SPC and Injection Moulding

[Howard Atkins]

This would appear to be another occasion where there is a blind use of tools without reason.
What is the polymer?
What is your quality history?
Do you think it will have any value?

 

[Jim Wynne]

I would just say to set the context that we are producing since 5 years the product without any SPC on...

2nd thing is that personnaly I would say that since a specification criteria (for example dimensional specification) got historically a very good CpK there is no interest to make any SPC as this means that the process is well capable...
Thanks

How can you determine the Cpk and that "...the process is well capable..." without having done some sort of SPC?

 

[Palt88]

This would appear to be another occasion where there is a blind use of tools without reason.
What is the polymer?
What is your quality history?
Do you think it will have any value?

Hi Howard,

The polymer is a simple ABS plastic.
The quality history is a capability calculation on each lot samples.
I have a doubt of any added value of a SPC on a well capable process...

 

[Palt88]

How can you determine the Cpk and that "...the process is well capable..." without having done some sort of SPC?

Hi Jim,

I don't understand what you mean.
The capability is calculated with the data's that we are collecting through all the years.
SPC didn't give you a capability value?

 

[Jim Wynne]

Hi Jim,

I don't understand what you mean.
The capability is calculated with the data's that we are collecting through all the years.
SPC didn't give you a capability value?

What have you done with the data? How have you used it to determine the capability of your process(es)?

 

[Palt88]

What have you done with the data? How have you used it to determine the capability of your process(es)?

The dimension of the components are measured through sampling.
So we just generate CpK with the data's against a specification...

The requirement of the customer is to have a certain level of CpK.

Now the customer starts to speak about SPC introduction in the moulding process.

If we are dealing with specification that are always over a CpK of 1.33 I personnaly don't understand the interest to "control" that through a SPC.

I would better understand that on fluctuating value with bad capability?

 

[Jim Wynne]

The dimension of the components are measured through sampling.
So we just generate CpK with the data's against a specification...

The requirement of the customer is to have a certain level of CpK.

Now the customer starts to speak about SPC introduction in the moulding process.

If we are dealing with specification that are always over a CpK of 1.33 I personnaly don't understand the interest to "control" that through a SPC.

I would better understand that on fluctuating value with bad capability?

Have you verified that the data are normally distributed and statistically stable? You need to do that before calculating Cpk. If you want to convince your customer that SPC isn't necessary, you'll need to provide some sort of statistically valid evidence to support your point.

 

[Palt88]

Have you verified that the data are normally distributed and statistically stable? You need to do that before calculating Cpk. If you want to convince your customer that SPC isn't necessary, you'll need to provide some sort of statistically valid evidence to support your point.

Do you mean the capability by sampling lot or by cavity ? (this is a 32 cavities mold)

By sampling lot the normality is ok (cavity #1 to #32 gives normal distribution) but by cavity it is not (I assume it normal due to raw material and moulding parameters variation between lots...)

But anyway a CpK could be calculated even if it's non normal distribution by using appropriate distribution that fits datas.

 

[Jim Wynne]

Do you mean the capability by sampling lot or by cavity ? (this is a 32 cavities mold)

By sampling lot the normality is ok (cavity #1 to #32 gives normal distribution) but by cavity it is not (I assume it normal due to raw material and moulding parameters variation between lots...)

But anyway a CpK could be calculated even if it's non normal distribution by using appropriate distribution that fits datas.

If you have data that should fit a normal distribution but doesn't, you will create more problems by cramming it into an inappropriate distribution. You need to talk to your customer and (with rational data in hand) explain why you think ongoing SPC is unnecessary.

 

[Proud Liberal]

Two comments:

1. Each cavity must be evaluated separately. Mixing streams of production will only confound the data. The delta Xbars between cavities will impact the evaluation of sigma.

2. Since injection molding is an auto-correlated process, the use of Xbar/R charts are invalid as the premise of independence has been violated (again sigma will not be accurately estimated). The more appropriate tool should be IX/MR.