Data Sensitivity / Data Importance

Which data to focus on? Data Sensitivity or Importance Analysis using Modla's Solvers and Asset Models.
Data Sensitivity / Data Importance

Overview

Northpower used Modla’s existing asset library models for Conductors and Distribution Poles, to determine the most important data to collect and manage for these fleets.

Asset Models: Conductors, Poles

Solvers: Data Sensitivity

Outputs: JSON, Excel

More information on asset models can be found here.

More information on solvers can be found here.

Project objectives

To determine the magnitude of the impact (risk) that the model inputs (data and information)have on the organisation’s population of assets.

Use Cases:

·       To inform missing data collection activities (Contractor Inspections),

·       To prioritise data transformation.

·       To inform business cases and justify which information to collect.

·       To enable better data governance and change control of sensitive data.

What was done

Workshopped sensitivity analysis methodology

Determining how to assess if an input is important, is the most important component of this exercise. A workshop was held, where many ideas were tabled and debated, including:

·       Assessing the remaining life, and the changes thereof,

·       Looking at changes in total probabilities of failure, and

·       Consideration of data quality, data completeness, and applicability to the population of assets

At the end of these workshops, one method stood out: % Overlap of cost profiles.

The intent of this method is to determine how much the resultant cost (including risk and probability) profile changes by assessing the overlap of various cost profiles with changing inputs.

Built and tested the sensitivity solver, and applied the solver to Northpower’s Assets

This solver was created as a composable function that can be attached to any existing asset model.

Some of the results are shown below:

Data Sensitivity 1

In this example, the changing of the model input for “Car Vs Pole Potential” from none to High, has an impact on the cost profile distribution. In this case it increases the likelihood of failure, and subsequently total cost. The difference is significant, but not drastic. In this case 78% of the resultant costs remain the same, with only 22%changed.

Contrast this with something like Pole Age, and you have the following:

Data Sensitivity 2

In this case, there is only a 4% overlap with the profile of a new pole and that of a 100 year old pole.

This means that “Pole age” has a marge larger impact on the resultant cost profile, than the “Car Vs Pole Potential”, and therefore is more significant to understand, and sensitive tochange.

This solver does not currently consider data quality, OR data completeness.

Results

  • Tabulated list of data points, and their overlap / sensitivity for each asset class.
Sensitivity Table

 

Summary

Northpower and Modla worked together to:

  • Define the method for determining sensitivity and importance.
  • Develop a modular solver to implement the method.
  • Apply this solver to their existing asset classes.
  • Generated SME informed focus areas for the data collection and governance.

 

Want to know more?

To discuss this case study, arrange a demo, or to chat to an asset analytics enthusiast, arrange a chat!