Mold Changeovers: Sequencing Injection Molding Work

Injection molding scheduling changeover has to model two different things at once: a tool swap that costs about the same every time, and a purge whose cost depends entirely on what ran before. A schedule that stores one average setup number gets both wrong. EDGEBIC by User Solutions prices changeover per press and per product pair, so the plan charges 20 minutes when you run light after light and charges the real number when you have to bring a press back from black.

For the mechanism described without the plastics framing, see the setup matrix explained. For the category-level view of this sector, see plastics and rubber production scheduling. For the map of which capability carries the load in which plant, start at how different industries use EDGEBIC.

The two halves of a press changeover

Ask a molding supervisor how long a changeover takes and you will get a question back: from what to what? That question is the whole feature.

The tool swap is broadly sequence independent. Unclamping, craning out, craning in, clamping, connecting water and hydraulics, first shot. A given press does this in roughly a fixed time, modified by whether the tool is on quick-change clamps.

The material and color transition is entirely sequence dependent. Natural to natural is nothing. Natural to a light color is a short flush. Light to dark is modest. Dark back to natural or a light color is the expensive one, sometimes by an order of magnitude, because pigment clings to the screw and the hot runner and shows up as streaks for shots after you think you are clean.

A flat setup field can hold the first number or an average of both, and either way it misprices most days. Underestimating a dark-to-light transition means the plan promises a finish that the floor cannot hit. Overestimating a same-family run means the plan hides capacity that the press actually has.

How the matrix holds both

EDGEBIC stores changeover as a lookup keyed by work center and by the pair of products involved, in minutes. Three layers make it maintainable.

Setup families group products that change over identically on a press. In molding the useful grouping is almost always resin plus color band rather than customer or part family. A working set is Natural, Light, Medium, Dark, and Specials for glass-filled or regrind-restricted material.

The family matrix is the per-press grid of from-family to to-family minutes. This is where roughly 95 percent of behavior lives. What a setup family is covers the concept on its own.

The product matrix holds specific from-product to to-product pairs that break the family rule, and it always wins over the family cell. One pigment that fights every light color afterwards gets a row here instead of forcing an extra family.

Crucially, a matched cell replaces the routing's flat setup rather than adding to it. That is the single most important practical fact when you load your first matrix: your cells hold the total changeover, tool swap included. If the swap is 45 minutes and the light-to-dark purge is 20, the cell holds 65. Put the breakdown in the cell's notes field so the reasoning survives.

A worked press grid

Press-2, quick-change clamps, tool swap 45 minutes. Five families. Values are total changeover minutes including the swap.

Read the asymmetry across the diagonal. Natural to Dark is 65. Dark to Natural is 150. The diagonal is the tool swap alone, because a same-family run needs no purge. Same product back to back returns zero automatically without any cell at all, so a repeat order on a press that is already set carries no changeover.

Twenty-five cells cover every combination of every part that runs on this press, however many part numbers that is.

Press-7, a twenty-year-old machine with a cold sprue and no quick clamps, gets its own grid with a 90-minute diagonal and longer purge times throughout. Same products, different physics, different numbers, and the scheduler charges whichever press the job actually lands on.

How the engine picks a number

For every operation, the engine finds what that press ran last and walks an ordered chain. First match wins.

Rule 1 catches people out. A cold start is not a free changeover: the press still needs its initial setup, so the routing's flat number applies. Only rule 2 returns zero.

Rule 3 is the escape hatch, and the documented example is exactly the molding case. Red belongs to the Light family, so Light to Dark would normally price at 60 minutes. Switching specifically from Red to Black needs an extra solvent flush, so a product-level cell holds 90. The engine hits that cell at step 3 and never consults the family grid. The stored reason on the scheduled operation reads: product matrix Red to Black, 90 minutes.

What honest changeover data does to a schedule

The published worked case uses a paint booth, and the arithmetic is the same shape as a press. Three jobs, one machine, one 8-hour shift.

With a flat 30-minute setup the plan showed a 12:45 finish while the floor needed roughly 11 hours, because two 240-minute reverse changeovers were never charged. With the matrix loaded and the same due-date order, the plan charged 330 minutes of changeover and honestly showed the work overflowing to 16:45. With the matrix loaded and the jobs campaigned like to like, changeover fell to 90 minutes and all three finished by 12:45 with three hours to spare: a 73 percent reduction from sequencing alone.

That is the two-step benefit. The matrix buys honesty. Sequencing buys capacity. A scheduler with no matrix has no basis for preferring one order over another, so the second step is unavailable until the first is done. The full six-job version of that walkthrough is in the paint-booth math, and it reads directly onto a press queue.

Reschedules and what really ran

Molding plants reschedule constantly, and the changeover question follows the job.

When a schedule is rebuilt, the engine anchors the "what ran last" question for each press to the last product that actually completed there, taken from logged actuals rather than from a planning assumption. If black genuinely ran on Press-2 this morning, the next job's changeover is priced against black even if the previous plan had intended something else. Completed operations keep the setup they were charged, because that is now history rather than a projection.

This matters more than it sounds. Without the actuals anchor, a reschedule that clears and rebuilds all future allocations would evaluate the next job as a cold start and quietly under-charge the very transition that is about to cost you two hours.

Five mistakes that show up in the first week

Each of these produces a schedule that looks fine and is wrong, so they are worth checking deliberately.

Only one product in a pair has a family. The family lookup needs both sides. Assign one and leave the other unassigned and the engine falls through to the flat routing default with no error. This is the single most common cause of "the matrix is not working".

Cells hold only the purge. A matched cell replaces the flat setup rather than adding to it, so a cell holding 20 minutes of purge quietly deletes the 45-minute tool swap from the plan.

Zeros entered down the diagonal for identical products. Unnecessary. A genuine repeat of the same product returns zero automatically. Those rows are dead data that someone will later mistake for a rule.

A family deleted while products still reference it. The system blocks the delete rather than orphaning the assignments, so reassign first. Deleting a family also removes the family cells that referenced it, and any product-level overrides survive: check whether they still make sense on their own.

Negative or joke values. Negative minutes are rejected at entry and clamped defensively during resolution, so a typo cannot produce a changeover that gives time back. A typo of 6 minutes instead of 60 will not be caught, though, which is why the first-week comparison against actual changeovers matters.

Rolling it out

1. Pick your two most contended presses. Not all of them. The presses that decide your ship dates are where the payback is.
2. Define four or five families by resin and color band. Resist modeling customers or part families: the press does not care who the part is for.
3. Assign every part that runs on those presses to a family. Anything genuinely unique stays unassigned and gets product-level rows.
4. Fill the grid with total changeover including the tool swap. Get the numbers from the people who do the changeovers, not from the routing file.
5. Add product overrides for the pigments and materials your team already warns each other about.
6. Bulk load if it is faster. The matrix exports to CSV and imports back, so a grid built in Excel goes in as one operation. The build guide walks the whole sequence.
7. Verify after the first run. Add the Setup Source column in the Job View. Rows showing a matrix source used the sequence logic. Rows showing the routing default did not, which almost always means one of the two products has no family assigned. The family lookup needs both.

Where it fits with the rest

Changeover is one of three constraints most molding plants have to model. The second is which press a job can run on at all, covered in routing molding jobs across a press pool. The third is how a long run spreads across shifts and machines without double-booking, covered in running molding around the clock. Changeover pricing and press pooling reinforce each other: once each machine carries its own matrix, the pool selection can prefer the press that is already set for the family you are about to run.

Bring one press, its real changeover times and a week of orders to a demo of plastic manufacturing scheduling software, and we will build the grid with you.