Iterate and convert datenums to date times, and find time durations

I have a matrix with wave heights (feet) that change over time. The time is a datenum in the form of yyyymmddHHMM. I need to find the times where the wave height is above a threshold, let's say 5 feet. So I need two things: How do I loop through the array and convert the datenums to datetimes? How do I find the time where the wave height crosses the threshold and goes below the threshold. (I need this to find the time duration that the wave height is above the threshold)

This is what I have so far. It doesn't really work though...

Thanks in advance.

% T is the datenum array
% Waveht is the Wave height array

%% Converting datenum array to date time array
vec1 = datevec('200007130000','yymmddHHMMSS'); %I'm not sure how to iterate through the datetime since this is a string input
a = datetime(vec);

%% polyxpoly
y_line = repelem(5,length(a)); %Creating the threshold line since polyxpoly needs an x value.
x = 0:1:length(a);

y1 = [T Waveht]
y_ps = [x y_line];
[xint1,yint1] = polyxpoly(y1,y_ps)

[KSSV]

You can use datetime class and use logicals to get what you want. If time is your datetime and h is wave height.

idx = h > 5 ;
t = time(idx) ; 

You can loop through datetime class, you need not to worry about it.

[Joy Shen]

Thanks KSSV,

So another question: https://www.mathworks.com/matlabcentral/answers/1594279-converting-datenums-to-get-time-durations

Answer 1

[Peter Perkins]

A couple of preliminaries:

xlsread is not recommended. readtimetable is what you want.

'200007130000' isn't a "datenum", that term means something very specific in MATLAB. Also 'yymmddHHMMSS' can't be write unless you are using zero-based month numbers, which ... don't. To convert that string to a datetime just do this (I'm just guessing on the format):

>> datetime('200007130000',"InputFormat","yyyyMMddHHmm","Format","yyyy-MMM-dd HH:mm")
ans = 
datetime
2000-Jul-13 00:00

So given that, you may not actually be able to use readtimetable, you may need to use readtable to read those timestamps in as text, and then convert them as above, and then use table2timetable.

OK, so you have a timetable, named tt. I don't really know specifically what you mean by "I need the time durations that the component is in operation, above a certain height", but you can use old-school logical operations to find runs of heights > threshold, and then use timetable and table operations to get their durations.

First set up some fake data.

>> WaveHeight = sin(0:.5:50)';
>> tt = timetable(WaveHeight,'TimeStep',minutes(1));
>> %plot(tt.Time,tt.WaveHeight);
>> H = .4;

Now find the rows of that timetable with tall waves, then find the start and end of the runs of tall waves, then remove all the other rows.

>> overThreshold = tt.WaveHeight > H; % rows of tt with height > H
>> tt.RunStarts = [false; diff(overThreshold)>0]; % rows of tt where height grew over H
>> tt.RunEnds = [diff(overThreshold)<0;false]; % rows of tt where height grew over H
>> tt.RunNumber = cumsum(tt.RunStarts) .* overThreshold;
>> runs = tt(tt.RunStarts | tt.RunEnds,["RunStarts" "RunEnds" "RunNumber"])
runs =
  16×3 timetable
     Time     RunStarts    RunEnds    RunNumber
    ______    _________    _______    _________
    1 min       true        false         1    
    5 min       false       true          1    
    14 min      true        false         2    
    18 min      false       true          2    
    26 min      true        false         3    
    30 min      false       true          3    
    39 min      true        false         4    
    43 min      false       true          4    
    52 min      true        false         5    
    55 min      false       true          5    
    64 min      true        false         6    
    68 min      false       true          6    
    77 min      true        false         7    
    80 min      false       true          7    
    89 min      true        false         8    
    93 min      false       true          8    

Now set things up conveniently, and se unstack to reorganize the data to have each row represent one run, and get the durations.

>> runs.RunStartEnd = categorical(1*runs.RunStarts + 2*runs.RunEnds,[1 2],["Start" "End"]);
>> runs.RunStarts = [];
>> runs.RunEnds = []
runs =
  16×2 timetable
     Time     RunNumber    RunStartEnd
    ______    _________    ___________
    1 min         1           Start   
    5 min         1           End     
    14 min        2           Start   
    18 min        2           End     
    26 min        3           Start   
    30 min        3           End     
    39 min        4           Start   
    43 min        4           End     
    52 min        5           Start   
    55 min        5           End     
    64 min        6           Start   
    68 min        6           End     
    77 min        7           Start   
    80 min        7           End     
    89 min        8           Start   
    93 min        8           End     
>> runs = timetable2table(runs);
>> runs = unstack(runs,"Time","RunStartEnd");
>> runs.Duration = runs.End - runs.Start
runs =
  8×4 table
    RunNumber    Start      End      Duration
    _________    ______    ______    ________
        1         1 min     5 min     4 min  
        2        14 min    18 min     4 min  
        3        26 min    30 min     4 min  
        4        39 min    43 min     4 min  
        5        52 min    55 min     3 min  
        6        64 min    68 min     4 min  
        7        77 min    80 min     3 min  
        8        89 min    93 min     4 min  

I'm just guessing at what you have and what you want, but this should get you started.

[Joy Shen]

Hi Peter, thanks for this extensive answer. I see what you mean, so run number is basically my component ID?

My data looks like this if it clarifies anything:

Comp ID Date Height

_________ ____________ ______

1 200007152311 2

1 200007152312 3

1 200007152313 4

1 200007152314 1

2 200007141111 0

2 200007151113 1

3 200007131115 0

3 200007141111 5

And how do I input the column of date strings to convert into datetime?

Date = datetime(T1(:,8),"InputFormat","yyyyMMddHHmm","Format","yyyy-MMM-dd HH:mm")

This obviously gets me an error.

[Peter Perkins]

I don't know what your component ID is.

And I can only guess what error you are seeing. Is it because your table doesn't have 8 variables? The one you show has 3. Is it because you've passed an Nx1 table into the datetime function, not the timestamps themselves? Is it because your timestamps are numeric, not text? The Date variable in the table you show is either numeric, or a big char matrix, which is a bad idea and I don't know how you landed there.

You are not providing enough information to go on.