Rapid Solve options

RNE Intensity

Rapid solves use an intensity measure to determine how much of the initial solution is fixed during the rapid solve. The RNE Intensity parameter controls how strictly the optimization model fixes certain decision variables, limiting flexibility in exchange for faster solutions. The RNE Intensity is a value between 0 and 5. Higher RNE Intensity levels lock down more variables, which reduces the complexity of the solution space and leads to quicker results by prioritizing efficiency over flexibility.

To define this option in a Scenario Item, select the Rapid Solve Model Options table and the RNE Intensity field. Set the value to a number between 0 and 5 as needed.

Intensity Levels and Fixing Strategy

Each RNE Intensity level defines a specific set of rules for fixing binary variables and flows. The rules range from fixing all binary and flow variables to leaving all variables flexible, providing a gradient of control based on the desired balance between solution speed and model accuracy.

Level 5: All Binaries + All Flows (0s)

• At this level, all binary variables are fixed based on the optimal solution from the staging model, and all flows with a value of 0 in the staging model are pruned out.

• This approach heavily constrains the model by ensuring that only non-zero flows and other continuous decision variables are being re-optimized in the Rapid scenario, while binaries and zero flows are fixed in the new optimization model.

• This level greatly reduces model complexity, providing faster solution times by focusing on only essential variables and reoptimizing allowed flows and continuous variables.

Level 4: All Binaries

• Fixes all binary variables based on the staging model but does not restrict any flows.

• This level locks down all binary decisions while allowing the model to have flexibility to adjust flows in response to the scenario requirements, maintaining some adaptability.

Level 3: Prioritized Pct Binary (70%) + ToCustFlows (0s)

• Fixes 70% of prioritized binary variables and fixes flows to customers (ToCustFlows) with a value of 0 in the staging model.

• The binary variables are fixed based on the top 70% prioritized binary variables choices, while customer flows with non-zero values, the remaining 30% binary variables, and the continuous variables stay flexible and are re-optimized. This approach provides a controlled constraint level, retaining some flexibility in the binary and flow decisions.

Level 2: Prioritized Pct Binary (50%)

• Fixes 50% of prioritized binary variables based on importance, leaving both the remaining binaries, the flows, and other continuous variables unfixed and being re-optimized.

• This level strikes a balance by constraining only half of the binary decisions, preserving flexibility in flow and remaining binary decisions for dynamic adjustment.

Level 1: Prioritized Pct Binary (30%)

• Fixes 30% of prioritized binary variables, providing light constraints to maintain higher adaptability.

• This level is suitable when a high degree of flexibility is necessary in the model, only fixing a small fraction of binaries.

Level 0: No Fix

• No variables are fixed, allowing the model to optimize all decision variables without any additional model reduction. This is similar to a regular scenario, but you can save time by skipping the regeneration of all input files and warm-starting the solution from the staging model.

• This is the least restrictive level, maximizing model flexibility and adaptability; however, this intensity level is expected to be slower than the other options.

Binary and Integer Variables Priorities for Fixing

The RNE Intensity levels follow a specific order for fixing binary and integer variables. This prioritization ensures that certain decisions are locked first according to predefined priorities, while other variables remain flexible and are reoptimized. When the RNE Intensity level is set to 3, 2, or 1, different percentages of binary and integer variables are fixed in the following order:

1. Selection variables related to facilities, work centers, production, sources, BOMs, and open-close decisions are the ones that get fixed first.

2. Next, binary variables from conditional minimum constraints are fixed.

3. Following these are binary variables from step functions.

4. Then, integer variables related to count constraints are fixed.

5. The last group of variables to be fixed are integer variables from transportation batches, assets, and production lots.

Notes

• Demand and Throughput Limits Changes: No flow fix is applied at intensity levels 5 and 3 when adjusting for demand and throughput limits. Additionally, when RNE changes are related to demand or throughput limits, integer variables from transportation batches, and production lots are not fixed at any RNE Intensity level.

• Prioritized Binary Percentage Fix: For prioritized binary percentage levels (e.g., 70%, 50%, 30%), the model starts fixing binaries from the highest predefined priority list until the specified percentage of constraints is achieved.