Setting Up SPC for Injection Molding Machines
- May 8, 2019 at 6:43 pm #238973
AhanceParticipant@Ahance Include @Ahance in your post and this person will
be notified via email.
To preface, I will be attempting to implement SPC on some injection molding machines (IMM’s). After a good amount of research online, it seems that SPC is rather redundant on something such as an IMM, as there is already minimal variance as well as a lot of input variables, making the likelihood of false negatives much higher. Knowing this, I am still going to attempt SPC on these IMM’s as it is something one of our customers has requested of us.
I am hoping to gain some insight as to how I may want to set up this SPC so as to not overwhelm myself, as I had never even heard of SPC before being assigned to this task. I will attempt to write out how I plan to implement this SPC, and would like some guidance/tips/tweaks to my plan from anyone experienced in the community.
To start, I plan to do a fishbone diagram as well as talk with some of our IM operators to determine both the special and common cause variability. From here, I will attempt to remove all special cause variability, and then conduct a Design of Experiments (DOE) to determine which common cause/independent variables have the most influence. Next, I will pick the dependent variables/outputs I want to statistically control, as well as the most heavily influencing independent variables/inputs that appeared in my DOE.
This step of choosing which variables you want to monitor seems to be a hotly debated topic for SPC on IMM’s. Because many of the variables in IM’ing are correlated, it seems that many people choose to do some sort of multivariable control, rather than univariable SPC. If anyone has more insight on this I would love to hear it, as I was planning on just looking up what people thought were the most useful variables to monitor and just choose a few so as not to make this overly complicated with many correlating variables I am tracking all individually.
Once I have the input variables I want to control, and the outputs I want to track variance of, what is the process of beginning to fine tune the inputs? After eliminating all (or as much as possible) special cause variance, and getting the process into a state of statistical control, how do I go about further improving the process? Is the best method to just start tweaking your inputs while tracking your outputs to slowly reduce variance? I don’t quite understand how to do this without the possibility of accidentally taking your process into an out of control state.
I appreciate any input, and let me know if there is any further information you need from me.
-Aaron0May 9, 2019 at 3:36 am #238981
Daniel SimsParticipant@Daniel.S Include @Daniel.S in your post and this person will
be notified via email.
I don’t suppose the customer gave you any indication of what exactly they want in SPC form? I guess it boils down to your Xs and Ys. I’m going to assume that your process is already in control, you’ll have setting sheets and have a grasp of what is important to ensure a good part out the other end. Off the top of my head I guess some of these things will be hot runner temperature, tool heaters, injection time, holding time etc.
I’m not as knowledgeable or experienced as many of the posters here, but my first pass at this would be identifying what metrics the machine can give you per cycle. Usually these have alot of info. Hopefully it can store a large number, I know some older machines only store around 20 cycles worth of data, in that case you’d have to have a collection plan. Once you can reliably track these settings I would first just run an individual chart. This just means plotting every point for given a period to get a feel for the data.
I’m going to assume that the customer gave you no indication of what they actually wanted from SPC, so I’m also going to assume they just want a reassurance that you have a handle on your processes. To meet this theoretical requirement I would take samples (for the sake of argument 8 cycles worth of data as the mould machines I’ve worked with before showed 8 at any one time), I would do this every three hours, which for my place would give me three snapshots per shift. Track them through an X-bar(R) chart and assuming you are in control you’d have to wait until parts are made out of spec and look for the relevant change in one of your inputs.
There are issues this won’t cover, for example flash from tool wear but I’ll assume you have competent toolmakers on site to diagnose and repair this. I’m sure there’ll be some pretty great advice from others on here, just giving you my opinion on it.2May 9, 2019 at 12:25 pm #238984
Mike CarnellParticipant@Mike-Carnell Include @Mike-Carnell in your post and this person will
be notified via email.
@ahance You need to begin with your data rather than a fishbone diagram. What type of defects are you seeing. If you have under weight parts or incomplete parts then you are probably getting short shots. Once you have something tangible to work on you look for the X’s that control the Y’s. remember Y=f(x). Effective SPC is on the x. If you control the x’s (independent variables) then the Y (dependent variable) will be good. Basically if I control shot size then I have part weight pretty well under control.
The problem you are going to have with injection molding machines is if they are relatively new they have some pretty good controllers on them. You can chart things like time to close but it isn’t going to change much and if it does it will be because that is what the controller allows. If I were you the first thing I would control is temperature of the water into the machine and the flow rate. That water temp and flow rate have a lot to do with how fast you can run the machine. Look at a profile of the program that runs the machine. Typically the time to cool is the longest time in the cycle. There typically isn’t much in the way of machine controls on the machine it assumes the chiller is taking care of that. The chiller controls are at the chiller not the machine so if you get a plugged line in the machine you will never see it beyond your run times getting longer. This a huge hole in injection molding control.
Same issue you saw on water. What happens when every machine is running at the same time. What is the psi at the machine. Your compressors may be putting out 120 psi but that has nothing to do with the machine.
Your leverage is in the common systems air and water affect the entire plant. Those systems can improve the performance of the entire plant.
Hydraulic systems are at a machine level typically. That oil needs to be clean and stay clean. You can chart it but from an operations point of view it makes more sense to get a cleaning system and clean the oil at least monthly.
Look at your cost of colorant. Very expensive. If it gets out of control it will cost you a lot of money. There are a lot of ways to deliver it to the machine depending on how you supply material to the machine. That is worth looking at in terms of control of amount and mixing.
Just a summary. Start with common systems – air and water. Colorant maybe depending on how it is delivered to the machine. Once the common systems are under control you will take the system level noise out of the machine. Look at the data by machine and by part. If I can run a mold on two different machines and have the same defects then check the mold. If the defects are different then start to look at the machine. You can put SPC on things that already have controls but it is a waste of time but make sure you have a way to evaluate the sensors for the controls. The controller will do what you told it to do but the sensors need to be accurate and precise (basic Measurement System Analysis – 5 factors)
This is a pretty simple process to control but you need to understand that there are 4 systems in every machine 1. electrical 2. mechanical 3. pneumatic 4. Hydraulic each system has to operate in control and then they have to operate together in control.
You can run SPC on your material or ask it from your suppliers. If you run regrind and you are one of those facilities that has this magic number for % regrind you could put SPC on that.
In general if your customer is just carte blanc asking for SPC and doesn’t have a clue what your process is they are relatively uneducated in terms of controlling a process. They are checking boxes so determine where you will gain something from the SPC and put it there. Have data available that demonstrates that you machine controllers are doing their job and SPC will be redundant and add cost. You will get the occasional person who will tell you Deming said that controllers will not do the job. Remind them that Dr. Deming passed 25 years ago (1993) and never saw the technology we have available today.
Just my opinion2May 10, 2019 at 1:21 pm #238993
You must be logged in to reply to this topic.