Model Tuning

Introduction

In the Cassandra Default Road Network model, the logic that controls key aspects such as Candidate Selection, Treatment Triggering, Increments etc. is coded into the C# Project which you can access online (contact Lonrix for details).

The logic encoded in the domain model references a set of thresholds and constants which are held in the lookup.xlsx file in your project’s inputs folder. By manipulating the thresholds and constants held in the model’s lookup tables, you can significantly alter the behaviour of your model. This allows you to calibrate and tune your model on a project-by-project basis without having to modify the underlying model C# code.

Although all values held in the lookup tables are held in a single table, they are also grouped into Lookup Sets (column name ‘lookup_set_name’) which makes it easy to filter the values in Excel for display and editing purposes.

Each Lookup Set defines a logical grouping of the model thresholds and constants. An overview of the various Lookup Sets used in the default model is provided below. For details on the individual lookup parameters and values held in each lookup set, refer to the detailed documentation using the pages listed on the left.

Important

Note that Options discussed below are by default suited for New Zealand roads. If you are using Cassandra in another country you may find that some of the terminology does not apply directly to your situation.

In such cases, you may have to modify the underlying C# model and introduce new lookup values and remove redundant ones.

Talk to Lonrix Ltd or ASC Ltd if you need help in configuring your model for a custom situation not covered by the default model.

Project Level Lookups

Project level lookups contain thresholds and constants that generally need to be carefully reviewed and refined for each Network. These parameters can be used to calibrate and tune your model to better represent the behaviour of each network. You should not assume that the parameters held in project level lookups apply to every network.

These parameters are kept on the lkp_project sheet of the lookups.xlsx file.

General

Parameters held in under the general lookup set key comprise constants that are used in various stages in the model. These include aspects such as the model base date used to calculate consistent values for Surfacing and Pavement ages based on the Surface and Pavement Dates held in the input data.

For details on parameters held in the ‘general’ lookup set, refer to this page

Candidate Selection

Parameters held in under the candidate_selection lookup set key comprise constants that are used to filter out non-viable elements before the model proceeds to the Trigger and Optimisation stages.

For more details about Candidate Selection, see this page

For details on parameters held in the ‘candidate_selection’ lookup set, refer to this page

Treatment Suitability Scores (MCDA Model)

Parameters held in under the treatment_suitability_scores lookup set key comprise constants that are used in the assignment of Treatment Suitability Scores. These parameters are only used in the Multi-Criteria Decision Analysis (MCDA) model. treatment_suitability_scores For details on parameters held in the ‘treatment_suitability_scores’ lookup set, refer to this page

Treatment Triggering (MCDA Model)

Parameters held in under the mcda_treatment_triggering lookup set key comprise constants that are used in the triggering of treatments in an MCDA model.

For details on parameters held in the ‘mcda_treatment_triggering’ lookup set, refer to this page

Maintenance Prediction

Parameters held in under the maint_pred lookup set key comprise constants that are used in the assignment of Routine Maintenance.

For details on parameters held in the ‘maint_pred’ lookup set, refer to this page

Expected Surface Lives

Parameters held in under the surf_life_exp lookup set key comprise the expected Surfacing Lives (in years) that are used in the determination of the Age At Distress Initiation (AADI). These are important values that need to be determined rigorously for each network based on observed historical trends.

The expected surface life assigned when a surfacing is reset will be a key factor in the determination of how long the surfacing will be relatively free of distress. See the section on the S-curve model of distress for more detail.

For details on parameters held in the ‘surf_life_exp’ lookup set, refer to this page

Domain Model Parameters

Lookup values falling under the grouping of Domain Model are lower level parameters that control and calibrate reset values and increments for the domain model. These parameters are kept on the lkp_main sheet of the lookups.xlsx file.

Generally, you will not need to modify Domain Model lookup parameters for each network. However, you may want to review and adjust some of these parameters if you encounter a network with very different pavement structures or environmental conditions compared to what you have been working on before.

Settling In Values

These values play a role in the estimated historical deterioration rate. See this page for more information.

Surface Class Reset

Parameters held in under the treat_surf_class lookup set define the Surface Class assigned after treatment. The two assigned classes are either ‘cs’ (Chipseal) or ‘ac’ (Asphalt) dependent on the treatment type. See this page for more information.

Surface Material Reset

Parameters held in under the treat_surf_materials lookup set define the Surface Material assigned after treatment, dependent on the treatment type. See this page for more information.

Surface Thickness Resets - New

Parameters held in under the surf_thickness_new lookup set define updated Surface Thickness assigned after a Rehabilitation treatment. See this page for more information.

Surface Thickness Reset - Add

Parameters held in under the surf_thickness_add lookup set define increase in Surface Thickness assigned after a treatment if the treatment is not a Rehabilitation. This value depends on the treatment type. See this page for more information.

Effective Preseal Area

Parameters held in under the preseal_effective lookup set define the relative percentage of the total area that is estimated to require Heavy Maintenance as part of Pre-Seal repairs (in the case of Chipsealed pavements) or Heavy Maintenance (in the case of Asphalt surfaced pavements).

This lookup is essentially a table that converts the Pavement Distress Index to a fraction of the total element area that will be assigned heavy maintenance.

See this page for details.

Road Class

Parameters held in under the road_class lookup set define the conversion of an ONRC Road Category to a High (‘H’), Medium (‘M’) or Low (‘L’) road traffic class. This assigned traffic class is then used in various downstream lookups - for example, to determine which Rehabilitation treatment should be considered for a specific element.

See this page for details.

BCA Objective Weighting

Parameters held in under the bca_weighting lookup set define the weighting factors used to weigh the objective function based on traffic volume and urban/rural situation.

See this page for details.

Pavement Expected Life

Parameters held in under the pavement_expected_life lookup set define the expected pavement life (in years) assigned to a pavement after Rehabilitation. Pavement life is assigned based on a combination of the Urban/Rural and Road Traffic Volumne class (Low, Medium or High). This is an important lookup since the assigned pavement life is used to determine remaining life which in turn affects the probability of distress which in turn affects the time to distress initiation.

See this page for details.

Rut Resets

Parameters related to Rut Resets for different treatments are held in lookup sets ‘reset_exceed_thresh_rut’, ‘reset_perc_improv_facts_rut’ and ‘rehab_resets_rut’.

See this page for details. Factors defined in these lookup sets are explained at this page.

Roughness Resets

Parameters related to Roughness Resets for different treatments are held in lookup sets ‘reset_exceed_thresh_naasra’, ‘reset_perc_improv_facts_naasra’ and ‘rehab_resets_naasra’.

See this page for details. Factors defined in these lookup sets are explained at this page.

Pothole Boost Factor

Potholes reflect a serious and urgent situation which should highlight a need for treatment. However, because individual potholes are fairly small, the percentage area for potholes is relatively small compared to other distresses.

For example, it is not uncommon for a road element that is littered with potholes to show that - relative to the total element area - only 2% potholes is recorded. By comparison, a road that has a similar extent of crocodile/mesh cracking will exhibit 20% or more of the total area as cracked.

Because of this, we allow the quantity of potholes to be ‘boosted’ by a specified factor in the calculation of indexes such as the Pavement Condition Index (PDI) and the Surface Condition Index (SCI). Applying this factor simply means that the recorded pothole area is multiplied by this factor and then that increased area is used to represent the pothole quantity in the calculation of indexes. For more information about the PDI and SDI indexes, see this page

The boosting factor for potholes is held in the ‘distress’ lookup set against the key ‘poth_booster’. A typical boosting factor for potholes would be 10, meaning that the pothole percentage is multiplied by 10 and then the increased quantity is used in the calculation of the PDI and SDI.

Parameters related to the progression and reset of distresses such as Flushing and Alligator cracking are defined in the lookup sets discussed on this page.

Distress Progression

Parameters related to the progression and reset of distresses such as Flushing and Alligator cracking are defined in the lookup sets discussed on this page.