Merged Discussion of femr's video data analysis

The aim in this case is to bring as much accuracy and realism to the WTC7 early acceleration profile as possible. I'd like to see your best effort toward that aim.

Consider the physical property of inertia. The reality of inertia provides physical constraints on what can be expected in an acceleration profile.

How do your acceleration profiles account for the reality that this is a flexible body with high inertia in the object being measured?

................

Other researchers have used similar methods to extract acceleration trends from discrete displacement data before this moment. I gave a few examples. Wouldn't it be smart to look at how they went about doing that and see if they spotted some issues that your graphs don't address at present?

By your current standards, did the authors of those studies make similar mistakes to Femr2? Do those authors use various techniques like Femr2 or do they provide results that look more like yours?

Since it is the first time you are doing this, and since this is connected with the physics of real systems, not just pure mathematics, wouldn't you want to check how other people go about solving the same problems?
 
Last edited:
...
Since it is the first time you are doing this, and since this is connected with the physics of real systems, not just pure mathematics, wouldn't you want to check how other people go about solving the same problems?
LOL, you have no idea what you are talking about. Another failed statement, like your conclusion in the "book".

You have never done engineering, so your first try failed. Reading the papers you supplied, they explain why and how they used the SG filter. Until someone posted what the SG filter is, a few posts ago and you failed to read that post too, you never had a chance to understand you failed to read your own papers posted. Your posted papers expose you can't explain why you would use an SG filter. You are not an engineer, so you don't realize the papers you posted show how and why it is not suited for femr2 data.

Once again your gish gallop using nonsense, has failed, and exposes you don't know what you are talking about. Good luck with the goal free "book" with no rational conclusion.
 
Last edited:
Consider the physical property of inertia. The reality of inertia provides physical constraints on what can be expected in an acceleration profile.

How do your acceleration profiles account for the reality that this is a flexible body with high inertia in the object being measured?

Is high inertia what makes heavy objects accelerate more slowly due to gravity than light objects? Oh wait, that doesn't happen..or does it?

Wile E. Coyote, for instance, must have had a particularly high inertia and could often stand in mid air for several seconds before the acceleration due to gravity took effect.
 
INow... what might be the basis for femr2 to say that the 2-pass curve is more useful?
I already told you...more useful during the earliest moments of descent.

Based on what?
Based upon analysis of the early motion from several locations, several viewpoints, several methods and correlation between acceleration behaviour and velocity behaviour.

It turns out that the 1-pass curve matches much better the trend that was arrived at independently by an experienced engineer by a different method.
...and which includes early acceleration UPWARDS up to 10ft/s^2.

So I must logically ask, more useful for what purpose?
:rolleyes:

The match is a confirmation that femr2's two-pass method to derive the graph is objectively worse, in that it introduces deceiving artifacts.
I beg to differ.

I'd even go farther and say that this casts doubts on the technical merits of femr2's analyses.
Of course you would :rolleyes:
 
How do you know?
Multiple traces in multiple locations, from multiple viewpoints, together with correlation between displacement, velocity and acceleration data generated by various methods...like I said.

Short memory ?
 
...
Since it is the first time you are doing this, and since this is connected with the physics of real systems, not just pure mathematics, wouldn't you want to check how other people go about solving the same problems?
LOL, you have no idea what you are talking about. Another failed statement, like your conclusion in the "book".
I have to agree with beachnut.

Thirty-five years ago, I spent an entire summer doing signal processing research for an aerospace company. Although its applications are classified, one might reasonably assume they were connected with the physics of real systems.

My very first journal article had "time series" in its title. That article was mostly pure mathematics, but the article began by explaining its connection to real systems. Here's a more recent example of applying that research to a real system:
S. N. Burrows, S. T. Gower, M. K. Clayton, D. S. Mackay, D. E. Ahl, John M. Norman, George Diak. Application of geostatistics to characterize leaf area index (LAI) from flux tower to landscape scales using a cyclic sampling design. Ecosystems, volume 5, issue 7, pages 0667-0679.
http://dx.doi.org/10.1007/s10021-002-0110-z​

Argument by experience or credentials is fallacious, but lying about others' experience and credentials is even worse. I find it hard to understand why pseudonymous posters so often think it's the best argument they have.
 
I have to agree with beachnut.

Thirty-five years ago, I spent an entire summer doing signal processing research for an aerospace company. Although its applications are classified, one might reasonably assume they were connected with the physics of real systems.

My very first journal article had "time series" in its title. That article was mostly pure mathematics, but the article began by explaining its connection to real systems. Here's a more recent example of applying that research to a real system:
S. N. Burrows, S. T. Gower, M. K. Clayton, D. S. Mackay, D. E. Ahl, John M. Norman, George Diak. Application of geostatistics to characterize leaf area index (LAI) from flux tower to landscape scales using a cyclic sampling design. Ecosystems, volume 5, issue 7, pages 0667-0679.
http://dx.doi.org/10.1007/s10021-002-0110-z​

Argument by experience or credentials is fallacious, but lying about others' experience and credentials is even worse. I find it hard to understand why pseudonymous posters so often think it's the best argument they have.

I think you are taking this information in a paranoid way. In that sense I see a lot of similarities between your posts and those of Beachnut.

Obviously, one would look at previous attempts to convert discrete displacement data into acceleration profiles. Does the research you cite deal with the reconstruction of acceleration trends from displacement data?

Isn't that the current application? We want to know how others attempted the same thing in the past.

Why not cite a research effort in which acceleration profiles are extracted from position data so that we can see if their techniques are similar to yours or similar to those of Femr2?
 
Last edited:
Major_Tom said:
The aim in this case is to bring as much accuracy and realism to the WTC7 early acceleration profile as possible. I'd like to see your best effort toward that aim.

..1 hour later:

Major_Tom said:
...Does the research you cite deal with the reconstruction of acceleration trends from displacement data?

Isn't that the current application?

Hmm. This guy is confused about what he aims for. I think he would make a good addition to my ignore list.
 
I did a quick search in my Debian repository for Savinzky-Golay and found a package called QtiPlot. It's a graphing package that claims to offer similar features to Origin.

On the positive side, it's easy to use and has several smoothing methods including SG, and several data processing tools including derivatives. Also, it is free software and thus the source is available.

On the negative side, its SG filter does not have any option for producing the derivative output, and its maximum window size is limited to 51. Being free software, it should probably be easy to edit the source to get rid of that limitation, though I haven't tried yet.

Its SG implementation seems a bit weird, in that it accepts different distances for the left and right sides of the window (max. 25 for each). But that should not be a big concern.

So, armed with that tool I went ahead and started experimenting. I was able to reproduce a graph with a much more similar trend to that of femr2's, by using SG with a window of 51 and polynomial degrees 1 and 2, instead of the same for each pass. Here's a comparison for reference:

femr2-qtiplot-comparison.png


That result makes me think that that's what femr2 used, together with some additional smoothing technique, judging by what he's said on the subject. I have no explanation for the time shift, as he has not provided enough data to explain it.

But did he use SG of degree 2 first, and then SG of degree 1 on the derivative, or the other way around? Since in qtiplot the derivative step and the smoothing step are separate, I tested several orderings and obtained this:

SG1-SG2-2Dif.png


Except for the initial 25 samples, for which the algorithm needed to access points outside of the data and the program used zeros instead, causing artifacts, and the final 25 samples, for the same reason, which I have cropped in the graph, the curves match quite accurately. I tried the same with two SG passes of degree 2 and window size 51, and the number of samples I had to truncate was 50, which is what I expected. I'm not sure why it's just 25 here.

There goes femr2's "Second pass is with a different dataset" claim. True but irrelevant.

Anyway. With the above data in my hand, I can say now that femr2's acceleration curve is indeed an acceleration curve, not 'Some Obscure, Undefined Quantity That Femr Mistakenly Calls "Acceleration" (souqtfmcA)'.

The only problem is that he didn't show the actual position and velocity data he derived his graph from: the velocity data he showed was the unsmoothed version, not the smoothed one whose derivative gave the acceleration graph. Thanks to the above result, though, I am now in a position to provide something very close to it. By performing two smoothing passes first, I get the filtered data, and by then performing two derivatives, I get self-consistent velocity data and acceleration data that closely matches his.

Here is the graph of smoothed position data:

SG1-SG2-Vertical-Pos.png


Velocity:

SG1-SG2-Vertical-Vel.png


And acceleration:

SG1-SG2-Vertical-Acc.png


So, while femr2 has not provided these graphs, truth is his graph did represent acceleration. Now it can be questionable whether performing two SG passes in order to smooth the data is proper or not. But criticizing whether his graph represents acceleration is out of place.
 
Last edited:
Its SG implementation seems a bit weird, in that it accepts different distances for the left and right sides of the window (max. 25 for each). But that should not be a big concern.
It is quite common.

That result makes me think that that's what femr2 used
Not right I'm afraid, though I'm glad to see you're digging into these things yourself.

I have no explanation for the time shift, as he has not provided enough data to explain it.
I suggest you look to your implementation for answers.

With the above data in my hand, I can say now that femr2's acceleration curve is indeed an acceleration curve, not 'Some Obscure, Undefined Quantity That Femr Mistakenly Calls "Acceleration" (souqtfmcA)'.
tfk says lots and lots of silly things.

The only problem is that he didn't show the actual position and velocity data he derived his graph from: the velocity data he showed was the unsmoothed version, not the smoothed one whose derivative gave the acceleration graph.
I've provided displacement, velocity and acceleration graphs countless times, for years...

485183061.png

06-05-2011

The actual position data has been available since 14-08-2010.

You cannot generate "unsmoothed" velocity data.

You appear to be basing your comments upon your personal procedure of generating derivatives separately to applying SG smoothing. Your implementation does not generate derivatives via differentiation of the polynomials.

By performing two smoothing passes first, I get the filtered data, and by then performing two derivatives, I get self-consistent velocity data and acceleration data that closely matches his.
There's an awful amount of trend going on ;)

So, while femr2 has not provided these graphs
Again, I've provided umpteen displacement, velocity and acceleration graphs countless times. :(

Have a peek at pages 1 and 2 of this. Each graph has been posted numerous times within this forum.

criticizing whether his graph represents acceleration is out of place.
Thankyou, though I doubt it will make a dent in some others opinions.
 
Last edited:
Just to show that femr2, pgimeno and I have found areas of agreement...

criticizing whether his graph represents acceleration is out of place.
Thankyou, though I doubt it will make a dent in some others opinions.


As I had already stated, but with my agreement highlighted:

All of my graphs show accelerations computed via Savitzky-Golay differentiation, together with femr2's alleged Savitzky-Golay curve. All of those curves, including femr2's, are representations of acceleration for the NW corner of WTC7. All of those curves are faulty. Before I could say which of those curves are best for your purposes, you'd have to explain your purposes.
 
I have been following this discussion and the related parallel track on another thread.

This summarises the key point which has been evident to me for some time:
Just to show that femr2, pgimeno and I have found areas of agreement...

clap.gif

I have been watching for the emergence of areas of agreement rather than persistent re-statements of areas of disagreement. I have abstained from substantive comment because the topic is, quite frankly, beyond the level at which I could confidently add anything to the discussion without doing a lot of research and revision. So I let those with either the pre-existing expertise and/or the energy lead the discussion.

Therefore I commend pgimeno for the effort he has obviously expended to reach this conclusion:
.... But criticizing whether his graph represents acceleration is out of place.

Then this comment by WDC IMNSHO goes to three of the "key issues" (If I dare use the word "key" in a thread where M_T has posted. ;)):
... All of those curves are faulty. Before I could say which of those curves are best for your purposes, you'd have to explain your purposes....

My "key issue" #1:
"...you'd have to explain your purposes." Fully agreed - it is one of the two central issues of a long standing contention I have with Major_Tom. Because "your purposes" clearly relies on two definitions - (a) Who "you" is therefore what purpose that specific "you" has for use of the data.

My "key issue" #2:
- of specific interest to "me" for my "purposes" is the fact that all of the curves seem to agree on the broad factors which are in the area of interest to me and I suspect to a lot of other members viz:
(i) They show a broad range of accelerations;
(ii) Where a period of small "movements" precedes a rapid increase in acceleration;
(iii) which goes beyond "G" for a period of time and for the specific single point being tracked.

My "key issue" #3:
- if all of the "curves are faulty" the range of error does not seem to be big enough to negate:
1) There was "early movement' before "release"; AND
2) The acceleration for the specific single point/small area being measured did briefly exceed "G".

Further discussion down that track takes us outside the scope of the topic. ;) :rolleyes:
 
I suggest you look to your implementation for answers.
I'll double check, but I doubt it's a problem of my implementation.


I've provided displacement, velocity and acceleration graphs countless times, for years...
And you keep missing my point.

How many of these countless times have you posted a velocity graph whose derivative is the acceleration graph you posted? And how many of these countless times have you posted a displacement graph whose second derivative is the acceleration graph you provided? I've suggested you to do so in at least two occasions; my exact suggestion was that you post an integrated version of the acceleration graph.

How many times have you be called on the fact that your velocity graph didn't match your acceleration graph? I think 'countless times' fits here too.

How many needless explanations could you have saved if you posted a velocity graph consistent with that acceleration graph?

That's what I've done above.
 
Last edited:
And you keep missing my point.

How many of these countless times have you posted a velocity graph whose derivative is the acceleration graph you posted? And how many of these countless times have you posted a graph whose second derivative is the acceleration graph you provided? I've suggested you to do so in at least two occasions; my exact suggestion was that you post an integrated version of the acceleration graph.

How many times have you be called on the fact that your velocity graph didn't match your acceleration graph? I think 'countless times' fits here too.

How many needless explanations could you have saved if you posted a velocity graph consistent with that acceleration graph?

That's what I've done above.
You have already stated that your implementation will not generate derivatives from the polynomials. Therefore you have generated your derivatives via a different method. As has been highlighted to tfk countless times, generating derivative data from the polynomials does not output the same results as, say, symmetric differencing. tfk's "inviolate laws" aren't.

Have a peek at how different symmetric differencing is...

352123489.png


...or perhaps a very old symmetric difference based acceleration plot...

82136974.jpg


...quite different from the SG plots...

621711303.png


...one of which you have full details to replicate.

That you don't "believe" what I've said, based on "stuff" you've absorbed from others...is not my problem.

That tfk makes assertions about "instantaneous" values for velocity and acceleration, drawing conclusions using ridiculously small timespans, is also not my problem, and I would advise making sure you avoid replicating such.
 
You have already stated that your implementation will not generate derivatives from the polynomials. Therefore you have generated your derivatives via a different method. As has been highlighted to tfk countless times, generating derivative data from the polynomials does not output the same results as, say, symmetric differencing. tfk's "inviolate laws" aren't.
I don't see how the last sentence follows from the rest. Also, do you have a point? I fail to see any, especially in relation with what I posted.


That you don't "believe" what I've said, based on "stuff" you've absorbed from others...is not my problem.
Do the quotation marks indicate you're literally citing me? I don't recall having said I don't believe you.

For the record, tfk is not the only one who has called you on the mismatch of your velocity graph and your acceleration graph. Others have done the same; C7 comes to mind as an example. Even I did so myself, and it took me long to understand the reason, and once I understood it I wondered why would you not just provide a velocity graph whose derivative was your acceleration graph, instead of posting a nonsensical velocity graph unrelated to the acceleration one even if being purportedly of the same thing. If it was not a deliberate attempt at causing confusion, it was pretty close and certainly had that effect.
 
I don't see how the last sentence follows from the rest...

It doesn't need to follow - it is a stand alone comment. So, for the pedants, it possibly should have been a separate paragraph... ;)


However:
Isn't it true that the four methods in the latter graph show:
1) Some movement in the 11-12 sec interval (the two SG methods only - the polynomial ones obviously out of range);
2) Rapid acceleration in the 12-13 second interval(all four);
3) decreasing accelerations after 13 seconds (all four); AND
4) a period in excess of "G" around the 13 second mark? (all four)

If those four are true is it also true that the arguments over femr2's "accuracy" does not move the bounds of solutions outside those four broad parameters?

I'll leave the obvious consequences till I get either confirmation or reasoned disagreement. :)

If I get neither I may need to point out the bleeding obvious....... :(
 
Last edited:
Is high inertia what makes heavy objects accelerate more slowly due to gravity than light objects? Oh wait, that doesn't happen..or does it?

Wile E. Coyote, for instance, must have had a particularly high inertia and could often stand in mid air for several seconds before the acceleration due to gravity took effect.


Is this the best the group has to offer on the realities of inertia?


To put this on a childlike level of accessibility, The reality of inertia is why mosquitoes can be expected to follow different trajectories than freight trains.


Is this too hard? Mosquitoes can move and vibrate with much quicker alterations in trajectory than freight trains.

How can one apply this to the WTC7 NW corner? Well, think of the corner of the building like a freight train rather than as a mosquito.

Is this simple enough to understand? The corner of the building is ...like...real big. Real heavy, much inertia.


If anyone shows me acceleration profiles that resemble the trajectory of an insect over a 2 second interval and thinks it applies to WTC7, I know that cannot be real.


The property of high inertia places realistic physical constraints upon the system under study.




All of my graphs show accelerations computed via Savitzky-Golay differentiation, together with femr2's alleged Savitzky-Golay curve. All of those curves, including femr2's, are representations of acceleration for the NW corner of WTC7.


No they are not. Some are better than others when realitic physical constraints are applied to look for the best representation. This is not just a math problem. The result is to reflect the most realistic representation of the acceleration of a large and heavy physical object.

It is a physics problem. You can't forget the part about realistic physical constraints.
 
Last edited:

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom