iSixSigma

Operator Error

Six Sigma – iSixSigma Forums Old Forums General Operator Error

Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #35945

    Citizyn
    Participant

    Greetings,
    I am in the Measure Phase of DMAIC training. My project is to reduce scrap in the metal turning process. I have2 questions, both pertaining to human error:
    1. On my Data Collection Plan, 3 of my measurables are directly tied to human variables: Setup, Operation, and Training. I am utterly clueless as to how to quantify or measure them.  
    2. According to Deming, only 2.5-5% of problems with a process are attributable to human error – the rest is process error. One book says that there is no such thing as Operator Error. That’s great. But my project’s process is rife with human interaction, and the cause of over half the scrap cost in historic data is Setup/Operator error. I know that mine is not the first Scrap Reduction project to post, so how do I attempt to eliminate opportunities for defects in this circumstance?

    0
    #102123

    [email protected]
    Participant

    The first thing I would do is perform an Analysis of Variance, given the different characteristics you want to study.  To simplify matters, assign a low value and a high value such that you have 8 different combinations to study, ie:
    Setup        –  +
    Operation  –  +
    Training     –  +
    This will give you the following matrix:

    Setup
    Operation
     Training

    Exp 1

    +

    Exp 2

    +
    +

    Exp 3


    Exp 4


    +

    Exp 5
    +
    +

    Exp 6
    +
    +
    +

    Exp 7
    +

    Exp 8
    +

    +
     
     

    0
    #102135

    DaveG
    Participant

    Your measurables are far too vague.  I recommend you:
    1) Pareto the defect / process types associated with scrap:  Small OD, Large ID, Undersize Width, Part Number families, etc.  It would be wise to pareto ALL defects to ensure you don’t miss potential root causes – a specific root cause could affect dispositions other than scrap, so you kill 2 birds with 1 stone.
    2) Pareto by criticality:  safety, delivery, cost, etc.
    3) When you’ve identified your drivers, do exhaustive fishbone root cause analysis.
    Regardless of the problem to be solved, I always ask “Is there a system which, when complied with, consistently yields an acceptable result?”  You want to ask that question to ensure you actually have a problem.  After all, whoever designed the process made it incapable of producing 0 scrap – maybe that is what they intended.
    IMHO, most problems are caused by lousy systems (original idea by W. E. Deming) and a failure to adequately define and offset risks before going into production (FMEA).  Combine that with Product Inspection used as the chief means of Product and Process Evaluation, and a company is always playing whack-a-mole.

    0
    #103137

    Tim Torrey
    Member

    What kind of metal turning process are you dealing with?  How complicated is the setup?  Is the setup causing the scrap, is the part being mismachined because the setup was done incorrectly?  How are you measuring training?  Has the amount of training for each operator been documented? Is there a way to reduce the possibility for error in the setup thru tooling or instructions or a setup drawing?
     

    0
    #103144

    rodrigues
    Participant

    Citizyn,
    My suggestion would be for you to look at the process inputs that affect operator performance such as training and competency validations; failsafe methods to reduce operator error such as overylay templates for hardware installation, forming plates to prevent forming parts backwards, special tooling, fixtures,  etc.  Setup sheets and accuracy, job routing accuracies, etc.
    I would also ask the operators what they feel is the greatest contributor to errors and what would help to reduce them.
     

    0
    #103148

    Ron
    Member

    In the measure phase I would focus my data collection of quantifying the amount of scrap and the shifts it is produced on the machine it is produced on and the operator running the machine at the time. To get into a root cause is inappropriate in the measure phase.
    Once you have a “Big” picture of where the scrap is coming from you can narrow down your search in the analyze phase to determine the root cause of the problem.
    I may be wrong but in your statement it sounded like you were looking for the operator to be the culprit. In my experience I have categorically eliminated the “operator” from any root cause analysis. If they have poor training, are not following the training proceedures, or have improper setup it is not the operator it is management that has allowed this condition to exist.

    0
    #103173

    V. Laxmanan
    Member

    Dear Citizyn:
    If you wish, send me an email and I will try to help you. Regards.
    Laxman
    [email protected]
     

    0
    #103181

    Citizyn
    Participant

    Wow – lotsa good responses to this posting!
    No blame. Here’s how it works: We’ve got CNC Turning machines with 30-year-old controls (but the machines are rugged, reliable, and repeat). We cut a lot of alloy steel and cast iron with indexible carbide insert tooling. As the carbide wears, the size changes. As the size changes, the operator edits the settings on the machine (tool offsets). This happens constantly. Eventually what happens is the operator (I’ve done it several times myself) enters a positive number when it should have been negative (or visa versa); or they mis-key a number, missing a decimal place. There are myriad other mistakes that can be made while making insert changes and other mundane tasks. There are so many opportunities for error, I think it’s miraculous that we don’t scrap more than we do (to the operators’ credit).
    Currently I’m measuring scrap in 2 fashions: 1. With a ratio of parts scrapped over parts produced; 2. With a ratio of $’s lost to scrap over $’s earned by production.
    Agreed, it is premature for me to dig for solutions.

    0
    #103184

    Peppe
    Participant

    Dear Citizyn,
    I read in a article, that every people that work 8 hours per day, typically, do 2 errors per day. This was confirmed many time by experience. Now if you have setted the production line as poka yoka, you can contain this errors, otherwise the produce fail or scrap.
    My suggestion is : take care of this and put in place a strong FMEA and cause/effect analysis, so you can, with your engineering dpt, organize the production flow in the way to prevent that errors. 
    There are many way to do it, but you must analize and understand which is the best for your environment.
    I suggest also to talk with operators and ask them why they are often wrong, from their answers you could find many good suggestion.
    Rgs,
    Peppe

    0
    #103243

    V. Laxmanan
    Member

    Dear Citizyn:
    Currently I’m measuring scrap in 2 fashions:
    1. With a ratio of parts scrapped over parts produced;
    2. With a ratio of $’s lost to scrap over $’s earned by production.
    I like your second measure. Keep it up! 
    Many people lose this focus when they get into “fancy” ideas of what the “process” improvement is supposed to accomplish.  It is not percent scrap or reduction in percent scrap, but the lost $ that really matters. 
    So, I say, y = hx + c = hx – W, where y is profits and x is number of parts and W = – c is the work function for your operation.  If you really know how to do what you say under 2 above, you have accomplished what no one has accomplished before.  But, now you must understand the meaning of the work function for your operation. Regards.
    Laxman
     

    0
    #103246

    Robert Butler
    Participant

    Your second post sounds like the basis for a good data collection scheme which, if properly done, has the potential for providing critical information about your system.  It sounds like you, and the operators, are aware of a variety of mistakes which could have a major impact on the process
      (“Eventually what happens is the operator (I’ve done it several times myself) enters a positive number when it should have been negative (or visa versa); or they mis-key a number, missing a decimal place. There are myriad other mistakes that can be made while making insert changes and other mundane tasks.”)
      I’d recommend taking advantage of their expertise by sitting down with them and asking for a prioritized list of these mistakes.  You don’t need to list all of them, just the top 10-20 and make sure there is a column marked “other”. (This last is important – I’ve worked several projects where we did as described above and in very short order we found that the single biggest category was “other”.  Because everyone was in on the effort and everyone could see what was happening, it was an easy matter to not only stop what we were doing but change the direction of our efforts. This last was easy to do because the initial check sheet had incorporated everyone’s “best guess” with respect to root cause.  When none of these seemed to be a clear winner it was easy for everyone to let go of their pet theories and focus on the problem and not the office politics)
      After you have the list, set up a check sheet and a procedure for shift-by-shift collection of these sheets.  Marking a check sheet is extra effort. In order to get buy in for this effort you need to have a definite start and stop date and everyone involved needs to know the duration and reason for the exercise.  You will probably have to go out and gather the sheets yourself.  This may seem foolish but there’s nothing like seeing the team leader on the floor actively collecting sheets and asking questions to drive home the seriousness of your interest.
    A check sheet with mistake type and item being worked will permit you to evaluate the mistakes, their frequency of occurrence, and the cost of the mistake from the standpoint of part being worked at the time. You should make a daily summary of the check sheets. Pareto charts of error type, dollar cost of scrapped part, etc. will go a long way towards helping you see what is going on. More importantly, making these charts available to everyone will allow all who are participating to see how things are going as well.  

    0
    #170068

    Brad
    Participant

       For anyone still reading this post; we are a production shop running approximately 60 CNC machines (plus many more manuals) and have the same problem as Citizyn. We machine a lot of specialty alloys in the High strength non-magnetic stainless steels and Inconels such as 718 and 925. What is amazing is surface foot for surface foot we generate many more “human errors” on CNC equipment. This has always seemed to me to be a paradox.
        The only successful way we have dealt with the problem of offset errors are to prevent the need for the human entered offsets in the first place. We have accomplished some of this by:  

    Purchasing more consistent inserts – many commercial inserts are not very accurate (+/- .002) This can quickly translate to ± .004 on a lathe X-axis. 
    Premeasuring & sorting inserts.
    Programming for automatic tool wear adjustment. Defining carbide change points.
    We have also begun installing automatic tool compensation software which can be added onto any existing CNC control, relatively inexpensive. Please feel free to contact me for info on this.
    SPC tracking of adjustments has minimized the number of adjustments beingmade as well as reduced the variability created by constant tweaking.
     

    0
Viewing 12 posts - 1 through 12 (of 12 total)

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