I have taken the same code and same data as you have taken , but when you are making Confusion Matrix then your Confusion Matrix is filled with data, meaning it shows that in normal condition the accuracy is 1, there is something else in the inner race but when I When I am performing the same data and code, I am getting only 1 row in the Confucian Matrix, that too the top one, the remaining 13 are coming blank, no data is coming in it, will you help me?
What do i do if i took measurements for 6 axes of data? I am collecting vibration for water pump Xyz for one location and xyz for another location, how do i combine this to test with model?
Hi, just like in case of 1 channel you extract several features from the data. For 6 axes you extract 6times more number of features from the 3 axis data and finally add them to a single dataset
@@Mohankumardash thank you, i concatenated the data from all axis into one array and performed the neural network. Would it be important to first scale data from each axis from -1 to 1? Because my raw data sometimes can reach high levels due to vibration spikes Also if i have readings such as temperature and current measurement, how do i add it in?
HI, nice demo. I have a similar case where the signal is a current signal of the grinding process. When the job approach griding wheel current will rise and after done its become an idle current. like wise for 1 hour window i will have 720 job(5 sec/ job) . can i able to you the same approach here? pls comment
Hello, so I believe your data is a single time series of current signals. And you want to break it into small windows of 1hr. This method should work for your case too, but you will have to adjust the hyperparameters accordingly. And the results will highly depend on the features you extract from the current signal. Hope it helps
hi,great speak! thank you! can you give a book or literature about segmenting the vibration signal into small portions using a window? i wanna read it and cite it once i use this method.
@@abulfahadsohail466 you can use a couple of LSTM layers at the start and then fully connected network, at the end a classification layer with softmax activation for classification task
![Eminem - Fuel (Shady Edition) feat. Westside Boogie & GRIP [Official Audio]](http://i.ytimg.com/vi/k1OquBida9s/mqdefault.jpg)
very nice work...very helpful
Thank you for your videos
I am glad it was useful
I have taken the same code and same data as you have taken , but when you are making Confusion Matrix then your Confusion Matrix is filled with data, meaning it shows that in normal condition the accuracy is 1, there is something else in the inner race but when I When I am performing the same data and code, I am getting only 1 row in the Confucian Matrix, that too the top one, the remaining 13 are coming blank, no data is coming in it, will you help me?
What do i do if i took measurements for 6 axes of data?
I am collecting vibration for water pump
Xyz for one location and xyz for another location, how do i combine this to test with model?
Hi, just like in case of 1 channel you extract several features from the data. For 6 axes you extract 6times more number of features from the 3 axis data and finally add them to a single dataset
@@Mohankumardash thank you, i concatenated the data from all axis into one array and performed the neural network.
Would it be important to first scale data from each axis from -1 to 1?
Because my raw data sometimes can reach high levels due to vibration spikes
Also if i have readings such as temperature and current measurement, how do i add it in?
HI, nice demo. I have a similar case where the signal is a current signal of the grinding process. When the job approach griding wheel current will rise and after done its become an idle current. like wise for 1 hour window i will have 720 job(5 sec/ job) . can i able to you the same approach here? pls comment
Hello, so I believe your data is a single time series of current signals. And you want to break it into small windows of 1hr. This method should work for your case too, but you will have to adjust the hyperparameters accordingly. And the results will highly depend on the features you extract from the current signal.
Hope it helps
@@Mohankumardash Thanks for the reply bro
hello sir , When i apply 2hp there is some error can you resolve it please
Hi, how to solve the overfitting problem
bro from where did u get the all_faults.csv file i am unable to find it?
plz help me out
Hi..there is no all_faults.csv present....u will have to create it by running the CWRU_Bearing_data_preprocessing.ipynb. It is present in the github
@@Mohankumardash yeah i tried to run the preprocessing but i am unable to do it , so could you share the file directly please it would help me a lot
can u please apply the 2 hp in place of 0 hp please? Please make a video taking 2hp load in place of 0hp load.
In this classification is rnn helpful
hi,great speak! thank you!
can you give a book or literature about segmenting the vibration signal into small portions using a window?
i wanna read it and cite it once i use this method.
I did all just from online resources, not from specific book
Thank for ur support
While visualizing the data I am getting the ufunctype error
How to solve this bug
U sure can try with LSTMs
@@Mohankumardash but how
@@Mohankumardash lstms are for prediction but we want classification.
@@abulfahadsohail466 you can use a couple of LSTM layers at the start and then fully connected network, at the end a classification layer with softmax activation for classification task