Descriptive and inferential statistical methodologies have proven effective in creating a proactive ‘alarm’,

accurately identifying Pump Failures with Horse Power (HP) and Pump Efficiency (PE) emerging as key
variables of interest with deviations of 15 HP and > 3 % PE being our core signal thresholds.

Statistical Alarm plot

45.00 1.20
40.00
1.00
35.00
30.00 0.80
25.00
0.60
20.00
15.00 0.40
10.00
0.20
5.00
0.00 0.00
2 167 332 497 662 827 992 11571322148716521817198221472312

Horse Power Pump Efficiency

PUMP FAILURE (1 or 0)

Correlation Coefficient Pump Failure Rolling

Stdev
0.7
0.68
0.66
0.64
0.62
0.6
0.58
0.56
0.54
1

Horse Power Pump Speed (RPM) Pump Torque

Ambient Temperature Volumetric Flow Meter 1 Pump Efficiency
Volumetric Flow Meter 2
1
Descriptive Analysis has enabled us to clearly identify particular signature abnormalities showing
clear signature changes in both Rolling Standard Deviation and Rolling Mean Datasets when observed
over the respective failure period of interest.

DRT Rolling Mean Plot DRT Rolling Stdev Plot

300.00 6.00 10.00 10.00

250.00 5.00 9.00 9.00

200.00 4.00 8.00 8.00

150.00 3.00 7.00 7.00

6.00 6.00
100.00 2.00
5.00 5.00
50.00 1.00
4.00 4.00
0.00 0.00
3.00 3.00
Volumetric Flow Meter 1 Volumetric Flow Meter 2
Pump Speed (RPM) Pump Torque 2.00 2.00
Ambient Temperature Pump Efficiency
1.00 1.00
Horse Power PUMP FAILURE (1 or 0)
0.00 0.00

Volumetric Flow Meter 1 Volumetric Flow Meter 2 Ambient Temperature

Pump Efficiency PUMP FAILURE (1 or 0) Pump Speed (RPM)
Pump Torque Horse Power

2
Further segmentation of the data via binary means (Pump Failure = 0 or 1) illustrated through box
plots, shows a clear signature difference between that of normal behavior and that of failure with
Pump Torque, Pump Speed (RPM), and Ambient Temperature showing the 3 largest variances.

Rolling mean
0
1

Rolling Stdev
0
1

3
 Correlation analyses across datasets yield interesting insights with Volumetric Flow Meter 2 and
Pump Efficiency negatively correlated with Pump Failure in the Rolling Mean (30 Minute) Data, whilst
Horse Power and Volumetric Flow Meter 1 show a subsequently strong positive correlation in the
Rolling Standard Deviation (30 Minute) Dataset.
RAW
Rolling mean
0.5
0.4
0.4
0.2 0.3
0 0.2
Correlation Coefficient
-0.2 0.1
0
-0.4 Correlation Coefficient
-0.1
-0.6
-0.2
-0.8
Horse Power Pump Speed (RPM) Pump Torque
Horse Power Pump Speed (RPM) Pump Torque Ambient Temperature Volumetric Flow Meter 1 Pump Efficiency
Ambient Temperature Volumetric Flow Meter 1 Pump Efficiency Volumetric Flow Meter 2
Volumetric Flow Meter 2

Stdev
0.7
0.68
0.66
0.64
0.62
0.6
0.58
0.56
0.54
Correlation Coefficient

Horse Power Pump Speed (RPM) Pump Torque

Ambient Temperature Volumetric Flow Meter 1 Pump Efficiency
Volumetric Flow Meter 2 4
Lastly, analysis of the model fit reveals that with an R2 of 0.7789, a linear model is a good fit for the data with variables Horse Power,
Volumetric Flow Meter 2, and Pump Efficiency having the largest coefficients, indicative that these variables have the most immediate
relationship with respect to Pump Failure behavior.

1
0.8
0.76484538773
0.6
2065
0.4
- - -
0.2 0.34236588972 0.01786733605 0.01890199470 0.03577966745
4241 16365 88853 08333
0 0.05093394392 0.01820688729 1
-0.2 70242 6729 -
-0.4 0.61136432959
0946
-0.6
-0.8

Horse Power Pump Efficiency Volumetric Flow Meter 1 Ambient Temperature

Pump Speed (RPM) Pump Torque Intercept Volumetric Flow Meter 2

Regression Analysis Summary: Pump Failure Behavior

•Model Fit: 2R2 of 0.7789 indicates a strong linear fit.
•Key Predictors:
• Horse Power: Highest positive influence.
• Volumetric Flow Meter 2: Notable negative impact.
• Pump Efficiency: Significant positive relationship.
Insight: These three variables play pivotal roles in determining pump failure behavior.