The system produces a wide range of deterministic models for groups of similar pavements. The software is highly flexible in terms of allowing these models to be developed and analyzed automatically as well as visually examined for their accuracy and reliability. User-defined models are also allowed to supplement cases where the data is inadequate to develop statistically valid models.
The general methodology for building performance models in the system is as follows:
- Prepare a set of data composed of Y (condition index) and X (pavement age) from the performance master file.
- Transform Y (both the linear transformation, as applicable, and the 0 - 100 scale transformation).
- Divide the dataset into groups based on the Performance Class Variables.
- In each group, run regression analysis (linear or non-linear, as applicable to the model type) without removing any data points.
- Identify outliers.
- Remove all outliers from the data set and run the regression analysis again, if desired.
- Review the statistical outliers to determine if a user-defined set of outliers should replace the statistical outliers. Run the regression analysis again. Select the model or define user parameters until better data is available for the data set.
Several data elements must be configured before models can be specified:
- A set of Performance Index (PI) definitions must be developed.
- A set of Performance Model Types (i.e., Linear, Exponential, etc.) must be established.
- A set of Performance Class Variables must be defined.
Performance Class Variables define the homogeneous sets of pavement sections within the roadway network. (Homogeneity in this case means that, for performance class variables, pavements with similar road structure category perform in a similar manner and provide more uniform data for performance models.) During configuration of the system, AgileAssets and staff from your agency determined which variables are available for creating performance classes. In the analysis, you can use all or none of the available variables to create performance models by making them active or inactive.
A performance class (model group) is defined for each combination of the values of the activated performance class variables. For example, if you have AADT and Functional Class as Performance Class Variables, and there are eight functional classes defined and three levels of traffic, then this produces a combination of 8 x 3 =24 groups for creating models and results in a matrix of groups as shown in the table below.
Low Traffic AADT ≤ 8000 | Medium Traffic 9000 < AADT ≤ 20000 | High Traffic AADT > 20000 | |
|---|---|---|---|
Interstate Urban | 1 | 9 | 17 |
Interstate Rural | 2 | 10 | 18 |
NHS Urban | 3 | 11 | 19 |
NHS Rural | 4 | 12 | 20 |
Arterial Urban | 5 | 13 | 21 |
Arterial Rural | 6 | 14 | 22 |
Non-Arterial Urban | 7 | 15 | 23 |
Non-Arterial Rural | 8 | 16 | 24 |
In the system this matrix is represented using a tree structure in the Default Model Structure window. Caution is recommended when choosing to use a Performance Class variable for modeling. When the number of performance classes is increased, the number of models that must be maintained increases exponentially. When selecting performance classes and defining their levels, keep in mind that depending on how you structure the model tree, the total number of models required by the system may be as many as the number of Performance Classes x the number of Defined Performance indices. Therefore, if you have 10 PIs and 24 Performance Groups you will need to create and maintain 240 models.
The following sections details how to setup performance models in the system.
