FRIDAY, OCTOBER 31, 2014
Font Size
Topic SPC limitations in machining enviroment

SPC limitations in machining enviroment

Home Forums Old Forums General SPC limitations in machining enviroment

This topic contains 61 replies, has 1 voice, and was last updated by Profile photo of Alto Alto 5 years ago.

Viewing 62 posts - 1 through 62 (of 62 total)
  • Author
    Posts
  • #114878

    Does anyone have experience in dealing with the limitations of using tranditional SPC (X-bar & R, etc) as process control in the machining / metal removal environment? By limitations, I’m referring to; SPC as a static-sampling based process control not able to adequately control – or even consider fundamental factors in a dynamic metal removal process. For example, one major drawback in applying SPC is that the tool-wear (cutting edge breakdown) as a non-linear relationship is not considered. Another major drawback is that SPC is not real-time in that it requires sampling history before a decision is derived. 

    #114879

    Jimm,Usually in our machining lines we use near real-time feedback from the gages, with tool wear comp programs. So for example, on our crankshaft machining, we have Marposs gages that measure a part in the station just past the machine tool, and send the adjustment for the next part back to the machining station. Some of the grinders are real time, measuring as the part is ground down. In both examples, SPC is also tracked to look at differences in raw stock, tool set-up, etc.On machines that don’t have in-line gages, they will have a manual off-line gage that measures samples and puts them into an x-bar/R chart. Since we know that the tool wear generates a sloped trend line (which does happen to be pretty linear) we have sloped control limits. Tool change happens long before the part gets out of spec.

    #114884

    I would recommend looking into EWMA charts.  In my experience, these are highly effective in detecting trends in processes.  You can also control the sensitivity of the charts.  I won’t get into the math and stats behind EWMA, but definitely do some research.  Minitab offers a decent explanation.
    V

    #114917

    As six-sigma practitioners, we are trained to think outside the box. Consider the original question and situation carefully. Understand that within the machining environment there are mulitple factors; variable and non-random which affect the desired outcome – i.e. dimensional stability and surface finish per customer (VOC) requirements. Given all these dynamic factors at work at any given time; cutting tool edge breakdown (which is often non-linear and cannot be represented by simple stepwise regression), tool pressure, tool heat, tool rigidity, material constiuency, material hardness, coolant (if employed) flow and lubricity, etc, etc, here is my question….
    How can we expect SPC – with the static sampling rules – to tell us – using predictive analysis for each individual tool as it performs the work – when the process must be altered to achieve stability? SPC does not employ nor consider modeling for tool wear – which given the process dynamics cannot be grasped using simplistic sampling nor EWMA techniques.
    Note the specific requirement of: predictive analysis for each individual tool as it perfoms work. Realize that if you change the tool cutting edge, replace a perishable insert, or perform an offset – you now have a different process and the tool wear model is different.  SPC cannot recognize this and allow for correct process control. Within the machining environment there is simply not enough history available (things happen too fast) to allow for static average plotting and data collection of subgroups (SPC) to be effective. I need a process control methodology which is more real-time and representative of actual conditions to predict tool wear and signal when to; stop and correct or leave the process alone.  SPC does not and cannot acomplish this requirement.

    #114918

    As six-sigma practitioners, we are trained to think outside the box. Consider the original question and situation carefully. Understand that within the machining environment there are mulitple factors; variable and non-random which affect the desired outcome – i.e. dimensional stability and surface finish per customer (VOC) requirements. Given all these dynamic factors at work at any given time; cutting tool edge breakdown (which is often non-linear and cannot be represented by simple stepwise regression), tool pressure, tool heat, tool rigidity, vibration and harmonics, workpiece material constituency, material hardness, coolant (if employed) flow and lubricity, etc, etc, here is my question….
    How can we expect SPC – with the static sampling rules – to tell us – using predictive analysis for each individual tool as it performs the work – when the process must be altered to achieve the desired size and finish? SPC does not employ nor consider modeling for tool wear – which given the process dynamics cannot be grasped using simplistic sampling nor EWMA techniques.
    Note the specific requirement of: predictive analysis for each individual tool as it perfoms work. Realize that if you change the tool cutting edge, replace a perishable insert, or perform an offset – you now have a different process and the tool wear model is different.  SPC cannot recognize this and allow for correct process control. Within the machining environment there is simply not enough history available (things happen too fast) to allow for static average plotting and data collection of subgroups (SPC) to be effective. I need a process control methodology which is more real-time and representative of actual conditions to predict tool wear and signal when to; stop and correct or leave the process alone.  SPC does not and cannot accomplish this requirement.

    #115639

    It depends on what you are trying to achieve.
    SPC intended for detecting special cause variation (i.e. the variables you don’t know about) in the presence of common cause variation. How you design your rational subgroups or collect your data will define the effectiveness of your chart. SPC is perfectly capable of fulfilling this role, and has efficiently allowed the development of 18 to 25 sigma machining processes in Japanes industry from 1948 onwards. If you have known variables you cannot, or don’t want to, eliminate from your process (for example tool wear) then model the effects of these and SPC chart the residuals.
    If, on the other hand you want to use methods to CONTROL your process, by providing feedback and adjustment, then you may be better using charts designed for that. Take a look at “Statistical Control by Monitoring and Feedback Adjustment” by Box and Luceno. This is more Engineering Process Control than SPC.
    Define your objectives and then choose your tool. But please don’t criticise SPC for not being able to achieve something it was never intended to, and please don’t couch your personal prejudices in the false guise of a question.

    #115643

    Jimm,
     
    It would appear that you have stumbled onto a truth that is the basis of six sigma.  One cannot throw a SPC chart on the finished part and expect it to correct all of the issues.  There are no output measurement systems that will do this.  The SPC at on the finished product is intended to let everyone know that there is something wrong within the process.  By the time a data point out of acceptable tolerance is collected the part has already been made.  Consequently, we agree, we cannot control any process by measuring the output.  The output will only tell us when the process is not were it should be.
     
    So what can we do to make this situation better?  I would suggest we go back to the basics. 
    Y=function(X)
    You listed a number of different inputs that may have an affect the output SPC measurement.  The basis of six sigma is to define exactly what impact those inputs have the output. 
    1.  What impact? 
    2.  How much impact? 
    These are the questions we need to answer.  Once we have the data and can answer these questions, we have the ability to begin to control the process.  The output SPC then becomes the alarm that notifies us that something in the process is not in control. 
     
    I have found that I like to have controls and measurements on the inputs I have found to be important.  This allows me to be proactive instead of reactive.  If I can control the important inputs the output is already controlled.  Thus, the adage, crap in crap out can also be stated good in good out.

    #115649

    If this is a standard machine shop scenario, then this would be a course of action I have taken in the past that worked well. Start by collecting data for your part(s) with 100% inspection (if possible) until you have a small accumulation. Run a Cusum Average on the numbers to determine the status of the part(s). Once the part(s) are in control, you can start reducing the inspection sample size. Once in resonable control, the sample can be done with First Article samples and a designated sample size after that. Now, at the same time you can focus on parts that have a history of rejection using your scrap history or rejections. Focus on this group using an XMR chart to track progress. Short term variation is captured well by this. The combination of these two approaches is very effective. 

    #115650

    Jimm,
    I agree with Dreemr 100%. While gathering data to establish Control Limits, you can still use your charts by relying on the known Specification Limits and trend analysis to understand the effect of any common causes (like tool wear) or special causes. Then you can decide if any corrective action is needed.

    #115685

    Philip,
    Why would you consider my question regarding SPC limitations as a personal prejudice? Is it not possible that the tried-and-true (often misused and over applied) staus quo SPC rules may not apply in a metalworking environment? Is it simply too far-fetched for you that one may even ask – or dare pose the question, or is the possibility too limited for you to consider? For that, I must offer an apology – I will limit my thinking and studying of the problem to that which you can grasp. 
    I had already defined the objectives; process control methodology which is more real-time and representative of actual machining conditions to predict tool wear and signal when to; stop and correct, or leave the process alone.
    My question, is necessarily biased in light of the limitations I have hands-on experience with; i.e various commercially-available SPC data collection software – InfinityQS, Zonetec, ASI Datamyte, KurtSPC, Statistica, VisualSPC, etc. None seem to offer this level of process control in regard to tool-wear modeling as a predictor of process control. Perhaps with this type process control desired, maybe SPC is not the correct tool – then does anyone have experience with another?
    I posed the original question to see if others have had similar experience with the general limitations of SPC in this type metal removal application.
     

    #115742
    Profile photo of Alastair Lea
    Alastair Lea
    Participant
    Reputation - 0
    Rank - Aluminum

    A normal Xbar R chart is best, because ……….
    Tool wear is just another variable in the process.
    If your tool is a 15 tonne steel press or a 20 gram mill or lathe bit the problem is the same. The tool will wear over time.
    The issue is how slow or fast is the tool is wearing in relation to the rest of the other process variables.
    EXAMPLE : if your tool requires replacement every 30 mins operating a 1 hour sample rate not appropriate. Every 5 or 10 mins would be better.
    The sample period must be designed to capture the variation in the process that is introduced by the tool.

    #116114

    Alistrair,
    I keep seeing references to using SPC charts to control lathe tool wear. Do you do this yourself? What has tool wear got to do with the tolerances of a machined component?
    If you agree tool wear has nothing to do with the tolances, either of a first off, or any other part – why do it?
    Blanco

    #162830

    STATISTICAL PROCESS CONTROL FOR PRECISION MACHININGMany folks have attempted to implement X-barR charts for years in precision machining applications, only to be frustrated by tightly compressed control limits or unexpected bimodal distributions. The problem turns out to be that the X-barR charts are the wrong charts for precision machining. They utilize the wrong statistical distribution, and they promote overcontrol. Precision machining is non-normal. There is a better way.What Is A Normal Process?A good starting point is to ponder whether a distribution can be expected to be normal. Normal distributions are a result of normal processes – and there is an emphasis on natural variation. One of Shewart’s examples was tensile strength. Random natural variation caused by a myriad of influences – chemistry, crystalline structure, surface flaws, etc. Sure, I’d buy that. The example I like to use is a processing line of loaves of bread. The height of the loaves of bread is controlled by so many variables – proofing, yeast quality, humidity, accuracy of ingredient ratios, etc. The net result is a natural variation -most a particular height, some less, some more. If a process can be expected to stay at a particular “level”, with some variation above and below that level – with NO operator intervention – until a special cause appears, it’s normal. That is the “voice of the process”. But, if you have to have someone adjust is to keep it there, then it is not normal – and you might as well start investigating what the distribution truly is. The problem is how many people assume normality, because they are attempting plug-and-chug statistics. The other problem is people performing transformations because they only understand normal data (barely), since transformations tend to mask valuable information. It is not so much their fault, they have likely not been trained well. What Is Precision Machining?First of all, it is sensible to define “precision machining”. For this discussion, it is a process where the most statistically significant variation originates from tool wear. So, any variations from worn bearings, measurement, etc., are all controlled to a level that is statistically insignificant.The Correct Distribution For Precision MachiningThe true distribution for precision machining as defined above is the uniform or rectangular distribution – not the normal distribution. You can tell if your process meets this distribution by performing a capability study. If, for example, you are machining a OD, set the process at the lower control limit [nominal - .75(tolerance/2)]. If – without operator intervention – the process increases to the upper control limit [nominal + .75(tolerance/2)], then adjust back down to the lower control limit. If this continues (until a tool breaks, or surface finish deteriorates – special causes), then you have the “sawtooth” curve, and it is the uniform distribution. Some normal-centric statisticians like to try to ‘normalize’ the data with transformations, but that is unnecessary, as well as a useless effort. The sawtooth curve is more meaningful as is to an operator – they understand tool wear.Calculating Control LimitsThey should be approximately 75% of the tolerance, centered within the tolerance. The control limits really never need to be adjusted. Compressing the control limits actually increases overcontrol – and therefore increasing variation. The slope of the line is the tool wear rate – which is meaningful information that would be masked by transformation. Notice the mean has no use whatsoever in the sawtooth curve – only the control limits. Don’t forget, most of the Western Electric rules – especially the one concerning runs, do not apply – they are for the normal curve.Since the probability of the uniform distribution is straight forward, 75% of the tolerance gives you well below the probability of +/- 3 std dev of a true normal distribution. You could use a higher percentage of the tolerance, but it is better to play it safe due to hysteresis concerns (you can never land exactly on the control limits). This follows AIAG SPC Chapter III Non-Normal Charts, last bullet point: use control limits based on the native non-normal form. (Also, AIAG PPAP section 2.2.11.5 states that the Cpk calculations are not applicable, since the uniform distribution is non-normal and the calculations are for bilateral normal distributions).What Chart Works With Precision Machining? I must add that X-bar-R charts are the worst for true uniform distributions. X-MR would be a little better. Again, the mean means nothing in the uniform distribution – both for the population or the sample, and there should be virtually no discernable variation between 5 consecutive parts, unless you are shredding up tools. The first problem people run into when running a machine capability is they measure one of the resulting part diameters. The emphasis is on ‘one’. How many diameters are there in a circle? There are an infinite number. So, how can you describe or predict an infinite number of diameters with one measurement? You can not. So, to resolve that problem, you need to pick a specific diameter – such as 24.000 mm from the end of the part (so that taper does not affect your data). You need to measure around the diameter, and determine the largest and smallest diameter measurements. Then, plot both of those dimensions on a X hi/lo -R chart, with UCL and LCL at 75% of the tolerance for X and 30% of the tolerance for R. The range is the difference – or the roundness. Continue that for about 100 pcs. (Make sure your gage R&R is less than 10%! Try to use the same material lot, if possible, operator, etc.) Is that enough pieces? Usually. If the process is ‘in control’ you should see the diameter increase as the tool wears (for an OD, opposite for an ID). You might see some fluctuation at the beginning as the machine warms up. That is a special cause that you cannot remove, but need to consider. If the tool wears until the X hi data is up to the upper control limit, then adjust until the X lo reaches the lower control limit. (Adjust during a capability study?? Yes.) If you get two cycles, you will have a real good idea of how the ‘machine system’ is going to perform. If the machine runs 100 pieces with no need for adjustment, but the data is gradually increasing at a steady rate, then you can likely extrapolate when the adjustment would need to be made. If you have the luxury to find out how long you can run before adjustment with more pieces, then all the better. There are occasions where the tool will wear to the point of breaking or poor finish prior to need for adjustment – and that information is good to know too. But, as long as you can keep the process between 75% of the tolerance, and you are capable to at least 1.33 [capability=(USL-LSL)/(UCL-LCL)].The next thing to review from the run is roundness (R chart). If the roundness is less than 10% of the tolerance, you have no worries. If it is greater than 30%, then you will have to watch that process like a hawk with frequent SPC checks. I would vote to resolve the roundness issue (which is most likely a machine or machine set-up issue, particularly in chucking or bearings). If it can not be improved, I suggest passing on that process and finding one that can maintain the roundness better. Now you have some real data on whether that ‘machining process’ will work for you! This is the short and sweet lesson, but I hope it corrects some of the misconceptions on the topic. The X Hi/Lo –R Chart Sounds Like More Work No, not really more measurements. Right now, most people measure a diameter on 5 parts. One diameter out of an infinite number of diameters. That would be statistically insignificant. Then they take the average of statistically insignificant data. Great… The range of those measurements more closely represents measurement error. All I ask is to measure 1 part 5 times, and report the highest and lowest diameter. After all, if you are really doing precision machining, 5 parts in a row really should not vary significantly. So, do not measure more parts,
    spend quality time with one part. As a quality professional, even if it were true that it was more work, one should recognize that less work does not trump correct. When I walk into a plant that is doing precision machining, they are so frustrated with X-bar – R SPC charts that they may be doing them, but they ignore them. I do not blame them. After training them on the correct method, they are relieved, they have a clearer idea of what the process is doing, how to control it, and how to improve it. No wonder SPC gets a bad rap – people are trying to rubber stamp the wrong charting methodology creating havoc.

    #162840

    I’m not sure where to start on this other than to wonder why you are addressing a 3-year old topic, but there are so many issues with the note that I feel compelled to add a response.
    On the use of SPC for “zig-zag” type machining data, there are ways to address this statistically, and not just using a guardband approach as you have suggested.   Perhaps the most direct way is to develop a regression model of the degradation in the machining behavior, and then use the model as a way to remove the trend in the data.  The resultant chart plots the residuals after the model has been fit.  I have done this technique and know that it can be successful.  There are other solutions as well.
    As to your knowledge of SPC (and Statistics), it appears to be rather superficial.  For example, X-bar-R charts can certainly be applied to Uniform data, or for that matter, most any other distribution.   In fact, in many cases this helps as the mere act of forming a subgroup for non-Normal data and taking the average develops a Normal distribution via the Central Limit Theorem.
    Also, control limts don’t need to be at 75% of the tolerance, that is just your  expectation to limit adjustments while remaining inside the tolerance band.   They should be whereever appropriate analysis of the data suggests they should be.  That is the basis of empirically determined control limits.
    The first couple of sentences replicated below in the “What Chart Works with Precision Machining” section are not correct either, at least in a general sense.  “I must add that X-bar-R charts are the worst for true uniform distributions. X-MR would be a little better. Again, the mean means nothing in the uniform distribution – both for the population or the sample, and there should be virtually no discernable variation between 5 consecutive parts, unless you are shredding up tools.”  

    #162842

    I am pretty sure I know where to start on this response. 
    First, I am responding to a three year old topic because it needs to be addressed three years later.  There may be many more of these buried in the forum – this one got lucky.  Consider it a probability function.
    Now I would like to address your rather superficial your knowledge of SPC (and Statistics).  First, you pose the regression model of the degradation in the machining behavior, and then use the model as a way to remove the trend in the data.  That approach ignores two major common causes to try to find lesser “normal” causes.  It totally ignores the tool wear itself.  Yes, it is a common cause, as it affects all parts – unlike tool breakage that is a special cause that affects only some parts.  It is a major cause that needs tracked and needs adjustment when the data reaches a control limit.  Second, you ignore another major common cause – the measurement error.  For example, if you are charting a round feature, you measure one diameter.  One diameter out of an infinite number of diameters is not only statistically insignificant (unless you are a lottery winner) – but totally ignores roundness issues.  Any variation you chart will be roundness error, not process variation.  X-bar R, X-MR or X trend -R techniques ignore this fundamental error.  Also, the use of statistics designed for the Gaussian curve encourages overcontrol – another mistake. 
    Unfortunately you may believe the trend technique is successful, but by ignore to major contributors to total variation, it really lacks the robustness of X hi/lo-R which captures and controls them very readily. 
    As far as the Central Limit Theorem – you forgot to read the fine print. The central limit theorem (CLT) states that the re-averaged sum of a sufficiently large number of identically distributed independent random variables each with finite mean and variance will be approximately normally distributed (Rice 1995).The continuous uniform distribution that arises from tool wear is neither random nor independent. So, CLT must not be assumed to apply to every distribution.
    As far as your claim to have used the regression model technique and know that it can be successful – rest assured I have used the X hi/lo-R technique and found that not only was it successful, it was easy to use and understood very readily by machine operators that were charged with its use.

    #162843

    It is possible I don’t fully comprehend your specific situation entirely, however you’re still making some incorrect statements such as this one: 
    “First, you pose the regression model of the degradation in the machining behavior, and then use the model as a way to remove the trend in the data.  That approach ignores two major common causes to try to find lesser “normal” causes.  It totally ignores the tool wear itself.  Yes, it is a common cause, as it affects all parts – unlike tool breakage that is a special cause that affects only some parts.  It is a major cause that needs tracked and needs adjustment when the data reaches a control limit.
    The modeling of which I speak removes the primary wear effect on the data, allowing the residuals to be analyzed.  This is the major component of variation in this situation and what causes the primary difficulty on any chart that works off the the original data.  Your statements to the contrary appear to me to be incorrect.

    #162844

    My point is that removing the tool wear and analyzing the residuals is rather academic – as it ignores two most significant common causes to seek out lesser issues to monitor for control.  There is really nothing incorrect about that. Use of the original data deals with the two major common causes (as well as any lesser ones buried in the data) and assures all common causes of the dimension are controlled.   X hi/lo-R does this rather simply and elegantly.

    #162845

    What nonsense

    #162846

    For starters Bob, tool wear charts have been addressed for at least 60
    years. Duh.Setting control limits to 3/4 of tolerance? How stupid is that? Depends
    on the inherent capability of the specific tool and machine, doesn’t it?I suspect you have never actually done precision machining just give
    advice from some academic pulpit. Is that correct?

    #162847

    Of course not.  I developed this technique in the field, and very successfully utilized it to control CNC machining and grinding to the micron level. 
    Setting control limits to 75% of tolerance is very clever, actually, as it is based on the correct continuous uniform distribution that the sawtooth curve tool wear and its associated adjustment creates.  Theoretically, the continuous uniform distribution would allow you to run to the specifications, but realizing that you are controlling total variation (such as measurement error) and you have sampling error (you can not catch the adjustment precisely at the control limit) the 25% ‘guardband’, if you will, allows for that. It also provides protection for the special cause of tool breakage. 
    The inherent capability of the specific tool and machine has nothing to do with the control limits – the adjustment for the tool wear is the major common cause – if you are doing precision machining.  If the variation of the machine due to loose ways or bad bearings exceeds the variation from the tool wear, then you simply are not doing precision machining.  You can use any of the charting based on the normal distribution, as that error is very natural and will be Gaussian.

     

    #162854

    Your 75% is dumb. Your guard-band is going to be determined by
    your process and measurement capability.Your claim of a uniform distribution is wrong as well. You have two
    distributions – one is the inherent distribution, the other is the
    distribution of the degradation of the tool. To assume it’s constant
    (uniform) should be verified.Your claim of “clever” is rather self serving, don’t you think? Your
    method is only better than just throwing a x-bar, R chart at it.
    Again, the subject of tool wear charts have been documented for
    over 60 years.

    #162856

    “Your 75% is dumb. Your guard-band is going to be determined by your process and measurement capability.”
    75% gives you 1.33 capability.  If you want higher capability, you can go to a larger ‘guardband’, but it increases overcontrol and has no economic benefit. If you performed your gage R&R and ensured that your gage was capable, then that distribution becomes statistically insignificant to the issue.  The hi/lo approach eliminates the measurement error that makes X-bar-R, X-MR and X-trend-R useless for precision machining.  
    “Your claim of a uniform distribution is wrong as well. You have two distributions – one is the inherent distribution, the other is the distribution of the degradation of the tool. To assume it’s constant (uniform) should be verified.”
     That is absurd.  If it is precision machining, and you are controlling the process correctly you will get the sawtooth curve, which yields the uniform distribution.  If it is precision machining and you have a normal distribution, you are out of control.  Most likely the person operating the machine is making adjustments with regard to the control limits.  At that point the person has become the process, not the machine.  That is out of control.
    “Your claim of “clever” is rather self serving, don’t you think? Your method is only better than just throwing a x-bar, R chart at it. Again, the subject of tool wear charts have been documented for over 60 years.”
    Dwelling on old techniques that have been shown to be effective is your option. Suit yourself.  This has been used for over 10 years very successfully. Arguments as “dumb” or “self-serving” are pretty hollow and show that lack of depth your consideration of the technique really is. 
    Show me one technique other than x-bar -R that can provide the following information in precision machining. The amazing thing, however, is how incredibly effective the X hi/lo-R charting methodology is.  Using the OD grinding example, you get the following benefits:
     
    1. You are charting the full GD&T of the characteristic – not just a statistically insignificant sample of data (one diameter out of an infinite number of diameters one finds in a circular feature.
     
    2. The limits tell the operator when to make an adjustment – a significant signal, indeed.  They are instructed not to make any adjustments until the curve hits the control limit, reducing overcontrol.
     
    3. The slope provides a very accurate representation of the tool wear rate – the most important function for continuous improvement, not compressing the control limits!
     
    4. The range is the roundness.  The roundness has been shown to be a leading indicator of tool wear.  As the tool wears the cutting becomes less efficient causing the roundness to increase.  A control limit on the roundness gives a great signal to change the tool!
     
    5. It makes sense – it truly is a clear visualization of what the process is doing.  Machine operators can easily make decisions based on the signals it provides.  X-bar and R charts simply frustrates them.  Rightfully so.
     
    Can you get this information from any other charting methodology?  And, if not, what is the point of employing the others in precision machining?

    #162857

    The bold statement below is a correction to the previous post:
    Show me one technique other than x hi/lo-R that can provide the following information in precision machining. The amazing thing, however, is how incredibly effective the X hi/lo-R charting methodology is.  Using the OD grinding example, you get the following benefits:
     
    1. You are charting the full GD&T of the characteristic – not just a statistically insignificant sample of data (one diameter out of an infinite number of diameters one finds in a circular feature.
     
    2. The limits tell the operator when to make an adjustment – a significant signal, indeed.  They are instructed not to make any adjustments until the curve hits the control limit, reducing overcontrol.
     
    3. The slope provides a very accurate representation of the tool wear rate – the most important function for continuous improvement, not compressing the control limits!
     
    4. The range is the roundness.  The roundness has been shown to be a leading indicator of tool wear.  As the tool wears the cutting becomes less efficient causing the roundness to increase.  A control limit on the roundness gives a great signal to change the tool!
     
    5. It makes sense – it truly is a clear visualization of what the process is doing.  Machine operators can easily make decisions based on the signals it provides.  X-bar and R charts simply frustrates them.  Rightfully so.
     
    Can you get this information from any other charting methodology?  And, if not, what is the point of employing the others in precision machining?
     

    #162859

    Wrong on the capability and wrong on uniform disribution.Ever heard of GR&R? Do you have a clue there is a distribution around
    the machining in addition to a distribution around degradation (not
    necessarily uniform).And yes your “clever” is self serving. I think you have never really done
    this.

    #162861

    “Wrong on the capability and wrong on uniform distribution.”
     
    Nope, I have the data to prove it, and your guess about it (rather than an attempt to verify the process in a real situation) is lacking a rigorous consideration.  “Doing things the way we always have” is an approach that subverts continuous improvement.  Are you sure you are on the right site?
     
    “Ever heard of GR&R?”
     
    Yes I have, and that is how you prove you have the correct gage system for the job.  In fact, if you are doing SPC, you should have an ndc >10 to insure that you have the statistical resolution to make the normal distribution that it contributes statistically insignificant to the total variation – as well as 10 categories within the range of the control limits.  It sounds to me that you have heard of it, but do not understand its impact and how to address it correctly.
     
     
     “Do you have a clue there is a distribution around the machining in addition to a distribution around degradation (not necessarily uniform)”.
     
    Any distribution exists other than the uniform distribution is evidence must be less than the tool wear, or you are not doing precision machining or it is not in control.  And, perhaps that is your issue – you are not familiar with properly controlled precision machining.  You clearly missed the definition in the first post.
     
    “And yes your “clever” is self serving. I think you have never really done this.”
    I can easily make the same assumption about you via your comments – but I have a better theory.  You have done it- but not correctly.  Not only that, you are likely spreading your incorrect approach to other poor victims.  Don’t feel bad – you are in good company.

    #162862
    Profile photo of Chad Taylor
    Taylor
    Participant
    Reputation - 0
    Rank - Aluminum

    Bob I have to agree with Stan on this one, 75% of tolerance as a control limit is the dumbest thing I have ever heard of Especially in precsion machining at the Micron level. I bet the Shop floor supervisors just love you…………….
    Also love your cut and paste abilities
     

    #162863

    Bad theory. I knew how to solve this problem two decadess ago.

    #162864

    Bad theory.
    I knew how to solve this problem two decadess ago.
    I can tell by your well framed arguments. Thank you.
    Bob I have to agree with Stan on this one, 75% of tolerance as a control limit is the dumbest thing I have ever heard of Especially in precsion machining at the Micron level. I bet the Shop floor supervisors just love you…………….
    Yes, shop floor supervisors do appreciate it, as the control limit calculations for X-bar/R, X-MR and X trend-R are for the wrong distribution and artificially compress the control limits for no valuable reason.  The variation they track is also incorrect – they track measurement error – totally masking any process variation.  Pretty useless.  The 75% is derived fom the process spread defined by the uniform distribution.  By the way, it is very easy to control a process to the micron level with X hi/lo-R – in fact I used it to a 6 micron tolerance with no problems. 
    Also love your cut and paste abilities
    Yes, I agree, this is not the most friendly forum software.  But, we make do.  Also, cut and paste is efficient…lean, if you will.  Heard of it?

    #162865

    Mr DOEring,Tool wear charts have existed as long as you and me.Your “cleve” negates understanding GR&R and the inherent
    capability.I know you jumped on here thinking you would impress someone,
    but you are just one of a long line of much ado about nothing.Your well framed argument is “hey look at me, I think I am clever”. I
    think you have a lot to learn.

    #162866

    Mr Doering,
    Tool wear charts have existed as long as you and me.
     
    So much for continuous improvement. 
     
    Your “clever” negates understanding GR&R and the inherent capability.
     
    Sorry – it is easy to prove the opposite is true. But, since you wish to attempt to dismiss it based on academic puffery rather than factual data, then I doubt there is much that will help you.
     
    I know you jumped on here thinking you would impress someone, but you are just one of a long line of much ado about nothing.
     
    It is clear my arguments went over your head. Sorry.
     
    Your well framed argument is “hey look at me, I think I am clever”. I think you have a lot to learn.
    Your abject lack of argument is amusing at best.  Using “dumb” or “we have always done it this way” as substance of your retort is deep.  Good luck to you.
     

    #162867

    Mr. DOEring,You have solved a non problem. Congratulations.You have given incomplete, dangerous advice.Again, congratulations.I have given you every opportunity to explain the assumptions and
    limitations to your advice and you have chosen to stick with I am
    right with my arguments (I haven’t seen any defending your
    assumptions or disregard of advice for MSA or inherent capability).You are just a hack.

    #162868

    Mr. Doering,
    You have solved a non problem.
     
    Apparently you have no idea.
     
    Congratulations.
    You have given incomplete, dangerous advice.
     
    You are correct, as a forum response it may not be thorough enough to boil it down for you.  Sorry about that. As far as dangerous – you have not shown that to be the case at all.
     
    Again, congratulations.
    I have given you every opportunity to explain the assumptions and limitations to your advice and you have chosen to stick with I am right with my arguments (I haven’t seen any defending your assumptions or disregard of advice for MSA or inherent capability).
    You must be scanning instead of reading, because I explained the MSA issue concerning how using a gage with adequate ndc make gage error statistically insignificant in charting, and how the X hi/lo-R technique eliminates measurement error (not gage error) by addressing every diameter, not just a statistically insignificant sampling (as in one) of diameters in a measure of a round feature.  I also address the two key causes of variation in precision machining that you choose to ignore completely. 
    You are just a hack.
    I think your personalization of the discussion is very unprofessional, and I will let that speak for itself.

    #162874

    Bob,Look to the very first answer given by JimH. It is better advice than yours

    #162875

    Bob,
    Look to the very first answer given by JimH.
    It is better advice than yours
     
    Interesting, in that the algorithms for automated adjustment are very similar to my approach, except that they do not address measurement error (not gage error) from roundness and are actually overcontrol.  That is fine for machines, but a wasted effort for manual control.
     I established the error of using the sloped control limits (X trend-R) in my previous post.  They ignore two key contributing common causes that the X hi/lo-R technique does address.  It also holds none of the benefits of the X hi/lo-R technique that I have identified.  That being the case, I do not see how that advice is better.

    #162877

    Bob, let it go. This is Stan’s entire life. He must believe he is the king of this forum and knows more than anybody else in the world. He has nothing else in life. Be kind and let it end.

    #162878

    Bob,I’ve been trying to follow this thread. I don’t understand why tool wear or real time measurements are important. When I use a lathe, I just cut until I reach the end stop. When it takes too long or it doesn’t cut cleanly, I sharpen the tool.

    #162880

    Bob, let it go. This is Stan’s entire life. He must believe he is the king of this forum and knows more than anybody else in the world. He has nothing else in life. Be kind and let it end.
    Thank you.  I can live with that.

    #162883

    Wow…
    Bob,
    Are you being serious? I mean grand-standing aside, your comments are way of base. As another Poster stated, you offer no substantive defense, or lnked logic behind your offerings.
    Anybody even thinking about going along with Bob’s take/advice had better get their resumes in order and depending on the application, you might consider retaining good counsel. man-oh-man, Bob.really?
    seriously?
    wow.

    #162884

    Wow… Bob, Are you being serious? I mean grand-standing aside, your comments are way of base. As another Poster stated, you offer no substantive defense, or lnked logic behind your offerings. Anybody even thinking about going along with Bob’s take/advice had better get their resumes in order and depending on the application, you might consider retaining good counsel. man-oh-man, Bob.

    really? seriously? wow.
    And I suppose you are also a fan of the X-MR trend chart for tool wear? 
     

    #162885

    I’m not fan of X-MR, but it doesn’t make a fan of your gross
    oversimplification. Don’t know about HeeBee

    #162889

    My beef has to do with the ham-fisted approach and the disregard for 60 yrs+ worth of actionable sources.
    Couldn’t care less about X-MR…

    #162890

    So essentially what you have here is just a dumber version of pre-control?

    #162891

    So essentially what you have here is just a dumber version of pre-control?
    Not really.  This approach is supported by the fact that the distribution from the sawtooth curve that is the main variation from precision machining is the continuous uniform distribution, and the distribution’s statistics support it.  That makes it smarter than pre-control.
     

    #162893

    The sawtooth is only the major source because you are giving it so
    much room. The real question is how much room dies the underlying
    distribution take.Your approach is ignorant of process knowledge and quite possibly
    could be costly in downstream ops.

    #162895

    The sawtooth is only the major source because you are giving it so much room. The real question is how much room lies the underlying distribution take.
    If the distributions of the underlying causes of variation are greater than the tool wear, you are not doing precision machining.  The underlying distribution you pick up in traditional methodologies is the measurement error (not gage error) from not accounting for roundness or parallelism.  If you want to chase that false variation, then it is an interesting notion, but of no value to the operation or the process.
    Are you capable of preparing a correct total variation equation, with the factors in order of the size of their effect and their associated distributions?
    Your approach is ignorant of process knowledge and quite possibly could be costly in downstream ops.
    You will need to explain that assertion further.  If I am controlling roundness or parallelism – which clearly going to affect downstream processes – as well as maintain all diameters or lengths in the parts within the properly developed process tolerance zone specified – what risk is left? And what protection does the traditional approach have for either of those conditions?
     

    #162897

    Ghost Rider: Tower, this is Ghost Rider requesting a flyby.
    Air Boss: That’s a negative, Ghost Rider, the pattern is full. Mr. Doering, I will land if you will. No more fly bys, no more sonic booms, the pattern is full and you have filled it. Please, sir, by all that is good and Holy, let it end; I beg of you. You win! You are right! Consider the rest of the plebes battered and beaten, but please just let it go.Ghost Rider out

    #162898
    Profile photo of Chad Taylor
    Taylor
    Participant
    Reputation - 0
    Rank - Aluminum

    If your customer uses short ppk studies as a goods inwards inspection release, then by selecting at random through the full run, then your product might be caught as apparently non-compliant. Good Luck with your customer

    #162899
    Profile photo of Chad Taylor
    Taylor
    Participant
    Reputation - 0
    Rank - Aluminum

    Bob has been spreading this non sense for a while now.http://elsmar.com/Forums/showthread.php?t=19595

    #162900

    Such erudition does little good to the folks on the floor, and those downstream, for whom the analysis is supposed to be of use.Perhaps, though, there is more than one way to skin a cat; or at least there may be more than one way to keep the process of skinning said cats in control. I’m done.

    #162901

    If your customer uses short ppk studies as a goods inwards inspection release, then by selecting at random through the full run, then your product might be caught as apparently non-compliant. Good Luck with your customer
    If the customer is automotive, I send them to the following passage in AIAG PPAP 4th edition that explains why that approach is statistically invalid – since the distribution is non-normal:
    2.2.11.5 Processes with One-Sided Specifications or Non-Normal Distributions
    NOTE: The above mentioned acceptance criteria (2.2.11.3) assume normality and a two-sided specification (target in the center).  When this is not true, using this analysis may result in unreliable information.
    If they are not automotive, I explain it using the same logic.

    #162902

    Ghost Rider: Tower, this is Ghost Rider requesting a flyby. Air Boss: That’s a negative, Ghost Rider, the pattern is full.
    Mr. Doering,
    I will land if you will. No more fly bys, no more sonic booms, the pattern is full and you have filled it. Please, sir, by all that is good and Holy, let it end; I beg of you. You win! You are right! Consider the rest of the plebes battered and beaten, but please just let it go.
    Ghost Rider out
    I really enjoy the culture of sarcasm  and drama I see on this site.  I have picked up some mildy humorus bits from some of the more baseless responses.  I knew normal-centrics were an odd bunch to begin with – but, plebes aside,  just in case there are some real quality professionals that are willing to look beyond what they have been rubber stamping for years, there is some meaningful discussion to be had. 
    Carry on, lad.

    #162903

    Sir, Carry on, Aye Aye, Sir.
    Left foot, Sir.
    Right foot, Sir.
    One, Sir, don’t bend the knee, Sir.
    Two, Sir, heels on line and touching, Sir.
    Sir, Carry on, Aye Aye, Sir.

    #162904

    What has tool wear got to do with maintaining all diameters or lengths in the parts within the properly developed process tolerance zone specified?I don’t know anyone who sets up a lathe machine time! Tool wear has nothing to do with the diameter of a part only the cutting speed.

    #162906

    Okay, Bob,Since you insist on not letting it go; here is one for you. If you can do this then I will bow at your altar and sing your praises from on high. If, however, you cannot do this, then I will assume that you are just another half-wit, self-deluded, egotistical, self-esteem-lacking individual who has the need to convince himself he is correct in something to make up for the myriad shortcomings he finds within himself…Here goes…Explain your most salient point in plain English, in less than five sentences, without all the grandiosity and $3 words and ideas, keeping the hyperbole to a minimum if you would. To be even clearer, pretend I’m a novice Six Sigma practitioner, but that I also hold $1 million in my hand, which will be yours if you can make your point while being Clear, Concise, Coherent, and by all means Cogent.It’s easy to spout a plethora of words in your argument; it’s much more indicative of intelligence to speak with as few words as possible while still making your point.Ghost Rider hopes Bob is up to the task

    #162907

    What has tool wear got to do with maintaining all diameters or lengths in the parts within the properly developed process tolerance zone specified?
    If you have set your speeds and feeds and leave the machine alone – as in do not overcontrol – the only significant variation you should have is the tool wear.  Consider an OD – as you machine part – with set speeds and feeds, correct coolant rates, etc all set – after part the tool still wears.  As the tool wears the OD increases.  At some point you will have to make an adjustment – and if you are smart, you will adjust it to the lower limit to allow the tool to continue to wear.    
    If you have so much random variation that you can not see the tool wear, then A) the operator has become the process from overcontrolling; B) your machine is so sloppy that it can not be considered precision machining or C) the variation you are detecting is from the measurement error due to not accounting for roundness (for round features) or parallelism (for lengths).
    Tool wear has nothing to do with the diameter of a part only the cutting speed.
    Most people adjust for tool wear with offsets, not speed adjustment.
     
     

    #162910

    C’mon Bob. You have time to vomit a load of rubbish; no time for a few simple words? Let me guess, now you don’t have time? Whatever? You are what everyone on this forum thinks you are; a hack. Bob, do us a favor, just go away.

    #162914

    Normal-centrics?How about BS-adverse?

    #162915

    Normal-centrics?

    How about BS-adverse?
    That appears to be mutually exclusive, son.

    #162916

    Hack!

    #162929

    Hack!
    Might want to try some throat spray for that. Should have gotten a flu shot.

    #162932

    I think part of the problem you are having Bob is that you are providing anecdotal evidence to support your method rather than sound statistics.  You continuously mention the normal distribution as if it is a prerequisite for SPC imlementation (it isn’t) and you focus on Xbar-R/X-MR charts as if anyone with any bit of statistical background would package them as universally applicable.
    The point of SPC is to evaluate the current state of the process versus what is expected to occur.  Your approach bases analysis on arbitrary percentages of specification limits and therefore cannot maintain the sensitivity that an appropriate SPC methodology can.  This does not mean that your method will be completely ineffective in protecting the consumer, it just means that you stand to learn very little about your process.

    #162934

    Bob,I don’t think this is correct. The OD is set by the position of the end stop.Flat distributions have nothing to do with saw tooth tool wear, it is due to the use of a single first off.

    Regards,
    Alto

Viewing 62 posts - 1 through 62 (of 62 total)

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

Login Form