Help, I need to know how to build a dataset. I have 7 txt file of bearing raw vibration generated. If can I need step by step. Already three days search.

Here attached with all raw vibration data mentioned above.

[William Rose]

@Rusyaidi Yusri,

I looked at the contents of the seven files you uploaded. Each file has two columns of numbers, and no labels. Column 1 appears to be time. The sampling rate is 1000 Hz for all seven files. The duration is 1 second for 4 files, and 10 seconds for 3 files. I assume column 2 is a measure of vibration, such as force or displacement or acceleration.

The file names indicate the seven recordings are made under seven distinct conditions. Please explain how you plan to analyze the seven recordings, and what information you plan to obtain from the database. I am asking because the structure of the database should be related to the intended use of the database.

What do you want to do with this data?

[Rusyaidi Yusri]

Hi, I need to make it as trained dataset for supervised model. But I lack of knowldge, how to do that.

Answer 1

[Tejas]

Hi Rusyaidi,

In supervised learning, models require the provision of expected values or classes alongside the data entries they will predict on. Therefore, one way to create a training dataset from the provided text files is to assign all entries within a text file to a specific class. Here is a step-by-step procedure to accomplish this:

• Begin by loading the contents of the text files

    data = cell(7, 1);
    fileNames = {'bearing_defect_ball.txt',  'bearing_defect_on_Retainer.txt'...
        'bearing_defect_Outer_Race_vibration.txt','bearing_vibration_defect_on_seal.txt',...
        'bearing_vibration_defect_on_shield.txt', 'defective_Inner_Race_vibration.txt',...
        'Normal_bearing_condition_raw_data.txt'};
    for i = 1:6
        data{i} = readtable(fileNames{i}, 'Format', '%f%f', 'ReadVariableNames', false);
    end
    data{7} = readtable(fileNames{7}, 'HeaderLines', 1);

• Next, add an extra column to dataset. This column should indicate the class to which each entry belongs.

    class = {'class_1','class_2','class_3','class_4','class_5', 'class_6','class_7'};
    for i = 1:7
        data{i}.Condition = repmat(class(i), height(data{i}), 1);
    end
    dataSet = vertcat(data{:});

• Divide the dataset into training and testing portions using cvpartition. For additional information on this function, consult the following documentation: https://www.mathworks.com/help/stats/cvpartition.html .

    cv = cvpartition(height(dataSet), 'HoldOut', 0.3); 
    trainData = dataSet(training(cv), :);
    testData = dataSet(test(cv), :);

• Proceed with training the dataset using a supervised model. In this example, a decision tree is utilized. More information on this can be found in the provided documentation: https://www.mathworks.com/help/stats/fitctree.html .

    model = fitctree(trainData(:, {'Var1', 'Var2'}), trainData.Condition);

This method represents one of several approaches to constructing a training dataset for supervised models. The final structure of the training dataset will be determined by the intended use of the data contained in the text files.

[Rusyaidi Yusri]

Oh my, thank you so much!