Hi All,

Is there any instruction which can help to kick out the noise in Power Meter values. I tried Floating average, but i just want something like continuous comparator and kick out this values which just happens for a second. Any already developed library available for these kinda filtering?

CPU 1212C

TIA V15

Thanks,

Jay

Low pass filter?
https://support.industry.siemens.com/tf/ww/en/posts/81093/
https://support.industry.siemens.com/cs/ww/en/view/6170416

Floating Average?
https://support.industry.siemens.com/cs/ww/en/view/39333120

hello #dp, 

I tried one of the links, but quite unsure whether that is really what I want. Maybe I am not understanding it better. whether those low pass filter compares the nearest values and kicks out large deviated values? Signal processing is what i am trying to understand.

Thanks,

Jay

Hello.

Jay.

hello #dp, 

I tried one of the links, but quite unsure whether that is really what I want. Maybe I am not understanding it better. whether those low pass filter compares the nearest values and kicks out large deviated values? Signal processing is what i am trying to understand.

Thanks,

Jay

The low pass filter is an averaging algorithm. The advantage is that it has "memory" of past events without keeping big record. Please review the detail presented on Wikipedia.

You could decide instead that what you see is in actual fact "noise" and you want to actively catch out this value. The FIFO table can help with that. BUT you have to come up with a rule that says how to evaluate that it is an unwanted value. What rules do you want to apply?

My opinion on the matter is that it is the job of the analytical equipment to decide with statistical analysis what to do with the readings. The readings should not be "corrected" at the recording state.

Here is a bigger problem. What makes you decide that the values that you marked is "noise"? You may see some additional trend if your sampling rate is increased. The readings may follow an ebb-and-flow that follows a sinus wave. Such behaviour should not be ignored.

Analytically there might be something the values are telling you that is important. Example: Switching on spikes and phase variants (heating elements).

If pushing ahead to do some filtering ->

One method is to sample at a higher rate than what you are recording. Taking 3 sample for example near the record trigger. Take then the average of these. Or decide on minimum or highest or any other algorithm you choose.

What are your follow up thoughts.

Thanks for the details #dP. I am thinking something like comparing a set number of values and kick the odd ones out with deviation heavier than 100.

Say for example, [10, 12, 15, 18, 190, 35,20,19]. I just thinking of a comparator window and just kick the odd one out. Because the noise it's quite big. It mayn't be ideal to do this during power up and down in a linear raise, but this power values are not gonna be like a temperature or pressure sensor. So it won't have a great influence in the system.

It is actually a chinese made Power Meter, which at times gives vague value. Floating Average , I didn't feel good about it. Maybe I should increase the sampling rate for Floating average ans see. But if i can work something with a comparator and buffer with FIFO table, then it will be great.

Thanks again for your thoughts,

Jay

Brainstorming picture below

Well it looks good for me. Should I try to replicate and do a test.?

Updated:

Looks like my requirement is like a conveyor belt program. Inspecting the values and rejecting the damaged ones. Only thing is that the comparison is happening between the things in the conveyor and the amount of damage it carries.

I am still thinking about how i can write a logic to take present and past incoming value samples , not really familiar with that kind of implementation.

thanks,

Jay

Jay.

Well it looks good for me. Should I try to replicate and do a test.?

Updated:

Looks like my requirement is like a conveyor belt program. Inspecting the values and rejecting the damaged ones. Only thing is that the comparison is happening between the things in the conveyor and the amount of damage it carries.

I am still thinking about how i can write a logic to take present and past incoming value samples , not really familiar with that kind of implementation.

thanks,

Jay

Perhaps make a program to test in PLCSIM. This allows you to play offline from the actual machine. It is part of the learning process to come up with the code yourself; you already have a concept suggestion.

Export readings you get now to a text file. Re-import that text file into the new test project. Use the imported readings to simulate the outcome of the designed program.

Continue forward by comparing the actual (export) values to the reprocessed values. Consider creating two overlapping graphs in a spreadsheet.

Power Meter is connected through RS485 connection and recivirec data through Modbus. Smoothing is an alternative, instead I would like to kick out these vague values happening just about for a second. 

Thanks,

Jay