iSixSigma

SPC for High-Mixed/Low-Volume?

Six Sigma – iSixSigma Forums Old Forums General SPC for High-Mixed/Low-Volume?

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #32851

    TM Aw
    Member

    The nature of my Production is of high-mixed, low volume using the same machine to produce a variety of product models. Sometimes, the production run for certain models can be as short as 4 hours or even shorter. We collect 5 pcs of samples for our hourly in-process inspection due to limited resources. As for trial run, we run less than 20 samples.
    Ques 1: How feasible it is to use Control Chart (eg: X-R Chart) to monitor such short-run process during trial run and actual production?
    Ques 2: Is it meaningful to continue collecting data on the same sheet of Control Chart for the particular product when there are some product changes cum parameter changes to the same machine over the past few days?
    If the answer to Ques ‘2’ is ‘YES’, then can I accumulate the data until I have enough data to do a Cpk study later on such process? If ‘NO’, then what would be the alternative method to monitor the process and do capability study using SPC method?
    Rgds
    TM Aw

    0
    #88197

    Michael Schlueter
    Participant

    Hi,
    My personal taste to Q1 is: doing a X(-R) chart on few samples is better than doing nothing. You never know what the data will tell you, unless you analyze them.
    Can you please give me some more details on the kind of product you make? E.g. what kind of parameter changes do you introduce? What kind of product changes do you have to cope with? There may be more specific options for you, which depend on your product.
    How important is the final quality level for your customer? I.e. how severe are variations? E.g. variability in brakes or medical equipment can have dramatic impact on peoples life, hence incure high (monetary) losses. While variability in toys may be of no consequence at all (well, kids may have a different perspective on this ;-)
    The answers to Q2, to my taste, again depends on your specific situation. E.g. when you measure your products density, and parameter changes do not affect density, than it is useful to continue tracking density. While, when you measure different product dimensions and the products size changes, I would regard these as two completely different situations (= processes).
    Best regards, Michael Schlueter

    0
    #88201

    Gabriel
    Participant

    I know there exist some short run charts (standardize chart, called Zbar-W, J. Wheeler “Short Run SPC”). Unfortunately, I only know they exist, so I can not provide more information. Michael or others…?

    0
    #88210

    David Oakley
    Participant

    You can do SPC charts with a sample size as small as one. The constants take care of the different sample sizes.
     
    If you are making adjustments to the machines, the text book answer is that you can’t do the statistics until the system is in control. If  you are making parts with the same parameter but with different values, you can do something with that. For example, if you have a cut to length machine, for each sample, you record the length difference from nominal instead of the actual length. That way, you can use data from different parts in your chart. I have used this before and been satisfied with the results.
     
    David

    0
    #88211

    Mikel
    Member

    For a HMLV environment, it is most meaningful to a a SPC type mentality for turning the process on. That, combined with quick change over (yes, they really do mean single digit minutes) are the most important controls for this.

    0
    #88322

    TM Aw
    Member

    Thanks for your reply. I am from the electronic manufacturing industry where we assemble different electronic components onto PCB boards. The process I am referring to are I) wire crimping process and II) Surface Mounting processes.
    We crimp an end connector to the wire end for the crimping process. We use different sizes of wires and different type of end-connector in this process for different product. The crimping strength varies for different type of wires. The crimping strength is verified with a pull test and the machine setting is done manually by adjusting the punch-head height to the required position. We also need to change the jig to hold the different type of end-connector.
    As for SMT process, we monitor the Solder Paste Dispensing machine and the Reflow Oven. We measure height and area of solder paste being dispensed at the dispensing machine. At Reflow Oven, we measure the peak temperature. There is a range of solder paste thickness to be monitored. We change the pressure and speed of the squeezee at the solder paste dispensing machine for every different product being run.
    As for the Reflow Oven, we change the temperature setting and conveyor speed for every different product being run. We measure these characteristics with a temperature profiler. There will be some minor adjustment to the conveyor speed and temperature setting during the run according to the reading taken from the temperature profiler.
    With such variability of material, specification and machine setting for each product, would it be meaningful to continue collecting the data for further analysis when there are many changes of product run in between for such high-mixed, low volume type of manufacturing environment? Is SPC control (eg: Control Chart, capability study) still feasible under such environment? Note: Each change of product requires easily ½ hrs of setup time.
    Rgds
    TM Aw

    0
    #88329

    Rocky
    Member

    TM Aw,
    SPC charting are used to verify the process are behaving the way you expected it to be. Having SPC around will  not solve the problem but will tell you the problem exists thus you can react swiftly on the matter. Are you getting tired of analyzing the data for your HMLV production ?? Hope not !
    I am in the same manufacturing industry as you are in now and handle almost all the process you’ve stated. Putting SPC charting will not hurt you as long as you do it right, analyze it well and provide sound corrective action. This is your fist line of defense against wide range of variations. If you insist of doing so but afraid in the potential variation in the process could go out of scale, perform a DOE in product specific manner. (Whoooo !! Do you have time??money??). From here, you can predict the acceptable variation in your processes. Document the parameter setting for everybody’s information and fool-proof the process as you can. These will perfectly work if you considered all important input factors (X’s) that affect your output (Y’s) (Well, are we done??.). Check your machine capability and your measuring equipment should be sound. And last but not the least, screen out all inputs into your process to be within specification e.g. PWB, temperature, humidity, viscosity and others.  
    I guess after you read this you will appriciate SPC. If not, then, I may not explain myself well… poor me !!!
    Rocky

    0
    #88361

    Mike Carnell
    Participant

    TM Aw,
    There is a difference between Statistical Process Control and Statistical Product Control. Just because you are changing products does not necessarily mean you are changing the process you make the products with.
    You might want to sit down a while and look at your process and process data. We assume that because things have different part numbers they behave differently. on my first wave solder machine we grouped product by designer because we had 4 major design groups and they had distinct styles and they all ran differently. It worked.
    Good luck.

    0
    #88380

    Michael Schlueter
    Participant

    Hi,
    Thanks for your details. They convinved me even further that your SMT  process is demanding for robust engineering:

     the SMT process runs smoothly 
     for a wide range of specifications
     in a controlled way.
    There is a certain approach to make this happen; but, as another member noted, it requires dedicated experiments – there are ways to do such experiments effectively, i.e. with little effort. In best case you can just feed the experimental parts during normal operations; it’s a matter of planning.
    It also does make sense to investigate your available data for defect analysis (“talking parts”) to generate clues about your processes (or to track interesting data for a short while). E.g. Bothe describes several useful techniques to identify trouble makers, like BoB-WoW analysis (best-of-best, worst-of-worse), paired comparision, scatter plots, variables search etc.
    I offer you analysing your existing data. I think you shouldn’t wait for future analysis, but do it now. I was tempted to encourage you to continue collecting data, but: what is it good for? Once we have narrowed down the important things to monitor, you can do more focussed data tracking. I mean, there may be more things to care for than you observed already.
    BTW, we didn’t talk about the defects which bother you, so far, did we?
    “Is SPC control (eg: Control Chart, capability study) still feasible under such environment?”
    Honestly, I can’t say at the moment. Which quantity would you subject to SPC? I think it must be either a universal or a transformed quantity, as far as I understand your process today: because SPC requires a single quantity, which should stay at exactly one target value; while the SMT process probably requires a set of individual (unknown ideal) target values.
    During my preparations for an answer I came accross this site: http://www.thepdfshop.co.uk/ppm/html/reflow.htm , which offers ppm-help besides galleries of trouble. Very interesting for me. Perhaps you can find an answer to your SPC question there?
    I suggest to contact me at:

     [email protected] 
    when you are interested in my offers regarding robust engineering and data analysis.
    Best regards, Michael Schlueter

    0
Viewing 9 posts - 1 through 9 (of 9 total)

The forum ‘General’ is closed to new topics and replies.