Data collected from different devices: how to combine for analysis?

[Mikki123]

Hi Everyone,
I'm working on a project where I have current values from three different devices when there is no arc and an arc generated by an arc generator. When I plot them, they all look different since the data is from different devices. Is there anything I can do to make them comparable, like make them look similar, so that I can perform further analysis?

 

[anuttarasammyak]

In your plot where y is current values, what is x ?

 

[Mikki123]

it is just indexes starting from 0 to the number of samples

 

[anuttarasammyak]

However you may get statistics, e.g. average, sandard deviation, as mathematical treatment, Number has no physical meaning. You had better pick up some phisical quantitiy from the samples for plot,e.g. same divice with different physical condition, same condition with different devices.

 

[Mikki123]

All good. I will try doing that. Thankyou

 

[Borek]

it is just indexes starting from 0 to the number of samples

Are they collected over time? Then x is time, isn't it? Are these samples collected in regular periods? Then it is just a matter of knowing frequency, no?

 

[Mikki123]

Hi Borek,
I just have a single column of 800,000 current values. The x values should be time, I suppose. I have the same from three different devices. But while plotting it, they looked so different. I wanted to train my machine learning model with this data for further processing. Since the data all looks so different, I'm getting such poor performance. Do you think the Fourier transformation for all three will make them look similar so that I can better train my model. I'm looking for any kind of preprocessing apart from normalization and feature extraction

 

[hutchphd]

The x values should be time, I suppose.

You should probably know this (?????) It matters that they be equally spaced with no " jitter ".
Then look at (the difference) fourier transform to find interferences.
When you say feature extraction what exactly do you mean?
Why do you expect similar results?

 

[DaveE]

Good data requires a detailed understanding of what it really means; how the measurement instrument works; what it is REALLY measuring. Good lab work is more about experiment design and selecting and/or researching the instrumentation than it is about collecting the data. This would be especially true if you are trying to represent the same physical quantity with different methods.

Without this prior engineering it is likely that the data is meaningless, or has unknown meaning. Bad data can be combined however you like. Garbage in, garbage out applies from the very beginning of experimentation and analysis.

A set of numbers isn't data, it's just numbers. Data has an associated meaning and context.

You will not get useful answers from us if we don't know, in detail, what you are measuring, why, and how.