Degradation Detection Process

Using time histories of  Geometry Data  to investigate
 Unwanted Track Changes and Trends


First the data base of New.g01 files has to be merged,
If we use *F.g01 files, process can be very simple



An Excerpt of the Merged file is shown Below:


Such Files Can be generated for each parameter to be investigated.
The  algorithm we use takes 200 ft of data at a time, and performs change analysis  in that data block.
Then it moves to the next 200 ft.
.


 ...


MORE DETAIL

Every several ft, in the 200 ft block,  time histories are processed to detect the the changes.
A detection is accompanied by a "weight" , a measure of goodness of fit of that observed
  for the type of change being evaluated.  The type of action (1-4)  with the highest weight in the 200 ft. block is reported,
 so long as this weight is greater than a threshold. Also note that saw tooth, if it has a weight greater than
 a threshold, is also Reported. The weight is  : " (measure of the magnitude change) /(Standard deviation of the fit to the Data)".
The Higher the weight the more obvious is the type of action.

An example of a spread sheet with the reported changes is shown next.
In this case there are 24 runs over about 5 years that are used.

...

Note that the column marked "Location" is the click or ft number of the beginning of the action
The Column "Click Number" is the first ft. number in the 200 ft interval analyzed. The Column between
"location" and "Parameter" is the number of runs that are consistent in the 200 ft. interval. In some versions
 of the software, runs eliminated are explicitly spelled out.


Next are plots for the location highlighted in "red" in the spread sheet above. It is  a Trend downward.
......

Color coding is : According to rainbow , Dark red the latest run, Dark Blue the earliest run
In Marked location, Left profile went for +.4 inches to -.6 Inches monotonically over 5 years


Next this same data is plotted in 3-D (Magnitude vs.distance,vs. time). It certainly looks impressive.
..

Next we pick another location and look at gage ( with mean subtracted)



The Above  descriptions should give you a feel for the type of analysis which can be done in scanning large data bases.
All processes have been coded.