CS-01

CS-01 · Visy Glass · 2020 to 23

Capacity recovery on a 24/7 glass line

A glass furnace is lit once and runs for years. Nothing stops so you can try something. Every step in this programme was taken on live production, with orders due and the rate being reported.

Result
~100k extra containers per year vs benchmark
A$500k recovered capacity, annualised
4 SKU projects in the programme
Role
Production Line Manager
Employer
Visy Glass Packaging, Melbourne · glass container manufacturing
Tenure
May 2020 to February 2023
Team
18 operators across three shifts
Shape
Four projects, one per SKU, delivered adaptively
Operation
24/7, continuous furnace, no maintenance window
Constraint
Process stability, pull rate, forehearth draw, machine cooling, conveying, palletiser
Status
Still running this way, to my knowledge

Context

Everything downstream of the furnace runs 24/7 across three shifts. An idle forming machine is molten glass with nowhere to go.

I managed one line and its eighteen operators. Any change I wanted had to be understood and executed the same way by three crews, on equipment that could not be stopped to explain it. That part transfers to project work.

The constraint

Run rate is several dials at once. Push the speed on a bottle design and you are testing process stability, pull rate, forehearth draw and machine cooling together, then conveying and the palletiser downstream. Any one can bind, and which one changes with the design. The question was the sustained rate at which all of it still held, without trading away quality or safety.

What I did

Four projects, one per SKU, and keeping them separate was the decision that made it work. Operating limits, quality response and equipment behaviour are specific to a product; bottle geometry and handling differ, and what you learn on one design does not carry to the next. One "line speed" initiative would have averaged four problems into one wrong answer.

Each ran adaptively, because the right operating point cannot be set on paper. Raise the rate a step. Hold it long enough for real production results. Review output, defect rate, downtime and equipment stability, get the operators' read, then decide whether to go again. Quality controls did not move. Extra containers that are rejects are worth nothing.

The discipline was knowing when to stop. Where a step caused instability or a quality concern we pulled the parameters back. A rate you can hold for an hour is a spike. A rate that survives a full job is capacity.

Result

The number is recovered capacity across the four projects. These jobs run a set number of days a year; at the higher sustained rate they produce roughly 100,000 more containers a year than the benchmark would have. At the plant's value per container that is above A$500k a year, and it meant the site could meet demand with less reliance on interstate supply.

To my knowledge the line still runs this way. That matters more than the number. An improvement that outlasts the person who made it is a different claim from a good quarter.

The before and after rates are the plant's numbers and stay off this page. The aggregate is mine to stand behind and I will walk through it in an interview.

What I'd do differently

Two related things. I moved too fast, and I did not consult enough of the people who knew parts of the process better than I did. The result was calling a problem closed because the symptom had gone, before the person who understood that part had confirmed it.

The fix is a rule. With this many interacting constraints, a change is closed when the owner of that part of the process says it is. Now I name who has to confirm what before I start.