Occasionally the Institute of Spring Technology (IST) has been asked why different companies get different spring load test results when testing the same batch of compression springs. In one particular instance, results recorded by IST were different from those of the springmaker who had used the same model of load test equipment while the ultimate customer had recorded yet another set of values. Although all three sets of results were within the load at length tolerances for the spring design, the variation had, unsurprisingly, raised concerns on the part of the end customer.
Many springmakers will appreciate that there is a natural non-repeatability in spring load calculations such that, every time they are tested, the result can be slightly different, but many companies struggle to understand why this happens. There is also the matter of tester to tester variation to take into account. All of this raises the question as to what would constitute a natural variation, and what would become significant difference. In other words, which set of results truly reflect the performance of the spring? The variation in spring load test results is made up of four main factors:
Length measurement effects
Load cell effects
Length Measurement Effects
This is the error produced by the length measuring system and the load frame deflection. All measuring systems have an error, and different test machine manufacturers may use length measuring systems with different degrees of accuracy. For example, the length encoders used by IST for its range of test machines have a resolution of 0.0002 in (0.005 mm) and are accurate to +0.00012 in (+0.003 mm), but that accuracy degrades when they are bolted into a testing frame due to imprecise deflection compensation and nonlinear deflections such as those from bearings. Hence a real life accuracy figure on IST machines is + 0.0004 in (+ 0.01 mm). ON a high rate spring, 0.0002 in (0.005 mm) resolution can produce significant load variation and IST regularly supply machines with 0.0001 in (0.001 mm) resolution encoders. The machine datum or zero setting can also introduce errors if the platens are not parallel or fi they are worn or damaged. A high rate spring placed offset from the machine center can cause the frame/load cell to deflect differently and hence introduce a length error.
Load Cell Effects
The type of load cell used in spring testers is important because many types of cell are badly affected by non-axial forces and moments. Irrespective of how careful the load cell selection has been made, all cells will exhibit some effect to a greater or lesser extent. It is thus good practice to always locate springs centrally on a test platen.
It is very difficult to quantify load cell effects, but they can be very significant and cause several percent differences in load measurements.
For spring testing, the machine load cell should be accurate to + 0.5% of reading or better. This means that, if the machine used by one company was reading high and that of another company was reading low, a difference of one percent of load could be seen. It is also significant that many calibration companies do not adjust the machine or even tell people if the machine is wrong when they are checking the equipment. Similarly many calibrators/ operators do not check the length systems, and IST has seen errors of greater than 0.0039 in (0.1 mm).
End squareness has considerable effect on the spring load as turning it upside down will usually give a different test result, as illustrated in figures 1 and 2. This means that unless the spring design is such that it will be tested one way up (e.g. for conical springs, beehive springs, springs that have a number of closed coils only at one end, etc.), it is advisable to test both ways up. In addition, springs should be placed in the same orientation on the test platen (i.e. with the end at the same position) in order to minimize possible variations.
Springs that are laterally stiff produce high side forces that will impact on the load cell and frame (as mentioned above). Repeatability and reproducibility tests (R&R) assume that the test pieces are absolutely stable and all variation is due to the machine/operator, but spring effects can make R&R test appear very poor even though the machine is good.
Springs themselves are not always stable items (design dependent) and are subject to recovery hen left unloaded or relaxation when stored compressed. Hence loads may appear to increase or decrease simply because of the delay between testing at each company.
There is no way to quantify what would be a natural non-repeatability in spring load as it will always depend on the design in question. The variation would ONLY constitute a significant difference when that variation falls outside of the load tolerance for the spring.
Mike Bayliss is the senior design engineer at the Institute of Spring Technology (IST) in Sheffield, England. He is principally involved with design and development of IST’s extensive range of spring testing machines. Readers are encouraged to contact IST at email@example.com with comments about this cautionary tale, and with subjects that they would like to be addressed in future tales.
Contributed and first published by Springs Magazine(Fall 2014)