harislulic / zema-machine-learning-tutorials Goto Github PK

During the last 7 weeks, we were working on uncertainty evaluation of ZEMA data in another repository which will be incorporated into this repository soon (Wednesday maybe)

I receive a

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-18-e491d44dd19f> in <module>
      8 for i in range((len(sensor_train))):
      9 
---> 10     sensor_train[i]=((sensor_train[i]*gain[sensor_num])+offset[sensor_num])*b[sensor_num]*k[sensor_num]

NameError: name 'sensor_num' is not defined

on execution of

#if you did not convert the signal to SI units, convert here
offset=[0, 0, 0, 0, 0.00488591, 0.00488591, 0.00488591,  0.00488591, 1.36e-2, 1.5e-2, 1.09e-2]
gain=[5.36e-9, 5.36e-9, 5.36e-9, 5.36e-9, 3.29e-4, 3.29e-4, 3.29e-4, 3.29e-4, 8.76e-5, 8.68e-5, 8.65e-5]
b=[1, 1, 1, 1, 1, 1, 1, 1, 5.299641744, 5.299641744, 5.299641744]
k=[250, 1, 10, 10, 1.25, 1, 30, 0.5, 2, 2, 2]
units=['[Pa]', '[g]', '[g]', '[g]', '[kN]', '[bar]', '[mm/s]', '[A]', '[A]', '[A]', '[A]']

for i in range((len(sensor_train))):

    sensor_train[i]=((sensor_train[i]*gain[sensor_num])+offset[sensor_num])*b[sensor_num]*k[sensor_num]

in 2_Features_extraction_and_selection.ipynb. Could you please have a look at the lines and fix this issue?

0_Data_Import_and_Visualisation.ipynb:

Subtitle: Basic visualisation for imported dataset
1st line: isto real-time - into real time

If you have problems with previous step, you can skip conversion into SI units by runing next cell.- double n is missing in "running"

1_FFT_and_Reconstruction.ipynb

If you have problems with previous step, you can skip conversion into SI units by runing next cell.- double n is missing in "running"

2_Machine_Learning_using_Best_Fourier_Coefficients.ipynb:
Subtitle: Converting into SI Units
If you have problems with previous step, you can skip conversion into SI units by runing next cell.- double n is missing in "running"

Subtitle: Feature extraction using BFC:
"This step an unsupervised extraction method" - "is" is missing

"A function is created, which takes as input:
data from one senzor sensor," - s insted of z

Subtitle: Pearson correlation
"Example: If element in 'senzor_n' is 5, and the element in 'feature_n' at the same position is 50, that means that you can access that feature in this way:" - s instead of z

Subtitle: Apply patterns for feature extraction and selection on testing data
"Further selection is based on the indices found with Pearason correlation method. " - Pearson instead of Pearason

Subtitle: Mahalanobis distance -
"The Mahalanobis distance is a measure of the distance between two points and unknown character" -What is with the unknown character, check this

3_Machine_Learning_using_Statistical_Moments.ipynb
If you have problems with previous step, you can skip conversion into SI units by runing next cell.- double n is missing in "running"
Subtitle: Pearson correlation
"Example: If element in 'senzor_n' is 5, and the element in 'feature_n' at the same position is 50, that means that you can access that feature in this way:" - s instead of z

On execution of 2a_Features_extraction.ipynb I get a

FileNotFoundError: [Errno 2] File C:\Users\jugo01\Desktop\sensor_units.csv does not exist: 'C:\\Users\\jugo01\\Desktop\\sensor_units.csv'`

on execution of

sensor=pd.read_csv(r'C:\Users\jugo01\Desktop\sensor_units.csv')

in 2a_Features_extraction.ipynb. Could you please include some documentation where to obtain this file and ideally replace the absolute, local path. I would really appreciate your support in getting things working!

In 0_Data_Import_and_Visualisation.ipynb the out-dated conversion factors are used, that lead to non-SI units, such as "g" for acceleration. The updated table is available on Zenodo and should be used instead.

Also, the scaling seems to be wrongly implemented for the plotting, because the amplitudes of the signals before and after the scaling are identical. For instance, the microphone data in Pa should be around +/- 0.4 instead of +/- 200000.