Repeatability: The degree to which a measurement process, procedure, or experiment returns the same results when said process, procedure or experiment is conducted over and over again.
Reproducibility: The degree to which all conditions affecting a measurement process, procedure or experiment can be duplicated from one execution to the next.
Accuracy: The degree to which a measurement process, procedure or experiment yields the truth.
“Repeatability does not accuracy make!” However, accurate readings should always be repeatable.
If repeatability is poor, but the process or tools used to obtain the measurement results is known to be good, then the cause of the non-repeatability (or repeatability outside of acceptable tolerances for deviation) must be poor reproducibility. For instance, if results don’t repeat, but you know your measurement tool is good and accurate (such as an Easy-Laser XT 770 Laser alignment system), there is a high likelihood of mechanical looseness (i.e. loose anchor bolts, or a loose bracket, or a loose rod in the bracket, or a loose coupling hub upon which the bracket is mounted), all leading to the conditions of the measurement not remaining consistent from one measurement to the next; alternatively, the conditions of the experiment are being affected by other external factors causing them to be continuously changing, such as machine temperatures cooling from a machine set that has recently been shut down, or the sun coming out from behind an obstruction and is now shining upon one machine but not the other, etc. etc.
Nevertheless, if reproducibility is good, and repeatability is good too, that still might not yield accuracy, because a repeatable error is intruding into the conditions of the experiment that compromises obtention of the truth. For instance, the laser head might be imperceptibly bumping or brushing against an obstruction to free rotation (such as a coupling stud in the base), or a protrusion of some sort (such as a bolt head protruding from the bearing housing), causing the laser head to be moved or deflected from its true position at rest just at that position, thereby compromising the truth of certain measurement points in the rotation. Thus, even though repeatable, the calculated results might still not “accurately” reflect the truth of the alignment, and these types of issues are particularly difficult to diagnose and pinpoint.
Filed under:
Alignment by Alan Luedeking CRL CMRP