data in real time and also at the end of the acquisition. And the plots are individual force values against time. E.g N against time, M against time etc.

This is what I just did playing around with the code but I’m technically stuck afterwards.

daq.getVendors

daq.getDevices

s = daq.createSession ('ni');

ch1 = s.addAnalogInputChannel('Dev1', 'ai0', 'Bridge');

ch2 = s.addAnalogInputChannel('Dev1', 'ai1', 'Bridge');

ch3 = s.addAnalogInputChannel('Dev1', 'ai2', 'Bridge');

ch4 = s.addAnalogInputChannel('Dev2', 'ai0', 'Bridge');

ch5 = s.addAnalogInputChannel('Dev2', 'ai1', 'Bridge');

ch6 = s.addAnalogInputChannel('Dev2', 'ai2', 'Bridge');

ch1.BridgeMode = 'Full';

ch1.ExcitationVoltage = 10;

ch1.NominalBridgeResistance = 350;

ch2.BridgeMode = 'Full';

ch2.ExcitationVoltage = 10;

ch2.NominalBridgeResistance = 350;

ch3.BridgeMode = 'Full';

ch3.ExcitationVoltage = 10;

ch3.NominalBridgeResistance = 350;

ch4.BridgeMode = 'Full';

ch4.ExcitationVoltage = 10;

ch4.NominalBridgeResistance = 350;

ch5.BridgeMode = 'Full';

ch5.ExcitationVoltage = 10;

ch5.NominalBridgeResistance = 350;

ch6.BridgeMode = 'Full';

ch6.ExcitationVoltage = 10;

ch6.NominalBridgeResistance = 350;

attached below is the document to help understand what exactly i am doing and how far i have gone. it also has the flow chart, diagrams and equations.

i really do anticipate your reply.

thank you so much

请先登录，再进行评论。