Everything varies.
Your vendors provide components with a range of values. Your production process varies, too.
Creating, monitoring, and maintaining process stability enhances your product reliability performance.
When I started my professional life as a manufacturing engineer, a senior engineer told me we take a product design and can only make it worse.
He said if we could make every unit exactly according to the nominal values of the drawing, every unit would work well.
This made sense to me as I’d been analyzing data gathered from every unit produced.
We had just installed a new measurement tool to an early stage of the production process.
The tool measured the resistivity of the core of the product (a heating cable).
The target for a product may be 500 ohms. There was not a flat line of readings at 500 ohms.
The reading bounced from 423 ohms to close to 600 ohms.
Sure the bulk of the readings were just above 500 ohms, yet a quick scan of thousands of reading did not find any exactly at 500 ohms.
The trick to process stability
This often eludes many.
You have to measure something that actually relates to the final product’s performance.
It is possible to measure many elements of a product in production, not all of them are important.
Start with the FMEA or a discussion with the design engineers. What are the critical elements of the design?
What has to be right on the specifications to best ensure the final product will work?
Sometimes we have to measure a surrogate or something that relates or predicts the final performance.
When we find and start measuring an element of the product that is the source of the final product variability.
We’re about to learn something- that not all products are the same.
They should be, and they are not.
Plan, do, check, act
Deming and others have long advocated process improvement based on data. So, given our measurements, we have data.
Now we need information.
A common approach is to prepare an experiment or adjustment concerning the process or materials.
In essence, this can be as simple as changing something, like polymer blend drying time from 30 minutes to 45 minutes. Do not just run down to your production floor and start changing things and see what happens.
That is a recipe for chaos.
Instead, think about what should occur and why.
Create an experimental plan. Prepare to collect sufficient data to allow your experiment to create clear results.
The sample size or run time is part of the experimental planning and using the appropriate statistical tools.
Run the experiment and collect the data. Do the analysis.
Check the results. Often with another run of the experiment or pilot run of production. Double check. Did the reduction in variability occur, or did the shift of the mean occur, or both?
Act. Implement the changes and monitor. At this point, I find we have other experiments to run as we notice other opportunities for improvements.
Statistical Process Control (SPC) and capability analysis
The measurement step is critical.
Once a meaningful measurement system is in place and the product is meeting specifications (mostly), we need better tools than a series of experiments and adjustments.
SPC tools allow us to efficiently monitor the variability of a process. SPC alerts us to changes (good or bad) soon after they occur.
If the process is stable, SPC will provide the evidence (the data) that the process is stable.
Once stable, it’s time to conduct a capability analysis.
This tool provides design and manufacturing engineers the information they need to create new products that the current process has the ability to actually produce well.
Summary
PDCA, SPC, and capability analysis are common tools to assist you to create products that perform as expected.
No two individual products from your assembly line are the same.
Different is okay within reason when it comes to product performance.
Measure, adjust, improve, and monitor your processes. Understand and correct causes of unwanted variability.
Keep your process under control and your customer’s happy.