Whirlpool Appliances Mixed Model Scheduling

Summary

Whirlpool Appliances was in the process of upgrading their legacy technology solutions to the Ignition SCADA platform. Many of their facilities have complex workflows consisting of multiple upstream part production machines feeding the assembly line.

The upstream production machines are set up to produce a given number of parts with various parameters like paint colors, and the assembly line needs to produce a specified allotment of finished appliances of various brands based on customer demand. To meet demand requirements the assembly line is fed from different upstream machines due to cycle times on those machines. This means running 1000 blue units then 1000 white units would starve the assembly line of parts.

To remedy this they utilized a “mixed model scheduling” tool with a subset of rules set up in their legacy platform. If there were 3 upstream machines with parts A, B, and C each identical except for paint color this allowed them to feed the assembly line in an ABCABC fashion, or many other options. If say 2 of the machines were set up with the same paint color then they could feed in an AABAAB fashion throughout the shift, or more complex mixes as needed.

Additional complexity in the mixed model scheduling system came from situations where other upstream raw materials like wiring harnesses, user interface screens, motors, etc. were fed to the line allowing them to produce different SKUs in assembly.

Benefits

• The ability to define complex mixed model scheduling for any given production line with dynamic upstream equipment configurations and production schedule rules

• Real-Time updates to the mixed model scheduling algorithm based on actual process data from upstream equipment and the assembly line to account for production slowdowns and downtime

• Mixed Model Scheduling fully integrated into the Ignition SCADA platform used to run the entirety of any given facility

Solution

Corso Systems was brought into Whirlpool’s Process Controls group to help solve this mixed model scheduling challenge.

We first worked with the production schedulers to understand all of the various rules they might apply to the legacy system when generating a mixed model schedule. This gave our team an understanding of the potential scheduling scenarios we needed to account for, and a framework for validating the schedule our system produced against real-world data before it was put into production.

Once we understood the overall scheduling framework we dove into development of our mixed model scheduling system.

The foundation of the scheduling system was built using Machine Learning tools in Python. This allowed the scheduling system to interact with the work orders for any given shift in the ERP system, then give the schedulers a screen in Ignition they could use to define the rules they wanted to use for any given production run.

Once the schedule was generated Ignition fed real-time process data back into the scheduling algorithm allowing the mixed model scheduling system to update based on actual production conditions, adjusting if there was downtime or slowdown on either the upstream machines or the assembly line. This accounted for slowdowns due to end of line issues if there was a backup in packaging or the automated testing line.

This system worked very well for Whirlpool, expanding greatly on the capabilities of their legacy system, and was soon rolled out to the various production lines requiring mixed model scheduling.