Process control for micro embossing : initial variability study

Abstract

The objective of this research is to study the dimensional variations in micro embossed parts. By measuring multiple parts produced with a fixed set of control inputs, it could be determined if the process is in statistical control, if the parts produced have any noticeable trends and if there are any other forms of deterministic or assignable disturbances that were overlooked. The experiment resulted in 50 sets of data consisting of 10 runs, resulting in 50 control charts. By using both classic SPC rules for and by observation it was determined that about 42/50 control charts show traits of a process that is stationary and in-control. In the remaining 8 charts, some distinct trends were observable. These trends were postulated to be produced by unintentional disturbances caused by the experimental procedure. There were some distinct observable trends in the results from the experiment. The first is the location and frequency of the occurrence of the 8 distinctive run charts mentioned above and 4 run charts that were also observed to have marginally trend-like characteristic though it seems more data points are required to make a more sound judgment. Out of these 12 run charts, 9 of them are from the left side of the part. Out of this 9, 5 of them are from the 3rd feature scale. This trend leads to a conclusion that the disturbance responsible for this behavior is localized to a graphic region of that part. The second observable trend is the strong correlation between feature scale size and the mean of the die-part difference. As the feature size increases, the mean difference between the die and part measurement increases. This can be because bigger features involve a larger volume of polymer material to form the shape and as the material

(cont.) shrinks after being embossing and cooled, the reduction in relative dimension is greater. The third observable trend is the strong correlation between the feature scale size and the standard deviation of the die-part difference. The variance in this dimension is larger as the feature size increases. As larger features produce a larger mean die-part difference, this might also produce an opportunity for a larger variation in this measurement.