Measurement error

measurement system analysisAll measurements involve uncertainty. Suppliers of measuring equipment usually specify the measurement uncertainty as a symmetrical interval around the measurement result:

Measurement result ± measurement uncertainty

Example: The length of a special measuring rod is 2000 mm ± 1 mm. With a 95 % confidence interval, which is normal to use, there is a 95% probability that the rod is between 1999 mm and 2001 mm.

Measurement System Analysis (MSA), also referred to as Measurement System Evaluation (MSE), can determine the measurements accuracy, precision and stability. This is done by measuring the same item repeatedly. Based on the repeated measurements, the accuracy, precision and stability can be determined.

Measurements are done many different ways, using different kind of technologies.

Lean Tech has evaluated subjective measurement systems like appearance, taste and smell. Surprisingly, the measurement error for a visual test was greatest for the quality manager, who made the decision when there was doubt about the quality. The quality manager did not perform the test frequently, only occasionally when the quality was questioned. Ironically, this person's evaluation was less consistent than for operators whom performed the test more frequently.

Maybe you use cameras / automatic optical inspection (AOI) to control product quality? The measurement can be performed based on reference points identified by the camera. Measures of distance can be made between reference points to control product quality. Measurement System Evaluations performed by Lean Tech for camera controls, show that they are both accurate and precise if they find the right reference point. Unfortunately, they become unstable if they fail to find the correct reference point and thus render incorrect measurement. 

Here is a video about measurement error for a measurement you probably can relate to: weight. 

In this video, Components of variation (CoV) is used to decide measurement error and within day variation.

For manufacturing companies, it might be relevant to use CoV to decide how much variation is due to the measurement, different measuring instruments, different analyst or operator, and different batches.  This is important to know if you are making decisions based on measurements that are affected by several factors.