SBS - The Best Value in QMS software

Analysis of non-normal stratified data for cpk/ppk? Rupture test

N

nwillnow

#1
Hello all, I'm having a bit of a problem.

I have a load of data (about 200 datapoints per week) taken during finished goods testing (1 in 40). The test is a rupture test with a specification of 294psi, and is performed by doing a staircase test with 30psi increments. This leads to a lot of data pooled at 300, 330, 360, and 390.

My problem is that I've been asked to look at all of this data across two of our manufacturing plants to acertain some sort of idea about the capability of our processes to make product meeting this spec, and of course the data is in no way normal. I've done just about everything I know of in Minitab (box-cox trans, johnson trans, distribution identification for normality, wiebull, exponential, log, etc), and nothing is giving me a normal dataset.

I've attached a sample dataset and some of the analysis I've performed on it so far. Am I stuck looking at this as attribute data due to the lack of resolution? Does anyone know of anything else I can do to understand the capability of my process?
 

Attachments

Elsmar Forum Sponsor

Tim Folkerts

Super Moderator
#2
Re: Analysis of non-normal stratified data for cpk/ppk?

One quick question to start.

The data is mostly pooled at multiples of 30 psi, but some data fall in between the standard values. How does this occur?

Was the "wrong" pressure incorrectly applied? Then you might need to check the proceedures

Did the disk rupture as you were ramping up the pressure? If you have the capability to ramp up the pressure and to measure the pressure along the way, then why not do that as the standard test?


Understanding the testing might help inspire some thoughts about improving the test and/or the analysis.

Tim F
 

Statistical Steven

Statistician
Staff member
Super Moderator
#3
Mr

Tim Folkerts said:
One quick question to start.

The data is mostly pooled at multiples of 30 psi, but some data fall in between the standard values. How does this occur?

Was the "wrong" pressure incorrectly applied? Then you might need to check the proceedures

Did the disk rupture as you were ramping up the pressure? If you have the capability to ramp up the pressure and to measure the pressure along the way, then why not do that as the standard test?


Understanding the testing might help inspire some thoughts about improving the test and/or the analysis.

Tim F
Tim, you beat me to the punch. Looks like the test itself needs to be look at before we can assess the process.
 
N

nwillnow

#4
I had a feeling that would be the first thing people asked about, I should have covered that in the first post.

The test ramps to a pressure, the pressure is sustained for a set time, then ramps to zero, then to the next stair, etc. So yes, there is a capability for the part to fail during the ramp, and that is what you are seeing when looking at datapoints that look like 339psi.

The issue with test method is that this is an approved medical device with a specific standard test method developed internally and filed along with the product. More unfortunate however is my inability to speak to it's creation and the reason for the incorporation of a fatigue-like test along with the ramp.

I'm afraid the answer you'll give is that I need to alter this test method for my purposes, which will mean creating new data, and given this is a destructive test I'll need to build units specifically to test. But if that's the only way to truely attain variable data that will give me an idea of the capability then so be it.

That being said, if the assumption is made that I can't get funding to cover building extra units for testing, and I need to draw a conclusion from my current test with the historical data, is there any other method I can put to use?
 

Tim Folkerts

Super Moderator
#5
I can think of several things I might try, but I don't know that they would pass muster as a formal test for a medical device.

There is "step-stress" analysis used in reliabilty analysis, but it is more for estimating operating lifetime at "normal" stress level. See
http://www.weibull.com/knowledge/weibullcom_index.htm#S

Do you have time data as well? Do you record when the parts fail? That would give a little bit more info. Do you hold at each pressure long enough for basically all the part to fail that are going to fail? In such a case, the parts that fail at, say, 360 psi, really are parts that would fail somewhere between 330 and 360 if you could measure more appropriately. Listing them all at 360 isn't really realistic. You could assume (yes, a very dangerous thing!) that the "true" failure locations are distributed between 330 and 360. If you picked a simple option like distributed uniformly between the two values, then your plots and analysis would look more typical. But again, this is going out on a pretty big limb.

A new test may be the best option at this point.


Tim F
 

Statistical Steven

Statistician
Staff member
Super Moderator
#6
Mr

nwillnow said:
I had a feeling that would be the first thing people asked about, I should have covered that in the first post.

The test ramps to a pressure, the pressure is sustained for a set time, then ramps to zero, then to the next stair, etc. So yes, there is a capability for the part to fail during the ramp, and that is what you are seeing when looking at datapoints that look like 339psi.

The issue with test method is that this is an approved medical device with a specific standard test method developed internally and filed along with the product. More unfortunate however is my inability to speak to it's creation and the reason for the incorporation of a fatigue-like test along with the ramp.

I'm afraid the answer you'll give is that I need to alter this test method for my purposes, which will mean creating new data, and given this is a destructive test I'll need to build units specifically to test. But if that's the only way to truely attain variable data that will give me an idea of the capability then so be it.

That being said, if the assumption is made that I can't get funding to cover building extra units for testing, and I need to draw a conclusion from my current test with the historical data, is there any other method I can put to use?
Given that the data is what the data is, I would suggest you use nonparametric methods to compare two sites. Use the Kruskal-Wallis or Friedman methods. If you need further assistance, please let me know.
 
N

nwillnow

#7
Thanks for your help and suggestions guys, and the quick replies. I think I'll try the nonparametric analysis for the line to line and site comparisons and might try smoothing the data for the probability of failure stuff the quality group is looking to me for (it's not for filing, merely engineering understanding of the process). Unfortunately the time of failure isn't logged in the test (actually the data is all logged manually ... pretty archaic), so I'm not sure i can avail of the step-stress analysis proposed. Actually, what I'm realizing through all of this is just how poor the test method is for our purposes. We aren't really trying to do any sort of accelerated test and our product doesn't need a lifecycle, so I don't get why the design group went with the step-stress in the first place.

But again, thank you both.
 
Thread starter Similar threads Forum Replies Date
K Non-Normal Data Analysis Literature, Websites, Books for Learning Quality Tools, Improvement and Analysis 2
S Capability Analysis of Non-Normal Data Statistical Analysis Tools, Techniques and SPC 23
J Is a t-test used in non-normal data analysis? Six Sigma 7
A Capability Analysis - Dealing with non-normal data in Minitab Using Minitab Software 8
A Non-replicable tests in MSA (Measurement Systems Analysis) Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 3
R Destructive Non Reproducibility MSA (Measurement System Analysis) Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 8
A Cnpk (Non-Parametric Capability Analysis) to assess whether the Process is Stable Capability, Accuracy and Stability - Processes, Machines, etc. 7
B Root Cause Analysis for Minor Non-Conformities identified in an Audit Nonconformance and Corrective Action 13
K Gap Analysis for a non ISO 9001 company ISO 9000, ISO 9001, and ISO 9004 Quality Management Systems Standards 8
S Root Cause Analysis Taxonomy - Categorising the reasons for non-conformances ISO 9000, ISO 9001, and ISO 9004 Quality Management Systems Standards 6
D Do all Non Conformance Reports need a detailed Root Cause Analysis? AS9100, IAQG, NADCAP and Aerospace related Standards and Requirements 13
S How to represent interactions in non-linear regression analysis in Minitab Using Minitab Software 5
B Minitab for Non-Quantitative Survey Analysis Using Minitab Software 1
M Root cause analysis on non-defect defect reports? Problem Solving, Root Cause Fault and Failure Analysis 1
A Non-conformance Root Cause Analysis Finding: Stupidity? Problem Solving, Root Cause Fault and Failure Analysis 11
D Need some Help on 8D fault tree analysis Problem Solving, Root Cause Fault and Failure Analysis 6
B How to exclude empty rows in full factorial design analysis? Using Minitab Software 0
B GR&R Destructive Data Analysis Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 11
H Risk Analysis and Probability of Occurrence ISO 14971 - Medical Device Risk Management 3
B Risk analysis for defective measuring or measuring equipment out of calibration General Measurement Device and Calibration Topics 2
P Benefit risk analysis on pFMEA ISO 14971 - Medical Device Risk Management 9
Q Process map Evaluation and Analysis Method Process Maps, Process Mapping and Turtle Diagrams 5
I QMS monitoring, measurement, analysis and evaluation requirement - Template ISO 9000, ISO 9001, and ISO 9004 Quality Management Systems Standards 9
S How to perform verification of the Statistical Analysis Software? Qualification and Validation (including 21 CFR Part 11) 7
J Starting a Weibull Analysis: Source on how to perform Weibull Reliability Analysis - Predictions, Testing and Standards 2
V ISO 10360-5: 2020 Gap analysis and Action plan Excel .xls Spreadsheet Templates and Tools 1
Q IATF audit - Root Cause Analysis results IATF 16949 - Automotive Quality Systems Standard 5
B Struggling with using the 5.6 version Ford Capacity Analysis Report APQP and PPAP 5
JoCam MDR Gap Analysis - Consultant fees EU Medical Device Regulations 10
silentmonkey Overall Benefit/Risk Analysis - Risk Management VS Clinical Evaluation ISO 14971 - Medical Device Risk Management 3
Awais Forklift Reliability Analysis Manufacturing and Related Processes 6
T Tactics for Root Cause Analysis Problem Solving, Root Cause Fault and Failure Analysis 4
lanley liao How to correctly understand the bullet list d) of 6.3 Analysis of Data for API Spec Q1 Oil and Gas Industry Standards and Regulations 7
R The term "Benefit Risk Ratio" in EU MDR, do I need to present benefit risk analysis as a RATIO Risk Management Principles and Generic Guidelines 4
B Two excellent examples of process capability analysis from Quality Magazine Capability, Accuracy and Stability - Processes, Machines, etc. 5
D Risk Analysis using Monte Carlo Simulation instead of Scoring and Heat Map Risk Management Principles and Generic Guidelines 2
A Attribute Analysis for colour matching scale Gage R&R (GR&R) and MSA (Measurement Systems Analysis) 5
M Recurrent event analysis software (python) General Auditing Discussions 2
E Normal Condition Hazards in Risk Analysis ISO 14971 - Medical Device Risk Management 3
M MDSAP and 13485:2016 gap analysis Quality Management System (QMS) Manuals 1
M Gap analysis on ISO 14971:2019 with previous revision ISO 14971 - Medical Device Risk Management 12
M Risk Analysis Flow - Confusion between ISO 14971 and IEC 62304 IEC 62304 - Medical Device Software Life Cycle Processes 8
R ECG Risk Analysis Standards ISO 14971 - Medical Device Risk Management 2
T Root Cause Failure Analysis - Not following Customer packaging Specification Problem Solving, Root Cause Fault and Failure Analysis 9
H ISO 13485:2016 Gap Analysis by NB ISO 13485:2016 - Medical Device Quality Management Systems 7
R Capability analysis - What is going on this chart? Manufacturing and Related Processes 15
adir88 Documenting Risk Control Option Analysis ISO 14971 - Medical Device Risk Management 8
MrTetris Should potential bugs be considered in software risk analysis? ISO 14971 - Medical Device Risk Management 5
K Analysis variation among the lots during cpk calculation. Design and Development of Products and Processes 1
DuncanGibbons Best practice for identifying "items" of parts for DFMEA analysis AS9100, IAQG, NADCAP and Aerospace related Standards and Requirements 2

Similar threads

Top Bottom