Doubt about the correct control chart

[buralbe]

Hello everyone!

In our manufacturing (Electronics) we would like to introduce a control chart to monitor the reflow oven parameters. The parameters to monitor are temperature and time
In specific: Soak Time, Peak Temp, Cooling time.

Currently we use a profiler to measure the parameters. The profiler uses 3 probes to measure the temperature and we have therefore 3 measures to fill the control chart.
The profiler is done once a month.

Which is the correct control chart? I-MR or X-bar / R-chart?

Thanks!

 

[Beth Savage]

It boils down to your subgroup size for each characteristic.

With an I-MR chart, each plotted value on the I (individuals) chart represents one measured value. Each plotted value on the MR (moving range) chart represents the difference between successive individual measurements.

If you are collecting one measurement for each characteristic (Peak Temp, Soak Time, Cooling Time), the I-MR chart would be appropriate.

With an X-bar/R chart, each plotted value on the X-bar chart represents the average (or mean) for a subgroup of data. The subgroup (or sample) must contain two or more measured values. Each plotted value on the R (range) chart represents the range (highest minus lowest) for the subgroup of data.

If you are collecting multiple measurements for each characteristic (e.g. 5 samples each for Peak Temp, Soak Time, Cooling Time), the X-bar/R chart would be appropriate.

 

[Steve Prevette]

I agree with Beth above - great post. I will say my personal preference is I-mR, which Dr. Don Wheeler is a big proponent of. But since you do have three readings of the same parameter - it is a great opportunity to see if they are consistent between the three readings and detect if one is drifting.

 

[Miner]

How are you using the three probes? Are these measuring different sections of the reflow oven? Are you recording a literal profile as it passes through the oven with the three probes in a right/center/left configuration such as the 6-probe example below?

 

[Bev D]

It boils down to your subgroup size for each characteristic.

Mathematically that is correct. Theoretically and practically that isn’t the best answer. The first step is to understand the variation and it’s homogeneity - lack of it. We rationally subgroup to place the random/homogenous variation within the subgroup. Then we can determine the bets sample size and the best control chart.

I have never used a control chart of temp profiles mostly because the variation is not all that random. My suggestion would be to really think about how the process varies and how it can change or go wrong. Plotting the data over multiple runs and looking at it will help. I think inner is going in that direction with his questions…

 

[Miner]

I think Miner is going in that direction with his questions…

I am indeed. The OP's answer will have a big impact on the type of chart, or even if a chart is appropriate.

 

[buralbe]

Thanks to all of you first of all!

To answer Miner: The three probes are placed on a PCB sample that is travelling through the oven. The probes are placed at different positions on the PCB (I am not sure where exactly).

 

[Steve Prevette]

I'd suggest try it multiple ways (ImR, xbar-R) and see which gives the best picture. Also, if you've had a known fault in the past, see what chart detects it. Likely they all will, but some may be better than others. You may want to try ImR on each individual temperature.

 

[Miner]

To answer Miner: The three probes are placed on a PCB sample that is travelling through the oven. The probes are placed at different positions on the PCB (I am not sure where exactly).

So the probes are either placed randomly on the PCB, or strategically on components with a high thermal mass. Either way, the variation between probes will consist of component to component differences + measurement variation. Within oven variation is excluded due to the relatively small size of a PCB.

I think both IMR and XbarR charts are an option, but this leads to the next question. How do you plan to collect each sample? Do you intend to take one sample at a specific time/location in the oven each time you run the profiler through, or do you intend to take multiple samples from one profiler pass through the oven? The first option would track oven temperature variation in the same location over time. The second option would track oven temperature variation over distance (location) in the oven.

 

[buralbe]

To answer to Miner: To goal is to track the stability of the oven temperature over time and we have therefore decided to take only one measure from only one of the 3 probes. As you pointed out, we don't want to check a specific PCB profile (and placing the probes strategically on components with a high thermal mass). Further, we intend to take one sample once a month.
It seems that IMR chart is the best one to detect points out of the limits.

Now, my next question is... When do I need to react with a corrective action? At every single point out of the 3-sigma limit?