Our company is CEM and we started to study Cp and Cpk analysis on the wave soldering machine's soldering pot temperature.And the results are attached as a document. We could not conclude anything out of it.Can anybody guide on this?
How we can further proceed?

Our company is CEM and we started to study Cp and Cpk analysis on the wave soldering machine's soldering pot temperature.And the results are attached as a document. We could not conclude anything out of it.Can anybody guide on this?
How we can further proceed?
I will leave the SPC stuff to Steve and others who are more skilled then I, However, can you tell me the size of your solder pot 600 lbs, 1200 lbs, etc. Also what style PCB's are you soldering, Through Hole, surface mount? Hole many layers and what is population density?
Or are these static measurements with no product through?
Thanks

Thanks for posting this, it gave me an oppurtunity to think a little.

I would like to ask some questions that might help to understand some things in the data.
What is the time frame between each sub group of 5?
Why in fact do you use a group of 5, what does it give, you are measuring a temperature in a bath and this various with time and of its nature changes all the time. It is not a manufactured part that 5 consequtive parts would give you a measurement of variation.
The differences are expected and the fact that there is such a correlation between the ranges of the 2 measurement systems shows that the connection is due to some external issue, maybe temperature in the room?
The range which which is up to 10 degrees some 4% is very large and this tends to confirm this supposition.
Have you got room tempretaure readings to compare, is the room temperature controlled?

I would assume (this is dangerous) that the manufacturer put his thermocouple in the most effective postion and you put yours in a convenient position. The differences are quite stable and this shows that the controller is not effective as the variations should not be so large.
This shows the agreement of the 2 measuring systems.
It does not matter where you measure or the actual temperature but rather that this is defined in a consistent manner and controlled from this point.

See attached file for some simple illustration

Where is your std deviation of 2.2 from? Is it from all the 110 values? It seems to me to be far to large.

The manufacturing process that you are contolling is an integral system that is controlled by a thermocouple and as such is constantly correcting itself. Why do you need to perform SPC at all?

The important aspect here is the calibration and position of the thermocouple and the sensitivity of the system.


This maybe is a question to Steve and all the other statisticians, is there really any relevence to SPC in a situation like this.

Hello Howe,

Our Solder bath capacity is 250Kgs.We are soldering SMT boards and SMT and THT boards also.
Thanks

Hello Atkins,

The Group of 5 data had been collected in a day and we the machine will be running 8 Hours a day.WE have chosen group of 5 as a default group size given in the SPC. We have chosen the solder bath temperature as a parameter as the temp of the solder bath is the critical parameter in the PCB solderability. Moreover the attached document also says the same. Please review the same.
The point of measurement using the digital thermometer is the same throughout the measurement.I also agree that there will be a variation as the point of measurement is different for the Machine's T/C and Digital thermometer used for the measurement.

The standard deviation is for all the 110 values. As the solder pot volume is more, I feel that so much of variation might have come.

The importance of the temperature is of course not in argue, I was talking more about the use of the statistical method.

For this situation the SPC should be on a single value taken at a fixed time. There is no justifiction for using a group of 5 non connected values and tying to connect them.
What your study has shown that there are very large fluctuations during the day and this is not related to the position of the sensor as it appeared in both measurements.
This is important not the SPC as SPC, you need to find why this variation exists, have you a temperature controlled environment?

What I was trying to say is that SPC for the sake of SPC is not usefull, you have data you have trends, these trends appear to be causing your proces to be out of control try and find the reasons, correct them and then monitor again.
Please keep us informed

Hello Howe,

Our Solder bath capacity is 250Kgs.We are soldering SMT boards and SMT and THT boards also.
Thanks

Hello Atkins,

The Group of 5 data had been collected in a day and we the machine will be running 8 Hours a day.WE have chosen group of 5 as a default group size given in the SPC. We have chosen the solder bath temperature as a parameter as the temp of the solder bath is the critical parameter in the PCB solderability. Moreover the attached document also says the same. Please review the same.
The point of measurement using the digital thermometer is the same throughout the measurement.I also agree that there will be a variation as the point of measurement is different for the Machine's T/C and Digital thermometer used for the measurement.

The standard deviation is for all the 110 values. As the solder pot volume is more, I feel that so much of variation might have come.

Let me ask a few more questions, if I may. Does the technical manual give you any indication on what the optimum solder pot temp should be? Does it give a tolerance for the controller of temp? Somehow there has to be a correlation between your readings and what the manufacturer states. If, for example, manufacturer states solder pot temp to be 240 +/-10 , then you should be able to prove that the machine is capable of this. First prove machine capability then move to processes.
If your readings are saying that the machine is not capable (will not hold temp) then I agree with Howard that there must be a cause(s) and you need to find out what they are.
May I ask if you are having solder problems and if so what are they? I am still trying to understand the problem.

Hi Guys,

I cannot open the SPC on Wave Solder attachment.May i know the file in waht format.

Thanks,

Looks like maybe a xbar/R chart is the wrong subgroup. There is no rationale for the subgroup. Try just a straight x/movR with S=movRbar/1.128.

What will not change is the mean value. This is the most troubling part of the capability analysis. The pot seems to run hotter than the target on a consistent basis. Might want to lower the set point.

Hi Guys,

I can`t open the attachment?may i know what format is the file?

Thanks,

Two things I'd like to share:

1. The digital thermometer readings decrease by time (see first attachment), the solder pot controller serves stable values (tested with linear regression model temp ~ obs).

2. The differences between the values of the digital thermometer and the solder pot controller has two modes, one at 2 and one at 4 (see second attachment). When plotting the difference vs. time there seems to be a break at obs no. 60/64. Perhaps the digital thermometer or the position was changed at this point?

First I would recommend to analyze the cause for the digital thermometer readings. (BTW: Why did you track both readings, from the solder pot controller and from the digital thermometer?)

Second: For valid conclusions drawn from capability indices you have to be sure that the temperature or the grouped temperature is normally distributed. This is not true for the temperature directly, so you could try to achieve a normal state if you're using more data grouped into subgroupes of size 5. There is a trend in the subgrouped data, too, although not signifikant. But with 22 data points conclusions should have be made carefully, because a slight (and afterwards significant) trend probably couldn't be detected out of 22 values.

BTW: Did you analyze the variation of your measurement system?

For further analysis you need more information / more data values.

Regards,

Barbara

It's been years since I wrestled with wave solders. This is bringing back some bad memories. <g>

The only thing I wanted to point out about the original data sheet was that there was no mention of solder level. I used to see biggest variation in pot temperature as a function of the solder level - especially when the level dropped and new bars are added.

So you may want to take note as to when the data is collected versus when solder is added. Has the temperature of the solder bath had a chance to normalize? It's pure speculation on my part, but that could be a possible cause of the bi-modal data.