AS/RS throughput stuck? Often it isn't the equipment. The same pallet keeps making the trip

Take one pallet of a common part. It comes out of the AS/RS, gets picked once at the outbound sorting station, and goes back. A little later the next order needs it: the stacker crane fetches it again, the AGV delivers it again.

All day the equipment is busy, and throughput still won’t go up. We later changed this piece of the wave logic, and overall efficiency rose by about 20%.

This post covers three things: how order grouping differs from wave building, how waves are set, and where the 20% came from.

1. First, order grouping and wave building are two different things

People often use the two words interchangeably. They sit at two different levels:

• Order grouping happens at the order level: before production orders are released to the WMS, orders of the same type that share common parts are grouped together. A common part is one that several orders all need.
• Wave building happens at the warehouse level: once orders are grouped and released to the WMS, the WMS splits the outbound tasks into waves by the orders’ number of units, common parts and outbound sorting station.

Order grouping decides which orders move together; wave building decides which pallets the warehouse pulls this round, and to which sorting station.

I wrote earlier about a site that asked, right after go-live, to switch automatic order grouping off and go back to grouping by hand. That was an order-level problem: the plan and the shop-floor progress did not match. Today is about the warehouse level: the orders are fixed, so how does the warehouse pick them out most cheaply?

2. How the WMS sets waves: by order, by time window, by sorting station

Our waves are set by three things:

• Basis: production orders, plus a time window; everything due in that window is counted together
• Grouping: by outbound sorting station, each station its own group
• Size: how many units per wave, then the part mix

The time window can’t be too long or too short. Too long, and a wave gets too big: the sorting stations and the lineside overflow. Too short, and waves fragment: the equipment shuttles between several waves.

How big a wave should be looks like one parameter in the WMS, but its root is in the PFEP table. The PFEP is the table with one row per part: which pallet it uses, how many go on a pallet, how many one unit uses. I wrote earlier about cutting that table from 30-plus columns to 8.

Pieces per pallet and quantity per unit decide how many pallets a wave pulls and how much each pallet gives up. If the PFEP is off, pallets end up half full, and the same part takes more pallet calls. Get the PFEP straight, and you get fewer half-empty pallets and fewer pallet calls. However you tune the waves, you can’t get around the PFEP.

3. The 20% from AS/RS outbound: once a pallet is out, serve everything before it goes back

The original logic was simple:

The stacker crane takes a pallet out of the AS/RS, the AGV brings it to the sorting station, this order is picked, the AGV takes it back, and the crane puts it away. The next order needs the same part, and the whole loop runs again.

Common parts suffer most, because several orders need them. One pallet might go in and out several times in a morning, and every trip ties up the crane once and an AGV once.

We changed that loop: once a pallet of a common part is out, the system first checks whether, as things stand, it can serve every sorting station’s need in the current wave; only after it has, does it go back to the AS/RS. Not back after every single pick.

To do that, the system has to work out two numbers before release: how much of this part all the orders in this wave need, and how much is left on this pallet. If the first is not more than the second, the pallet stays in the sorting area until picking is done.

Crane and AGV calls went down, and overall efficiency rose by about 20%.

But there is a trap: rush orders. A wave is built from the orders known at the time. When a rush order cuts in, a part that just went back gets called out again, and the wave breaks into pieces. The more rush orders, the more of that 20% you give back.

---

When a rush order cuts in, do you break up the current wave and rebuild it, or open a separate wave?