See our Art’s Tips to find earlier Tips of the Day.

See our Art’s Tips to find earlier Tips of the Day.

The attached paper came out of some discussions with customers about “how do you do a DOE?” In one case just a few minutes before he was starting one.

What is “D.O.E.?”

The acronym stands for “Design of Experiments” (“Doing of Experiments according to Bob Launsby). In the manufacturing context this means making process changes and measuring the results to see what process changes cause changes in the manufactured product. In our context I quote from the paper that a designed experiment means “Figuring out which of the four cavity variables to force in order to change the parts, how to force them, working the plan (doing the experiment), measuring the parts, finding correlations and setting alarms.”

In the past we have sometimes offered education in this department. Some of that education was a bit general and did not focus on plastics variables or on using the eDART to do an experiment. In any case this is not a regularly scheduled at this time.

So I got to thinking: How can I make a sort of one page crib sheet that would explain the essence so that people could get an experiment designed properly and not foul it up and have to do it over? Mike Groleau reminded me of much of the core. We tried it out when Melanie was teaching here and corrected some items. It comes to one page – front and back, though. It does not cover all of the off-line analysis (that is in the Alarm Settings Document and the Auto Correlator document). It has two additional pages with diagrams to help promote understanding of some concepts.

How to use the paper

This is by no means the definitive word on designing experiments. Rather this is a memory jogger to get you or a customer thinking about what steps to take to make sure that good in-cavity variables and values drive part containment. It is also an aid in not forgetting something such as starting the experiment and finding you have no way to label the parts.

If you are interested in doing a DOE go through each step in the planning and place a tick mark next to each step as you do it.

Since the annual sales meeting last March when this document was created I have been adjusting and adding to it. I thought it was time to send out the latest copy. This includes several things I have learned by doing. You can discard the old tip or document and replace it with this one.

First is that the word “Designed Experiments” tends to be reserved for a certain branch of the science of which our process is a subset. So I have renamed the document “Finding Correlations to Part Quality with eDARTs”. Some of the things I have added are as follows:

Under “1. C. Which plastics variables most likely affect those part characteristics?”

• Melt temperature and mold temperature are important in crystalline.
• Weight: pressure including packing and then pressure discharge after pack or at end of hold.
• Note: Time constraints may limit the number of variables

We did a “real” correlation study for a customer’s part and discovered some more practical things that I had not noted before.

• Setting the limits on values to change may require experimentation. For example, if you are trying to correlation to dimensions and you already know a part with sink or flash is a bad part then it makes no sense to push the limits of pressure out to where the parts cannot even be measured. Also it does not make sense to go so far that you exceed your machine’s capabilities or the capabilities of the machine on which the parts will eventually be made. There is a new section on setting practical limits.
• Plan part handling very carefully on a part with fast cycles or many cavities. You can have so many parts coming out of the machine that labeling them or even finding a place for all of them can be surprisingly difficult, especially in the thick of things.
• I added some wording about setting up the “Sequence Settings” to get good fill times and integrals.
• Process stability is another important point added to the new document: Getting the process stabilized in the first place, making sure the baseline (center) has not changed and also not stopping in the middle to make machine adjustments or label parts.

This has been an ever evolving document. The above are a few of the items included in the new one. While it is quite dense and takes some heavy reading and thinking, it contains a great deal of our practical knowledge on the subject. I hope it helps.

If you are not using the Summary Graph and want to add the Part Sampling tool to the eDART’s Main Menu then you can drag it onto the System space in the Architect. This works for any eDART tool that you might want to add to the main menu or remove from it (by dragging it off). I use Part Sampling as an example because the customer could not get the Part Sampling tool to start from the Summary Graph menu.

Adding Part Sampling to the Main Menu

First, open the Architect from the Main Menu like this:

Then click the System Tab on the left and the Production tab on the right. Drag your chosen tool into the blue working space. It does not matter if it overlaps tools already there.

    

Close the Architect. Click the Main Menu button on the toolbar once, then minimize it and then click it again. The Part Sampling tool should show up on the eDART’s main menu and you should be able to start it from there.

Returning to Part Sampling and Correlation…

One of our customers could not get the Part Sampling tool to start on a validated eDART (October 23, 2003 software release). We’re not yet sure why. It is supposed to be there. I wrote some help to him to get him going since he had a big DOE starting with a bunch of people looking over his shoulder.

This brings up the subject of what to do about part sampling if you forget to use the Part Sampling tool to mark the parts. We recommend that you always use it because it tells you exactly which part to take from the machine and lists the time stamps as each is sampled. Sometimes people forget and run through several runs and then have to go back and sort of retro-sample the parts. This can be done either on the eDART’s Summary Graph or on the Analyzer.

Or course you MUST record (on paper or a bag of parts) the shot time stamp at which each part group started. So here are the steps in brief:

• The shot time stamp on the upper right corner of the Cycle Graph updates at the end of Injection Forward. Record this time at the start of the group and at each sample that you bag up afterwards.

  Remember to wait until Injection Forward goes OFF. You have until the next injection forward goes off to get it written down.

• Enter a note on the summary graph at the shot time stamp at which the group of samples began. This is exactly what the Part Sampling tool does for you (as well as file breaks and rejecting parts – see Tip #34).

---

Part Sampling Entry without Part Sampling Tool

If you forget to run the the Part Sampling tool or it is not in the system or will not run then enter the sample groups manually on the Summary Graph as notes. Here’s how.

1. Take note of the exact shot time for the parts that you actually removed from the machine. This is the shot time in the upper right corner of the cycle graph. It updates with the new shot time AT THE END OF INJECTION FORWARD.
2. At some convenient time enter a note on the Summary Graph at that shot time. First position the cursor there and read the time stamp at the top. Then click “Notes” / “Add Note”. Verify that the time at the top of the note matches the time at which you first began taking parts from that run (which we call a “group”).
3. To make a part entry the first line must be typed like this (use proper case and the colon):

        Start Group: N Shots

        Where you replace the N with the number of shots you have in the group.

4. The second line in the note is the name of the group; e.g. Hi H – Low S – Low T
5. You can add more comments to the rest of the note just like any other. For example you might put the actual hold pressure or speed level on the third and fourth lines. Of course the eDART records a lot of those settings in the data automatically (e.g. Average Value / Hold Pressure and Average Value / Fill Flow Rate).
6. Save the note. Ensure that it gets saved (and does not have the “Not Saving Data” message). Here’s a sample:

You can also enter these notes later in the Analyzer on a windows computer if you do not have time to do it at the press. Just be sure you carefully mark the parts (bag them) and record the exact time stamp for each sample.

It came to my attention today that not everyone has got into the loop on how we correlate part measurements to in-cavity data. So I thought I would forward the two key documents (Auto Correlator User Guide, Alarm Settings) with a little discussion on how it is done and why. Not everyone was here for the quick overview at the last sales meeting.

While this may seem “advanced” you should think of it as another run on the value ladder. It does indeed require more learning and effort. But can raise the capability of the molder to deliver correct parts for a whole range of specifications, not just “full and burgundy”.

First: A Little Philosophy

Remember that RJG is all about measuring in-cavity data to detect bad parts. We know that a great many part quality problems can be found with in-cavity measurements. These measurements usually correlate to part characteristics.

A customer once suggested thinking of part characteristics in three levels:

Level 1: Shorts, Flash and possibly warp – characteristics that you can see with the naked eye, usually as soon as the part comes out of the machine.

Level 2: Dimensions, warp, textures – characteristics that you need to measure in a lab; often these require some stabilization time and special jigs, fixtures or equipment to measure.

Level 3: Impact resistance, tensile or compressive strength, chemical resistances – characteristics that often require destroying the part being measured. Hence you often cannot post-test every part because you would destroy them all.

Level 1 alarms usually do not require the sampling of parts and use of the autocorrelator. Usually peak cavity pressures and cycle integrals will indicate shorts and flash. The customer can just push his process up and down, forcing cavity pressures with the machine until he gets bad parts. He then checks the levels for peaks and cycle integrals and sets the alarms inside the levels at which the parts can go bad. He can then “tune” them later as need (a topic for another day).

Level 2 and 3 alarms are particularly suited to the post-measurement of parts and correlations to in-cavity variables.

Using Parts Measurement and the Auto Correlator

The attached documents describe this in detail. Here I will just list the 10 essential steps.

1.       Plan the experiment to “push” the four plastics variables as needed for the particular part.

2.       Use the eDART’s Part Sampling tool when running the experiment to mark in the data when each group of parts was taken.

3.       BE SURE TO COLLECT AND LABEL THE PARTS CORRECTLY TO MATCH THE NAMES GIVEN FOR EACH RUN!

4.       Stabilize and measure the parts

5.       Transfer the data from the eDART to a computer where you can use the Analyzer (latest version – 3.1). The eDART’s FTP backup service in the eDART Configuration can do this automatically if you have the eDART networked.

6.       Open the data in Analyzer and check to make sure everything looks sensible; i.e. nothing “broke” during the experiment, the experimenter actually did set the hold pressure high when he said he did and so on.

7.       Open the Measurement Entry tool, create columns for the measured values and enter the values (may be pasted from Excel)

8.       Save the entered measurements

9.       Run the Auto Correlator and look for interesting correlations, focusing on the in-cavity (“intermediate”) variables, if possible. For example increasing hold pressure could increase cavity cycle integrals but we want to alarm on the cycle integrals, not hold pressure (usually a constant).

10.   Export the interesting variables (with parts measurements) to Excel and use scatter plots and other tools to help with making good choices for alarm levels.