Optimization objectives parameterization of GA, GS and GSLP

 View Only

Optimization objectives parameterization of GA, GS and GSLP 

Fri March 18, 2022 06:48 AM

Optimization objectives parameterization of GA, GS and GSLP

 

The GA, GS and GSLP optimization models are managing the trade-off between multiple optimization criteria. This is controlled by order, weights, and unit costs.

 

As the number of work orders/operation may be large, it not always possible to reach a global optimality. The focus will be on generating quickly a good solution. The optimizer generates feasible plans wrt the hard constraints and will improve the solutions over time considering all objectives. Overall quality of a plan depends on the time limit used. 

 

By default, GA and GS are ordering the different criteria and performing lexicographic optimization. The first criterion is considered to be the most important one; any improvement of this criterion is worth any loss on the other criteria. The second criterion is the second most important one, and so on. The last criterion is the least important one. The importance of each criterion (and ordering) is defined by a model parameter weight, a higher weight means more important. If 2 criteria have the same weight, they will be considered together.

 

GSLP is by default using a weighted sum of the different criteria with the weight available in the optimization window. The importance of each criterion is defined by a model parameter weight and by internal unit cost parameters.

 

 

 

GA criteria

 

By default, the following objectives sorted by importance are considered by the GA optimization model. Priority is defined by an internal parameter weight that can be change in the “Additional Parameters” page.

 

  1. Maximize the total number of assigned work orders (unperformedCostWeight=100)
  2. Minimize the overall time between the start of the first work order and the end of the last work order (turnaroundTimeCostWeight=10)
  3. Minimize total travel time (Spatial only) (Ω=5)
  4. If the flag “Complete High Priority Work First” is “Consider”, assign priority work orders first (highPriorityFirstCostWeight = 1)
  5. If the flag “Match Skills?” is on, try to match work orders to closest skill level possible (allocationCostWeight=0.9)
  6. If secondary labor craft is being used, maximize the number of default laborers used in comparison to the number of non-default laborers [flag “Match on Secondary Labor?”] (secondaryCraftAllocationCostWeight=0.8)
  7. Minimize non default work zone allocation [flag "Allow Secondary Work Zone Assignment?"] (nonDefaultWorkZoneAllocationCostWeight = 0.7)

 

 

GS criteria

 

By default, the following objectives sorted by importance are considered by the GA optimization model. Priority is defined by an internal parameter weight that can be change in the “Additional Parameters” page.

 

  1. Maximize the total number of assigned work orders. By default, all non-directly infeasible work orders are mandatory. To consider this as a soft constraint and apply a priority/weight on this criterion you can set the model parameter relaxMandatoryActivity = true (unperformedCostWeight=100)
  2. Minimize the number of extra resources used to schedule the work (only with Capacity Planning) (overflowUsageWeight=20)
  3. Minimize the overall time between the start of the first work order and the end of the last work order (turnaroundTimeCostWeight=10)
  4. If the flag “Complete High Priority Work First” is “Consider”, assign priority work orders first (highPriorityFirstCostWeight = 1)
  5. If the flag “Match Skills?” is on, try to match work orders to closest skill level possible (allocationCostWeight=0.1)

 

 

 

GSLP criteria

 

By default, the following objectives are considered by the GSLP spatial optimization model. Weights can be changed in the optimization window.

 

 

  1. Minimize the overall time between the start of the first work order and the end of the last work order (turnaroundTimeCostWeight=10)
  2. Assign priority work orders first (highPriorityFirstCostWeight = 1)
  3. Minimize plan change, i.e., minimize the gap between the optimized plan and the previously published plan, focusing on the first days (based on avoid and limit change time windows parameters) (planChangeCostWeight = 1)
  4. Minimize tasks earliness or lateness based on target start and end dates (targetDateEarlyLateCostWeight=1)

 

By default, all work orders are mandatory in GSLP; in case of infeasibility a conflict set explaining the reason will be generated.

 

 

Multi-criteria as weighted sum or lexicographic optimization

 

This can be changed by changing the value of the parameter ObjectiveUseStaticLex. If not defined the value is true for GA and GS and false for GSLP.

 

 

Unit costs

 

Unit costs are important when a weighted sum is used or if 2 criteria in a lexicographic optimization have the same weights. Unit costs are listed in the parameter documentation document. This includes:

  • maximize assigned work orders: baseUnperformedPenaltyCost
  • minimize turnaround time: turnaroundTimeCostPerHour
  • minimize travel time: transitionCostPerHour
  • high priority work first: baseHighPriorityFirstCostPerHour
  • minimize plan change: avoidChangeCostPerHour, limitChangeCostPerHour, earlinessChangeCostPerHour
  • minimize gap to target start and end time: earlinessCostPerHour, latenessCostPerHour
  • minimize the number of extra resources used: overflowUsageUnitCost, overflowShiftOpeningCostPerHour
  • minimize overskill: overskilledAllocationCostPerHour

 

 

 

Statistics

0 Favorited
4 Views
0 Files
0 Shares
0 Downloads