This module is a part of “Engine Carbon Brushes Replacement” project. The purposes of the module are:
a) to build convenient for further work datasets based on the provided raw data;
b) preliminary separate exploratory data analysis of carbon brushes replacement of Cement Mill 3 engine;
c) care by outliers.

The input of the module is a csv file. The file contains an unordered dataset, with provided information of engine inspections since 2004 for each engine carbon brush. In case carbon brush was replaced due to deterioration - the event is signed by “1”, otherwise by “0”.

The outputs of the module are convenient for further work datasets based on the provided raw data.

Statistic information of maintenance periods and cases of carbon brushes replacement due to deterioration:

## Cement Mill 3 (January 2004 - May 2016)
## Maintaneince days summary:
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    8.50   20.00   29.11   31.00  450.00
## 
## Maintaneince days Outleirs:
##   lowerFarOut lowerOutliers upperOutliers   upperFarOut 
##        -59.00        -25.25         64.75         98.50
## 
## Replacement periods summary:
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     8.0   243.0   417.0   462.6   586.5  1973.0
## 
## Replacement periods upperOutliers:
## upperOutliers   upperFarOut 
##       1101.75       1617.00

Graphical Presentation of the Data:

Box Plots of the “maintenance periods” and “cases of carbon brushes replacement”

Low “FarOut” Actually, low farOuts points (days of carbon brush replacement) mean that carbon brushes were replaced on the same day or after few days. In both cases, low farOuts might be entered same information twice- and must be removed from dataset.

Low “FarOut” 

## [1] Mill      C         R         brushName brushType phase     fromDate 
## [8] toDate    Days     
## <0 rows> (or 0-length row.names)


First quantile - most frequently replaced carbon brushes.

Here, it is easy to find out places, where carbon brushes were replaced significantly more frequently than a majority of carbon brushes of this engine. Presumably, these carbon brushes were affected by local factors (like a local clamping spring) more than common reasons (for this engine).


Fourth quantile - most rarely replaced carbon brushes.

This chart provides information about “frozen” (rarely replaced) carbon brushes. Probably “frozen” brushes have a relatively weak clamping spring - as result, the carbon brush, apparently, did not do the job at all. -For example carbon brush C1R6 - was not deteriorated more than five and half years (The criterion for replacement is the size (deterioration) of the carbon brush).

We should take into account, that there are upper outliers in the maintenance list - the reason for extremely long intervals between maintenance is probably data loss of maintenance cases:

##    dateDiff  fromtDate     toDate
## 1       117 2004-04-09 2004-08-04
## 2        84 2006-01-05 2006-03-30
## 3        72 2007-04-17 2007-06-28
## 4        71 2008-02-19 2008-04-30
## 5       101 2008-10-11 2009-01-20
## 6        82 2009-09-08 2009-11-29
## 7        86 2010-03-16 2010-06-10
## 8        80 2011-11-10 2012-01-29
## 9        69 2012-02-27 2012-05-06
## 10       83 2012-05-07 2012-07-29
## 11      113 2013-02-04 2013-05-28
## 12      450 2013-10-13 2015-01-06
## 13      140 2015-01-11 2015-05-31
## 14       81 2015-09-09 2015-11-29
## 15       82 2016-01-03 2016-03-25

There is a chance that deteriorated carbon brushes had been replaced during those periods, but we have no information - and as result, we see the lifespan of them as outliers:

##   brushName   fromDate     toDate Days
## 1      C1R3 2011-03-16 2015-09-09 1638
## 2      C1R6 2010-08-09 2016-01-03 1973
## 3      C2R1 2012-12-12 2016-05-16 1251
## 4      C2R6 2012-02-27 2016-05-16 1540
## 5      C3R4 2012-09-05 2016-01-03 1215
## 6      C3R6 2012-09-05 2016-05-16 1349

In order to exclude the uncertainty- whether carbon brushes lifespan outliers are true, or results of information loss- we can test which of these time periods are overlapping (outliers of periods between maintenance and outliers of periods between carbon brushes replacement).

According to the table, all lifespan outliers might be a result of the lost information. For example, it looks like carbon brush C1R06 was not replaced for about five years, but this period is overlapped with eight periods of lost information, during which carbon brushes might have been replaced for several times.

All outliers are better to be removed from the dataset.


Current condition
In the same way possible to analyze current data in order to predict probable problem (“burned” or “frozen” carbon brushes). This chart is based on current data, taken at May 2016.

“Importance” (to put attention) provides visual information about the number of days since last replacement- lower than second quintile, or upper then third quintile. The importance is calculated by the following principles: Maximum value - 10, is equivalent to the maximum distance from the second quantile down (for the low values -“burned” brushes) and from the third quantile up (for the upper values- “frozen” brushes) found in the whole engine history. The minimum value “1” - the number of days since last replacement, lays in the second or third quantiles.

Colors:
green- the number of days since last replacement lays in third or second quintile;

red- “burned” brush - the number of days since last replacement and lifetime of the previous brush in this place lay in the first quintile;

blue- “frozen” brush- the number of days since last replacement lays in the fourth quintile.