SPC Variations in automatic seam analysis

Seam analysis is presently utilizing technology to improve the quality control process. First, video-based solutions replaced the old projection systems and manual gauges. These video systems used cursors which had to be moved manually in order to measure distances. These systems were inefficient due to the time it took to measure each seam. Further complicating matters was the fact that some measurements were still needed to be measured using calipers and other external devices in order to be measured accurately.

Information collected using these instruments was somewhat difficult to analyze. Reports were collected, but SPC information was much more difficult to obtain.

Recent improvements in computer software and hardware allows seam analysis to be performed automatically. Such automatic systems are both efficient in their data collection (i.e., require less work from an operator and take less time to measure each seam) and provide invaluable SPC information which can be collected over time.

There are many factors, however, which can disturb the accurate collection of SPC data – and which could easily render SPC data useless:

1. Operator influence. There are several key elements in the seam analysis where the operator has a significant influence over the seam measurement:

a. The saw. Manual saws require the operator to exercise a precise amount of force on the seam. Different operators will exertdifferent forces on the seam – consequently causing the seam to open more or less depending on the operator’s temperament and physical strength. In addition, the quality of the cut is critical to the quality of the automatic seam analyzer. An automatic saw with a clean cut is required in order to eliminate the operator’s strength from factoring in the SPC process.

b. The optical unit. Some optical units provide the operator with the freedom to place the seam in several positions. This could result in an imperfect focus and even angled measurements – for SOME operators. All of these could result in an inaccurate measurement of the seam. A centering device on the optical unit should completely eliminate the operator’s influence in placing the can.

c. Operator bias. In most cases, it is the operator’s duty to stop the line and fix problems, should he observe them in the seam analyzer. It is easier, in many instances to simply move the measurement lines instead of fix the seaming machine. Where such cases are suspected, security features in the software should prevent manual adjustment altogether, or to log such adjustments for review by a supervisor.

2. Repeatability. Measurement variances in the automatic seam analysis software can result in inaccurate or even misleading SPC results. It is critical that the software have an extremely low error rate and optimal repeatability and reproducibility (GR&R) results – both with the same operator and different operators.

3. Collection of data from several different measurement systems. Different measurement systems have a variance of both measurements and calibration. Each system behaves somewhat differently. This could introduce additional problems when collecting data from several different measurement stations and cross-linking the data to produce a single SPC statistic. This SPC graph or statistic could potentially have more errors resulting from variances in the test equipment than in the seaming machine itself.

4. Data collection at different positions on the can. When collecting data about a particular section (i.e., cut) on a particular can – it is critical that the measurements all be done at the same point on the can. Adding external gauges (e.g., seam thickness) introduce statistical errors when calculations (e.g., freespace) are done on the data – which assume that all of the measurement data was performed on the same exact spot on the can.

Variations in the position where the measurement was done can introduce significant errors to calculated fields. Particularly Freespace and Overlap %.

5. Roll and chuck replacement. Roll and Chuck replacement must be logged precisely. Performing statistics on measurements taken both before and after rolls/chucks were replaced – provides useless statistics. Each roll/chuck replacement should be logged and statistics should only be performed during measurement ranges (or dates) where the rolls and chucks have not changed.

It is important to note that the accuracy of the test equipment is the key factor in the SPC analysis. Inaccurate gauges, a non-repeatable seam analyzer software or operator influence can each invalidate the findings of the SPC statistics and graphs.

Measurement equipment should be tested for repeatability and reproducibility, as well as for their ability to minimize outside influences on the seam measurement.

Automatic seam inspection software, such as SEAMetal (Quality By Vision) and Accuseam 2000 (Wilkens Anderson & Co.) have advanced software which minimizes the errors described above for correct and consistant SPC results.

Ofer LaOr,
The author is the technical director of Quality By Vision

All rights reserved (c) 2004 - Cantech International magazine - published with permission.