Run metrics
- Last UpdatedAug 11, 2025
- 4 minute read
Click the Metrics button in the Views group of the Home ribbon tab to open a menu with the following options:
-
Solution Metrics: Opens a dialog box listing various metrics on the current solution.
-
Multi-Start Metrics: Opens a dialog box listing various metrics on the latest multi-start optimization.
-
Multi-Start Metrics Chart: Opens a dialog box displaying a series of charts with statistics on the latest multi-start optimization.
Solution metrics
|
Item |
Description |
|---|---|
|
Economic Objective Function |
The profit/loss of the model determined by running the optimizer solution as a simulation and calculating the difference between the sale of materials and purchase of feedstocks (along with any other costs such as utilities). |
|
Optimizer Objective Function |
The value of the Objective Function for the optimized problem, which is the calculated from the purchases and sales, but also includes penalty terms from infeasibility breakers. |
|
Active Infeasibility Breakers |
The number of active infeasibility breakers in use in the solution. |
|
Infeasibility Breaker Cost |
The penalty cost associated with the active infeasibility breakers. |
|
Model Equations |
The number of equations in the optimization problem. |
|
Model Variables |
The number of variables in the optimization problem. |
|
Model Non-Zero Elements |
The number of entries in the optimization problem matrix which are non-zero. |
|
Model MIP Entities |
The number of mixed integer variables in the optimization problem. |
|
Solution Time |
The time taken to solve the optimization problem. |
|
Total Run Time |
The time taken to construct the problem, solve the optimization problem and return the results to the user interface. The total run time will be affected by the Result Retrieval Settings. Where multi-start is being used, this will be the total run time of the longest running start point, whereas the solution time will be the optimizer solve time for the actual solution result, which is likely to be shorter. Therefore, there may be a significant difference between solution time and total run time, especially in poorly behaving models. |
|
Date of Run |
The date and time of the last optimization run. |
|
Optimizer MIP Nodes |
The number of branch and bound investigations performed during the evaluation of the MIP problem. |
|
Optimizer Iterations |
The number of successive passes of the optimizer needed. |
|
Optimizer Convergence Status |
Whether the optimizer has converged. |
Economic and optimizer objective function values
The Metrics windows show both an Economic Objective Function value and Optimizer Objective Function value. The Optimizer Objective Function value is the value converged upon by the optimizer during a run. This is the final optimizer value and may include penalty costs associated with infeasibility breakers.
The Economic Objective Function value is sometimes known as the Simulator Objective Function value. After optimization is complete, the solution is used to run a simulation (Post-Optimization Simulation), where the feedstock purchases, stream routings, and operating parameters are taken from the solution and applied in the simulation, and the Economic Objective Function value is calculated from the results of this simulation.
Differences between economic and optimizer objective functions
With no active infeasibility breakers and no warning or error messages present, the Optimizer Objective Function value and Economic Objective Function value should match to within numerical tolerance.
Without active infeasibility breakers, a mismatch between economic and optimizer objective can reflect a poorly converged solution.
MIP nodes and entities
The Run Metrics window reports the number of MIP nodes and entities in the problem. MIP evaluation can result from a number of locations:
-
Purchase Count variables
Where mixed integer (MIP) optimization is required the problem will be solved initially using relaxed MIP constraints (that is, without the actual integer constraint(s) applied). If the solution meets the integer requirements without constraining the variable values then it is not necessary to actually apply the integer constraint. However, if the required value is not an integer, it is necessary to perform a branch and bound analysis. This splits the MIP integers and runs the optimization again in order to determine the integer constraint value with the greatest objective function value. This will happen many times down a tree of options.
Example: A multi-base reformer model may have three bases: 98, 100 and 102 RON. The final
solution may be to interpolate between two bases, but these bases should be neighbors
(so either 98 and 100, or 100 and 102). So, in this problem, there are two MIP entities.
In the initial problem, no constraint is applied on the base activity and the final
solution might be 101 RON target. If this is reached using 50% 100 and 50% 102 RON bases then the solution is valid and no further analysis is required. Thus, even though
there are two MIP entities, no actual MIP constraint was required and hence there are no MIP nodes.
However, if the solution is reached through 75% 102 and 25% 98 (whose solution value
is still 101), then the solution is not valid (because 98 and 102 are not neighbors).
In this situation, you would need to run two further analyses with the bases explicitly
constrained to be neighbors: one for 98/100 bases, and another for 100/102 bases.
Thus, now two MIP nodes are required for analysis.