using 'controlchart' and 'findpeaks'

Hello everybody. I’m using the function “controlchart” to analyze something like the data in the picture, however I was wondering how to use the ‘’findpeaks’’ function to identify the peaks in the red circles and also the width of those peaks. Apparently the ‘’findpeaks’’ function doesn’t work properly in this case, it is possible to use both functions in one chart?. Thank you.

[Mathieu NOE]

hello

we could probably better help you if you share a code and data

tx

Answer 1

[Star Strider]

That is going to be something of a challenge, however it is possible.

Using an example from the controlchart documentation:

load parts
st = controlchart(runout,'charttype',{'xbar' 'r'});

Ax = gca;                                                   % Axis Handle
Kids = Ax.Children;                                         % Axis ‘Children’

x = Kids(4).XData;                                          % Choose Desired Variable, Get ‘XData’
y = Kids(4).YData;                                          % Choose Desired Variable, Get ‘YData’
[pks,locs] = findpeaks(y, 'MinPeakProminence',0.5);         % Peaks With Desired Characteristics

figure
plot(x, y)
hold on
plot(x(locs), y(locs), '^r')
hold off
grid

That should work, although it will likely be necessary to experiment with it to get the correct variable and the desired peak characteristics.

[Fercho_Sala]

@Star Strider , @Mathieu NOE ok, according to my question here is the code for the controlchart I have done, however is not possible to generate the peaks inside this plot, based on the curve showed before..

ax9=subplot(1,3,3); 
load parts;
st2 = controlchart(MATRIX1,VECTOR1,'Parent',ax9); 
ax9.YGrid = 'on';
ax9.XGrid = 'on';
ax9.Children(1).Color = 'w';
ax9.Children(2).Color = 'w';

%%%tile and labels
title('Power [dB]');
ax9.XAxis.FontSize = 7
xlabel('altitude/km','FontSize',11,'FontName','Arial','color','default');
ylabel('Power dB for a LT 2.5h (7:30 - 10:00)','FontSize',11,'color','default');
xlim 'manual'
xlim(ax9,[95 151]); 
ytickformat('%g dB');
ax9.XDir = 'reverse';
ax9.YAxisLocation = 'right';
camroll(-90);
legend('off');
hold;

[Star Strider]

@Fernando Salamanca Guerrero — Did you try the code I posted with your data?

I cannot attempt to provide any specific code or guidance without the data, since this is obviously specific to the data.

[Fercho_Sala]

yes @Star Strider , your code works so good it overlaps the picks in the second chart showing the peaks according to the ('MinPeakProminence',0.5 ) paramenter, however my variables for the case of 'XDATA' is something like (XDATA(:,151)) it's a matrix , and apparently this doesn't work properly . :) , in your code it will be something like this:

load parts
st = controlchart(XDATA,y,'charttype',{'xbar' 'r'});

Ax = gca;                                                  
Kids = Ax.Children;                                        

x = Kids(4).(y);                                          
y = Kids(4).(XDATA(:,151));                                          
[pks2,locs] = findpeaks(XDATA(:,151),y,'MinPeakProminence',0.5);         
hold;
plot(XDATA(:,151), y)
hold on
plot((XDATA(:,151)(locs), y(locs), '^r')
hold off
grid

[Walter Roberson]

plot((XDATA(locs,151), y(locs), '^r')

perhaps ?

[Star Strider]

@Walter Roberson — Thank you!

@Fernando Salamanca Guerrero — I cannot comment on how well my code works with your data. It may be necessary to change the code to work correctly with it, since it may have different characteristics from the characteristics of the data in the example code from the documentation that I used. I cannot help you with that because I do not have the data you are using, or your controlchart call. If you provide those, I can do more than guess as to what the correct approach would be.

Answer 2

[Fercho_Sala]

@Star Strider Here is the part of the code with the issue, it has been a bit difficult to identify the peaks in the 'controlchart' plot, so for that reason, I generated another plot for the 'findpeaks' function. However the idea is to mix the both functions in just one chart, overlapping by the 'findpeaks' results on the 'controlchart'. Tthe dataset is attached.

ax9=subplot(1,2,1); 
load parts;
st2 = controlchart(pws1,y,'Parent',ax9);
ax9.YGrid = 'on';
ax9.XGrid = 'on';
ax9.Children(1).Color = 'w';
ax9.Children(2).Color = 'w';

%%%tile and labels

title('Power [dB]');
ax9.XAxis.FontSize = 7
xlabel('altitude/km','FontSize',11,'FontName','Arial','color','default');
ylabel('Power dB for a LT 2.5h (7:30 - 10:00)','FontSize',11,'color','default');
xlim 'manual'
xlim(ax9,[95 151]); 
ytickformat('%g dB');
ax9.XDir = 'reverse';
ax9.YAxisLocation = 'right';
camroll(-90);
legend('off');

%2.5h (7:30 - 10:00)

ax11=subplot(1,2,2);
load parts;

[pks,locs,widths2,proms2] = findpeaks(pws1(:,151),y,'MinPeakHeight',30); 
findpeaks(pws1(:,151),y,'Annotate','extents','WidthReference','halfprom');
text(locs+1,pks,num2str((1:numel(pks))'));
legend('Filtered Data','Peak','Prominence','Width');

ax11.YAxisLocation = 'right';
ax11.XDir = 'reverse';
ax11.XGrid = 'on';
ax11.YGrid = 'off';
title ('Layering (7:30 - 10:00)');
ylabel('Layering identification','FontSize',11,'FontName','Arial','color','default');
xlabel('Altitude/km','FontSize',11,'color','default');
yticklabels('');
xlim(ax11,[75 95]);
ylim(ax11,[10.5 36]);
camroll(-90);

%number of peaks found

numberpks1 =(findpeaks(pws1(:,151),y,'MinPeakHeight',30)); 
numberpks2 =(findpeaks(pws2(:,151),y,'MinPeakHeight',30)); 

[Star Strider]

I was thinking of something like this —

ax9=subplot(1,2,1); 
% load parts;
st2 = controlchart(pws1,y,'Parent',ax9);
ax9.YGrid = 'on';
ax9.XGrid = 'on';
ax9.Children(1).Color = 'w';
ax9.Children(2).Color = 'w';

Kids = ax9.Children(4);
xcc = Kids.XData;
ycc = Kids.YData;

[pks,locs] = findpeaks(ycc, 'MinPeakProminence',1);

figure(9)
plot(xcc, ycc)
hold on
plot(xcc(locs), ycc(locs), '^r')
hold off
grid
xlim([min(xcc) max(xcc)])
ylim([min(ylim) max(ylim)+4])
title('Points Plotted On Control Chart')
xlabel('X Vector')
ylabel('Y Vector')
legend('Points','Peaks', 'Location','best')
text(xcc(locs), ycc(locs), compose('(%3d, %.2f) \\rightarrow ', [xcc(locs); pks].'), 'Horiz','right', 'Vert','middle', 'Rotation',-45)

producing this plot —

Experiment to get different results. (I had to use return after that code to get it to plot correctly, since it does not automatically open a new figure window, for some reason.)

[Fercho_Sala]

@Star Strider thanks, lets's say this is the best way to do it. :)

[Star Strider]

As always, my pleasure!

I definitely agree!