|
Fritz Jooste Administrator Posts: 81
2/13/2020
|
Fritz JoosteAdministrator Posts: 81
On the General sheet of the DMS file, there is an option "Trigger Horizon (years)". This post deal with this parameter, what it means and how it can be used in your model.
Note: Our research related to the use of a Trigger Horizon has been published in the Transportation Research Record. You can access the paper in which this research is summarised here.
The Trigger Horizon is used only in Benefit-Cost Analysis (BCA) models in which multiple Strategies are evaluated during the model run. If you do not tightly constrain your triggers, you may get a situation where too many treatments are triggered. This could make your model very slow, or JunoViewer may alert you that the number of allowed strategies has been exceeded.
The figure below shows a simplified example of how this can happen. You can see that an overlay is first triggered in 2021, but then the model keeps adding this treatment as an option for all subsequent years. For the example here, we are only looking at 9 years. But if you are running a model over 30 years, then the number of strategies added will grow exponentially (remember that each option may have several possible follow-up strategies, and these will also be repeated for all subsequent years).
The way to avoid excessive strategies from being generated is normally to close the trigger ranges (i.e. the condition ranges in which treatments are triggered) in your DMS file. However, this has to be done manually and can be a complex and time-consuming task.
You can avoid this problem by specifying a Trigger Horizon with a value greater than zero in your DMS. The trigger horizon specifies the number of years that a treatment can be postponed after it is first triggered. So for a trigger horizon of 3 years, it will repeat a treatment that is triggered in year 2021 for three years and no more (including the year in which it was first triggered). This greatly reduces the number of strategies that need to be evaluated.
For the example shown above, with a trigger horizon set to 3 years, the list of possible strategies will reduce to the following:
Using a trigger horizon can thus greatly (a) simplify your triggers; and (b) speed up your BCA model running times. This is especially important if you are using modelling periods above 15 years. The figure below show an example of how modelling period can affect the number of strategies generated (this figure is for a single treatment length and a specific set of triggers - it does not apply to all cases):
You should, however, note the following important limitations of using a trigger horizon:
1) When you use a trigger horizon, there may be a small number of segments where a less than 100% optimal strategy is chosen. This is because the strategy set does not evaluate all possible strategies, but only those located around the time when the treatment is first triggered. Lonrix internal research suggests that the use of a trigger horizon can get your overall network benefit to within 2% to 5% of a fully optimized model, but at a fraction of the running time.
2) When you use a trigger horizon, the treatment that is triggered in year N, is automatically repeated in year N+1, N+2, ..., [N + trigger horizon -1]. This means that for year N+1, N+2, etc., the model does not check if the treatment is triggered or not. It simply repeats the treatment without rechecking if the treatment meets the trigger conditions. This should not be a problem in most cases, but it could be a problem for second treatments that have to be placed in a certain year (e.g. a second coat seal that HAS to be placed in the year immediately following a rehab). When you use for example a three year trigger horizon, this could lead to the second coats being placed in year 1 after rehab, year 2 after rehab and year 3 after rehab. If one of the latter two options is optimal it will become the chosen strategy. Please do your own investigations on this issue to see what the impact is in such cases.
The suggested approach for using a Trigger Horizon in a BCA model is to first do a thorough exploration of the effect of using a trigger horizon. You can do this by using the Trigger Model Analyzer. You can then upload a DMS with a trigger horizon, check the strategy set on a specific treatment length, and then change your trigger horizon to -1 years (default), save and re-upload your DMS and then compare the resulting strategy set and selected optimal treatment.
Our internal research suggests that in most cases the selected treatment will be the same, except that the model with a trigger horizon of 3 or 4 years will be faster and more stable. However, this may vary depending on your chosen discount rate, trigger windows and many more parameters. So please do your own research to ensure the use of a Trigger Horizon is right for your model.
edited by admin on 10/13/2020
|