Ferry Manufacturing Snippets: 13th May - 19th May 2024

Here's the weekly digest of Ferry's manufacturing snippets!

Monday - 13th May

Slight variations in factory temperature can have a significant impact on manufacturing performance.

Particularly in quality control

We've heard many stories from seasoned operators who know from experience that quality rates are higher in certain seasons or times of the year, or in sites that reside in warm but dry regions.

Often it's to do with how machinery interacts with product & packaging. Maybe that pouch doesn't open as well when it's colder & more rigid. Maybe that food product clumps together under certain humidity conditions and jams the feeder.

It is often qualitatively noted but not quantitatively measured.

Which presents a potent source of efficiency gain if tracked over time, and process adjusted to match

Tuesday - 14th May

When people in manufacturing talk about AI they often think about vision in quality control

Namely, visually detecting defects during manufacturing, sorting product according to dimensions for allocating to downstream processes etc etc.

But it's so much more than that:

• Visually identifying breaches of safety to protect the frontline workforce

• Tracking how processes are performed (& how operators interact) for workforce training

• Crunching through data to find trends & correlations for identifying process improvements

• Automating report generation to reduce manual work

• Translating communications for multi-lingual teams

• Machine monitoring for predictive maintenance

• Enhanced production scheduling for optimal resource allocation

Often we won't see it, but AI will be embedded deep into the solutions that folks use on a daily basis

Wednesday - 15th May

In large manufacturing enterprises, it's quite common to have duplicate systems.

We've seen cases where manufacturers have dozens of "MES" applications (manufacturing execution systems) distributed throughout the company, sourced from different vendors

Sometimes this is due to acquisition, where companies get bought & consolidated with heterogeneous setups.

More often, these organizations practice decentralized decision-making, where individual plants can make separate decisions on what they want to use.

The inevitable challenge is this makes the job of unifying, standardizing and aggregating information from sites up to HQ very tricky.

One approach is to migrate every site to a common system standard. Doable, but brutally expensive and time intensive.

The other approach is to build an orchestration layer. Which feeds into systems to create a unified view of what is happening in production across sites & divisions, aligning teams through a common understanding of operations.

Faster to implement, cheaper, and gives the organization flexibility in the long run (i.e. whether to migrate site-by-site or keep a decentralized IT model)

Thursday - 16th May

There's a common saying that manufacturing data is inherently relational

That is, you can structure data in tables with rows & columns representing entities that can be joined together to form relations. Like batches of production being linked to a particular product / SKU

Yes, but....

There is an entire universe of unstructured data that is absolutely necessary to parse in order to truly understand the levers behind manufacturing.

One such good example is pdfs of product characteristics for SKUs (i.e. bag types, sizes, sealant form etc) that you can use for correlations against production performance.

Another is recorded notes & transcripts from operators can provide context behind production events.

It transcends the written word too. Video recordings can uncover deep insights into how operations are actually performed.

Relational data gives you the baseline to monitor performance.

Unstructured data unlocks the context to make it valuable & actionable