Discretization in IBP Optimizer

[Former Member]

Hello Experts,

I am having some doubts regarding Discretization in IBP optimizer. I have referred to SAP help, but did not get clarity.

https://help.sap.com/viewer/feae3cea3cc549aaa9d9de7d363a83e6/2008/en-US/74f362b10a5e475094e7a1f7416f...

What actually is the meaning of discrete decisions? Say for example - if for minimum lot sizes I set a discrete horizon for 3 months and if for a product min lot size =100 and demand = 50 for the next 12 months what is the impact on discrete horizon? What will happen after the discrete horizon in optimizer run i.e 4th month to 12 month?

The SAP document states "If you set the discretization horizon to n, discrete decisions only take place in periods 1 to n. As a result, constraints are not active in the remaining periods" - what is actually meant by this statement? If the discrete horizon is 3 months, then optimizer constraints remain active till 3 months only, from 4th month onwards constraints will not be considered?

Regards,

Suman.

[keyhani · Accepted solution]

Dear Suman,

adding lot sizes to your scenario means that your problem is a mixed integer linear programming problem. MILPs are harder to solve and require longer run times. To reduce the run time, you have the possibility to specify that only in the first periods of your planning horizon (e.g. 2 weeks) the lot sizes should be considered. For the rest of the horizon, you allow the optimizer to ignore the lot sizes (ignoring the lot size constraint, the other constraints are still considered). This reduces the complexity of your problem and can improve the run time.

You can specify such discretization horizons for transports and productions separately. So you can consider transport lot sizes for the first 4 weeks and production lot sizes for the first 3 weeks of your planning horizon.

In your example, in the first three months, the optimizer will not produce less than 100 per bucket/period. Afterwards (month 4 until the end of the planning horizon), the optimizer ignores the lot size and would be able to produce 50 in a period.

Kind regards,

Mohammad

[Former Member · Accepted solution]

Dear Mohammad,

Thank you for the explanation. It really helped me to understand the discretization impact in optimizer & I was able to reduce the optimizer runtime considerably by tweaking the discrete horizon.

Regards,

Suman