A Quality Engineers Guide To Inspection Reports
Overseas vendors often provide scant data in inspection reports. This can be challenging for any quality engineer looking to approve a shipment of parts. If there is confidence that the measurement data is repeatable, reproducible, and stability exists in the process described by the given variable, then statistical tools can be applied to gather a glimpse of the lot population. In this guide, we outline some of the tools to determine the amount of acceptable parts in a sample size and how engineers can apply these tools to their own unique projects.
Applying The Laws Of Statistics
At first glance, the small sample size typically associated with overseas inspection reports may not confer a high degree of confidence in determining the characteristics of the whole population. Process stability indicated by previous inspection reports paired with the larger sample size associated with PPAP documentation can provide additional clarity to interpretation of the latest inspection reports. Assuming normal distribution of the data set and then applying statistical tools to extrapolate the limits of the population, can provide insight into potential risk associated with that production lot. Most production processes are like most things in nature in that they follow normal distribution. Certain processes like heat treatment follow normal distribution to a striking degree, processes like this are ideal for statistical analysis as highlighted herein.
Bell Curve of Normal Distribution (below)
Measure Twice And Cut Once
If your inspection report had eight samples that all fell within the specification limits, someone uninformed of the laws of statistics may assume that this is an indication of the lot being acceptable, but this would likely not be true. Sample standard deviation can be utilized to create a probability assessment of the whole production lot. In our example the data is skewed to the right of nominal with the mean coming in at 2.375 above nominal, so we will focus on the distribution towards the upper specification limit. Adding the sample standard deviation in multiples to the mean will generate numbers that can be quickly interpreted to find how much of the population is likely to fall within specification limits. Using these calculations on the sample data set below and applying the aforementioned assumptions, it could be said that based on the available data about 5% of the part population would likely be over the upper specification limit.
Sample Hardness Measurement Data (below)
Relaying The Data Back To Your Team
When reporting these results to management and team members that may not fully understand what variables influence the confidence that can be conferred to a given data set, it should always be noted that your conclusions are based on the data available. These statistical inferences based on the likely population distribution can be used as an effective risk mitigation strategy, especially when applied to key characteristics.