| Product(s): | WaterSight |
| Version(s): | 10.00. |
| Area: | Documentation |
Workflow
1) Definition of the key aspects or criteria
Definition of the key aspects that can drive pipe prioritization. The user can choose any aspect taking into consideration both the data available and imported into the solution as well the specific context and requirements of the utility and the region. Those can include (but not limited to): pipe age, material, break rate, pressure changes, customers affected, diameter, proximity to main roads, etc.
Aspects can be defined in order to contribute to the likelihood or consequence of failure, and can be based on the original information provided in the pipe shapefile, hydraulic model results or other contextual information. More information is available under the likelihood and consequence of failure help pages.
To take advantage of the data driven approach, when creating the aspects the user needs to choose the type Predefined.
2) Calculating the scores for each aspect using a data driven approach
Individual scores for each pipe can be automatically calculated based on the respective value of the pipe property, being on a scale from 0 to 100 (100 being the worst), using the linear interpolation method. This approach can be successfully used for most numeric field properties such as (but not limited to):
- pipe age: older pipes usually tend to break more than new pipes;
- break rate: pipes with higher break rates tend to break more than pipes with lower break rates;
- customers affected: pipes that affect a higher number of customers when they are closed cause a higher impact or consequence than pipes that affect few or no customers at all;
- diameter: pipes with bigger diameter usually tend to cause more impact than smaller pipes when they fail (customers affected, potential flooding, repair costs, etc)
- pressure changes: pipes often submitted to higher changes in pressure usually tend to break more than pipes submitted to low pressures variation;
The linear interpolation method is used to automatically calculate the individual scores for each pipe, by linearly interpolating the scores between a minimum and maximum value defined. Results from the interpolation method can change depending on the method defined.
Absolute Method
Minimum and maximum absolute values are automatically calculated by the software from the full range of values belonging to the user defined key aspect. For the minimum value default score is 1 while for the maximum value default score is 100. Scores for all other values are then linearly interpolated between the min and max values defined.
Percentiles Method
Statistical analysis is used to automatically calculate the maximum and minimum value using the percentile 95 and 5 calculated from the full range of values belonging to the user defined key aspect. For the minimum value default score is 1 while for the maximum value default score is 100. Scores for all other values between min and max are then linearly interpolated. Values below the minimum value defined (percentile 5) will share the same score as the minimum value. Values above the maximum value defined (percentile 95) will share the same score as the maximum value.
In order to better understand how the linear interpolation calculation works, as well as the differences between selecting Absolute or Percentiles methods, please take a look at the below example:
Example
Please assume the given series values related to a specific key aspect (for example break rate, in number per length, for each pipe):
| Pipe ID | p-1 | p-2 | p-3 | p-4 | p-5 | p-6 | p-7 | p-8 | p-9 | p-10 | p-11 | p-12 | p-13 | p-14 | p-15 | p-16 | p-17 | p-18 | p-19 | p-20 | p-21 | p-22 |
| Values (break rate) | 0 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 18 | 18 | 18 | 30 | 30 | 35 | 900 |
We want to automatically score each pipe based on its value, knowing that higher numbers should be scored worst than lower numbers. Also please assume that utility already knows that pipes with break rates near or more than 30 should be already classified with a medium grade (or average performance), based on national regulator guidance.
For the absolute method, using the linear interpolation calculation, the results are the following:
| Pipe ID | p-1 | p-2 | p-3 | p-4 | p-5 | p-6 | p-7 | p-8 | p-9 | p-10 | p-11 | p-12 | p-13 | p-14 | p-15 | p-16 | p-17 | p-18 | p-19 | p-20 | p-21 | p-22 |
| Values (break rate) | 0 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 18 | 18 | 18 | 30 | 30 | 35 | 900 |
| Score (absolute) | 1 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 4 | 4 | 5 | 100 |
You can notice that the minimum value is 0 (with score 1) while the maximum value is 900 (with score 100). Values for all other pipes were linearly interpolated between these two extremes. Taking into consideration that all values are much closer to the 0 (and therefore to the score 1) then to the value 900 (and therefore to the score 100) it is expected that when using the linear interpolation method all pipes will have a very low score (score from 1 to 5 in a scale of 100).
This is just an example, but the goal is to illustrate how a unique and single high value (that can correspond to bad/wrong data or even a correct value but that only happened once, very unlikely to happen again and therefore represents an outlier) can influence all the results when absolute method approach is used. In this case using the absolute method all pipes (with the exception of the pipe with 900 value) would fall in the low grade or good performance (green range). This go against the utility guidance, where a value higher than 30 should already be not considered good performance or low grade. Therefore the absolute method approach should only be used for more homogeneous series of values.
Now, doing the same analysis using the percentile method approach:
| Pipe ID | p-1 | p-2 | p-3 | p-4 | p-5 | p-6 | p-7 | p-8 | p-9 | p-10 | p-11 | p-12 | p-13 | p-14 | p-15 | p-16 | p-17 | p-18 | p-19 | p-20 | p-21 | p-22 |
| Values (break rate) | 0 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 18 | 18 | 18 | 30 | 30 | 35 | 900 |
| Score (absolute) | 1 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 4 | 4 | 5 | 100 |
| Score (percentile) | 1 | 1 | 4 | 8 | 11 | 14 | 18 | 21 | 24 | 27 | 31 | 34 | 37 | 41 | 44 | 44 | 44 | 44 | 84 | 84 | 100 | 100 |
Using the percentile approach the minimum value is the percentile 5 of the series corresponding to the value of 5 (with score 1). The maximum value is the percentile 95 of the series corresponding to the value of 35 (with score 100). Values for all other pipes were linearly interpolated between these two extremes. Taking into consideration that values are now much more in the middle between 5 and 35, when using the linear interpolation method all pipes will now have a more meaningful score (values closer to 35 will be scored the highest while values closer to 5 will be scored the lowest). The value of 900 is treated as an outlier.
For some specific key aspects there may exist already some guidance from national regulator or through benchmarking with other utilities from which values (for example from which break rate value) pipes can be considered with poor, medium or good performance. In that case the user can also use the custom method (where he can define exactly the min and maximum values and the respective scores) or use the discrete score type or even use a decision tree method to score pipes.
More information about the predefined aspect can be found here.
3) Combining different aspects together
After the key aspects are defined (1), and scores for each pipe and aspect are calculated (2) the user needs to combine them together to build his multi-criteria decision ranking system. The multi-criteria decision ranking system is composed by the different key aspects (and scores) as well as the weight for each aspect. The final pipe score is obtained doing a weighted average of the individual pipe scores for each aspect.
The software is flexible enough for the user to be able to include every key aspect in the same decision ranking system (mixing risk - likelihood and consequence of failure - with asset performance) or by creating a ranking system to separately assess likelihood of failure (LOF) and another ranking system to assess consequence of failure (COF). Then the user can combine both LOF and COF results to calculate the risk score using a risk matrix.
Below an example of a multi-criteria decision ranking system, with the aspects and respective weights:
Figure 1 - Example of a multi-criteria decision ranking system to assess likelihood of failure.
4) Identification of the priority pipes
After the key aspects are defined (1), scores for each pipe and aspect are calculated (2), and the multi-criteria decision ranking system is defined (3), the priority pipes can be identified. Priority pipes are those represented with highest scores and by default are color coded red and represented with a high grade in the application.
Below an example of the results displayed in the map, with the identification of priority pipes at red. Results below are based on a decision ranking system specifically focused on consequence of failure (where proximity to main roads is the most relevant aspect). The user is also able drill in into the high-grade pipes to identify the exact scores and pipes ID.
Figure 2 - Results based on a decision ranking system specifically focused on consequence of failure where proximity to main roads is the most relevant aspect.