Tweet
Coordinate Measuring Machine Process Control
Since I retired, I have been thinking about putting some of my ideas about cmm’s and process control into a Word.doc that might shed some light on what process control can do to improve your bottom line. It will have to be somewhat lengthy and I do not know of another way to do it.
When we think about processes that are either in or out of control, we often only consider the cmm. Not the processes that provide the parts you measure. A proper application of some basic data collection and programming can help you determine if a MEASURING process or a MANUFACTURING process is out of control.
Many of you know that I am an avid shooter. I try to identify and reduce the impact of any identifiable variable in the process of shooting a bullet downrange. What I propose to do here is give you a couple of examples to help you develop a process control scheme to tell you when your measuring machine needs attention as well as when your manufacturing process needs attention.
The first case, in my opinion, needs to address the cmm. If you do not know how well it performs, you are doing a disservice to both your manufacturing supplier as well as your down stream customer.
Many of you have and have asked about a paper I presented to NCSL a few years ago on process control and coordinate measuring machines. I am sure some of you have it. I do not have it any more but it should be available from either NCSL or the Measurement Science Conference held in Albuquerque, New Mexico in 1998. In truth, I probably DO have a copy of it SOMEWHERE…
As many of you know, my scheme for determining the performance level of the cmms at the Boeing plant in Mesa consisted of using calibrated tools at A0/B0, A90/B0, A90/B90, A90/B-90, and A90/B180.
By finding the location and setting zeroes in the calibration sphere with the A0/B0 tool, I could measure the size and location in a looped program to gather the X,Y,Z, and D coordinates with all five tools.
What that did for me was to collect 5 values for each of the X,Y,Z,and D values for EACH run of the program. Those 5 values gave me a range and a standard deviation and could be plotted on an XBar and R chart.
I would run the program in a loop 25 times and collect that data over the short term to give me a look at the short term performance of the machine.
If the artifact was aligned along the Y axis of the machine, some very interesting things could be observed. We had the artifact with the little 45 degree bend in the shank. What we noticed over the short term was that if the temperature in the lab changed, the Z coordinate of the sphere location also changed as well as the Y coordinate. If artifact was placed such that the 45 degree bend was toward Y minus and the temperature rose, the Y coordinate of the sphere ALSO moved Y minus and the Z coordinate moved Z plus.
It showed up plain as day on the XBar and R charts. This variation over the short term was never as great as the variation over the long term. We would run the looped program 25 times usually before our machine was calibrated and then we would run the same program AFTER the cmm was calibrated to see if the machine was improved by calibration or not. We used that data to justify calibration interval adjustment.
We began collecting that data at the factory when the machine was new and before delivery. We then collected similar data after delivery and installation.
We also instituted a program to run the program once a week to collect a data point for the long term charts.
We had XBar and R charts for the 3 coordinates of the sphere as well as the diameter. We had 8 charts in all. The basic size of the artifact we determined by direct comparison to gage blocks and a Standard Measuring Machine.
Because we had both long term and short term charts, there was a LOT of data. It took quite a bit of work to maintain the charts but having the data, was far better than just having an opinion.
Not surprisingly, the long term charts were not as good as the short term charts. The reason for that is the variables involved in making the measurements had more time to make an influence that was detectable on the charts. For example if you made one run of the program at 66 degrees F. in the lab and another a week later at 68 degrees F., the results will be different. When you are looking at the printout and you are seeing .000030 in. one time and -.000030 in. the next time, most humans only see the number and NOT the sign.
The range therefore is .000060 in. Having the data plotted is MUCH better than trying to remember what the printout from a week or so indicated.
Keep in mind; this does not address every variable that influences your reading. It merely addresses the most common.
I am not going to try to bore you to death in one fell swoop. I will let you digest this and then tomorrow, I will try to put some thoughts down about getting your process control program to not just tell you if you are in control. Your machine probably is NOT in control. But the chart CAN tell you what to look for to make it BETTER. One of the things I will try to address, is the use of a test part to help you determine whether the measuring process is out of whack or whether the manufacturing process needs attention. Believe me, you do not want to tell manufacturing they have a hosed up process and then determine the cmm is the source of the problem. That will only have to occur once and you will have the devil to pay to redeem your self.
When we think about processes that are either in or out of control, we often only consider the cmm. Not the processes that provide the parts you measure. A proper application of some basic data collection and programming can help you determine if a MEASURING process or a MANUFACTURING process is out of control.
Many of you know that I am an avid shooter. I try to identify and reduce the impact of any identifiable variable in the process of shooting a bullet downrange. What I propose to do here is give you a couple of examples to help you develop a process control scheme to tell you when your measuring machine needs attention as well as when your manufacturing process needs attention.
The first case, in my opinion, needs to address the cmm. If you do not know how well it performs, you are doing a disservice to both your manufacturing supplier as well as your down stream customer.
Many of you have and have asked about a paper I presented to NCSL a few years ago on process control and coordinate measuring machines. I am sure some of you have it. I do not have it any more but it should be available from either NCSL or the Measurement Science Conference held in Albuquerque, New Mexico in 1998. In truth, I probably DO have a copy of it SOMEWHERE…
As many of you know, my scheme for determining the performance level of the cmms at the Boeing plant in Mesa consisted of using calibrated tools at A0/B0, A90/B0, A90/B90, A90/B-90, and A90/B180.
By finding the location and setting zeroes in the calibration sphere with the A0/B0 tool, I could measure the size and location in a looped program to gather the X,Y,Z, and D coordinates with all five tools.
What that did for me was to collect 5 values for each of the X,Y,Z,and D values for EACH run of the program. Those 5 values gave me a range and a standard deviation and could be plotted on an XBar and R chart.
I would run the program in a loop 25 times and collect that data over the short term to give me a look at the short term performance of the machine.
If the artifact was aligned along the Y axis of the machine, some very interesting things could be observed. We had the artifact with the little 45 degree bend in the shank. What we noticed over the short term was that if the temperature in the lab changed, the Z coordinate of the sphere location also changed as well as the Y coordinate. If artifact was placed such that the 45 degree bend was toward Y minus and the temperature rose, the Y coordinate of the sphere ALSO moved Y minus and the Z coordinate moved Z plus.
It showed up plain as day on the XBar and R charts. This variation over the short term was never as great as the variation over the long term. We would run the looped program 25 times usually before our machine was calibrated and then we would run the same program AFTER the cmm was calibrated to see if the machine was improved by calibration or not. We used that data to justify calibration interval adjustment.
We began collecting that data at the factory when the machine was new and before delivery. We then collected similar data after delivery and installation.
We also instituted a program to run the program once a week to collect a data point for the long term charts.
We had XBar and R charts for the 3 coordinates of the sphere as well as the diameter. We had 8 charts in all. The basic size of the artifact we determined by direct comparison to gage blocks and a Standard Measuring Machine.
Because we had both long term and short term charts, there was a LOT of data. It took quite a bit of work to maintain the charts but having the data, was far better than just having an opinion.
Not surprisingly, the long term charts were not as good as the short term charts. The reason for that is the variables involved in making the measurements had more time to make an influence that was detectable on the charts. For example if you made one run of the program at 66 degrees F. in the lab and another a week later at 68 degrees F., the results will be different. When you are looking at the printout and you are seeing .000030 in. one time and -.000030 in. the next time, most humans only see the number and NOT the sign.
The range therefore is .000060 in. Having the data plotted is MUCH better than trying to remember what the printout from a week or so indicated.
Keep in mind; this does not address every variable that influences your reading. It merely addresses the most common.
I am not going to try to bore you to death in one fell swoop. I will let you digest this and then tomorrow, I will try to put some thoughts down about getting your process control program to not just tell you if you are in control. Your machine probably is NOT in control. But the chart CAN tell you what to look for to make it BETTER. One of the things I will try to address, is the use of a test part to help you determine whether the measuring process is out of whack or whether the manufacturing process needs attention. Believe me, you do not want to tell manufacturing they have a hosed up process and then determine the cmm is the source of the problem. That will only have to occur once and you will have the devil to pay to redeem your self.
Comment