Cautionary Tale: Are Your Customers' Load Tests Identical with Yours?
by Mark Hayes
Not exactly, is the usual answer. When discrepancies are small and sell within given tolerances, it is often of no importance that your customer gets a slightly higher or lower result. When discrepancies are larger, they may be important, and IST frequently undertakes independent load and torque testing for both springmakers and spring users to resolve such differences. Assuming that both your torque or spring load tester and that of your customer are calibrated correctly, the reason for the discrepancy may lie with the reproducibility and repeatability (R & R) of load testing for springs.
My colleague, Mike Bayliss, senior engineer at IST, has prepared the following discussion to help resolve any differences between your customer’s results and yours.
The R & R Method
The R & R method of quantifying the suitability of a piece of test equipment makes the assumption that the individual test pieces are completely stable compared to the test equipment. This is simply not the case with springs, and several features in springs mean that they do not make reproducible or repeatable test samples.
The R & R results are examining a machine’s repeatability and reproducibility, and as a rule, only machines speciﬁcally designed for spring testing should be used.
If the R & R was undertaken using dead weights or gauge blocks most spring testers would give good results. When using springs to test the R & R, acompletely uncontrolled factor is introduced, which is totally ignored by the test analysis.
There is a wide variety of spring testing machines of differing quality, so it is important to understand what factors affect R & R results, including the spring. This will enable you to decide whether the machine or the spring is the problem when poor R & R results are found.
Spring Factors Inﬂuencing Results
Loss of Free length
If a spring is repeatedly load tested, it may shorten slightly, giving a reducing load characteristic. This feature depends on the stress level in the spring and the design and manufacturing process routes; hence, a progressive reduction in load can sometimes be seen as the operators test the spring
If the spring is left in the unloaded position, it may start to recover some of the lost length. This is a time-, stress- and temperature-dependent feature. Hence the time between different operators carrying out their test runs can inﬂuence the results.
Spring Orientation, Positioning and Location
If the spring is turned upside down the squareness and parallelism of the ends will cause the spring to sit differently, causing a difference in load characteristics.
Positioning of the spring offset relative to the central axis of the tester will induce bending moments in the loadcell and frame and inﬂuence the results.
If the spring is subject to buckling and has to be supported, this will produce considerable non-repeatability/reproducibility due to the spring/ support interface friction conditions.
It is recommended that the test springs be marked to enable all operators to place and orientate the springs repeatedly.
Non-axial Forces Produced by the Spring
When compressed, a spring produces shear forces and moments about the end plane. The magnitude and direction of these forces will inﬂuence the loadcell and measured results. Different types of loadcell behave differently when subject to these extraneous forces, and simply rotating the spring about its axis will also inﬂuence the results.
High rate springs require higher precision measurement systems and mechanics. This is because the test machine will stop at the correct test length, but with a possible error equal to ± 0.5 resolutions, so the recorded load will vary by the ±spring rate x 0.5 x length resolution.
When measuring low load, ensure the measuring system has sufficient resolution. For class 0.5 machines (per ISO 7500-1:2004), the lowest load to be measured on the machine is 400 x load resolution.
Springs that buckle must be guided by a rod of design such that it can be repeatedly ﬁ tted and that the rubbing surfaces remain undamaged. Lubrication may be used to help make the friction more consistent.
Test Machine Factors Inﬂuencing Results
The load measurement system must be stable and the choice of loadcell has a big inﬂuence on test results. As a rule modern electronics are very stable in respect to drift in calibration and do not usually cause problems.
The top grade for spring testers is Class 0.5, which means the reading is accurate to 0.5 percent. This will be over a range between fullscale and a lower limit dictated by the resolution of the display system. If a machine is used below this limit the resolution becomes too coarse and this inﬂ uences the R & R result.
Length Measurement System
Most machines use digital length encoders to measure the spring deﬂection. These are usually very good, and as long as it is ﬁtted correctly, will readily give a 0.01mm resolution and accuracy.
Machines that use rotary encoders on the ballscrews or motors can be subject to non-repeatability if there is mechanical backlash.
When measuring high rate springs the length measuring system must have sufﬁ cient resolution so that the resultant uncertainty in load calculated by Spring Rate x Length Resolution is a small percentage of the tolerance band. High rate springs may require resolutions of 0.002mm.
Loading frame rigidity, mechanics, deﬂ ection and compensation must be extremely good with high rate springs to minimize errors and non-repeatability.
The ease and precision of the length zero datum must also be good to achieve high reproducibility.
The IST’s load test machine meets the above requirements and is shown in Fig. 1.
Understanding the Results
Repeatability is the ability of a machine to measure a spring many times and give the same result.
Reproducibility is the ability of a machine to be reset and used by different operators and to achieve the same result.
If a bad R & R is achieved then it is important to be able to look at the results and ﬁnd where the fault lies. R & R results of less than 10 percent error are satisfactory. Results of 10 – 30 percent may be acceptable based upon the importance of the application, but results greater than 30 percent indicate that the measurement system needs improvement.
Table 1 above shows an R & R of 20.6 percent, which would not usually be accepted. This test used ten springs, three operators and tested three times.
Looking at the "% Tolerance Band" column it can be seen that the repeatability is the dominant problem indicating that the machine or springs are not repeatable.
Looking at the part variation shows that this is
1.5 times the required tolerance band, indicating that maybe there is a spring manufacturing problem or the tolerance is unrealistically small. Alternatively the machine may have an issue with noise, electrical or mechanical, or an ability to ﬁnd a test length repeatedly, because the mean values are good (reproducibility), but the results have a non-repeatable scatter about the mean. The IST’s load test machine does not have these faults and is shown in Fig 1.
The reproducibility is very good indicating that different operators can readily and accurately set up the machine.
Looking at Table 2 (above) it can be seen that the tolerance band is more than three times larger than in the previous example; now the R & R looks good. Part variability is still poor, indicating that the springs are out of tolerance and hence there may be a manufacturing problem.
R & R testing assumes the test pieces are stable — springs are often not stable.
High rate springs will require high-resolution length measuring systems.
High rate springs will require high rigidity loading frames and compensation systems.
Test samples should be marked to enable repeatable positioning by operators.
Poor reproducibility indicates that the machine’s set-up is variable. Often down to a length measurement problem or unstable spring lengths.
Poor repeatability indicates variable test length searching, or noise.
High part variability indicates the spread of spring results is great compared to the tolerance band and a problem in process capability is present.
If the reason for discrepancies between load tests at two sites cannot be traced, the IST can independently test the springs and resolve the problem. When the IST load or torque test springs, the result is regarded as definitive.
Mark Hayes is the senior metallurgist at the Institute of Spring Technology (IST): The International Independent Centre of Excellence for Spring Technology. He manages IST’s spring failure analysis service, and all metallurgical aspects of advice given by the Institute. He also designs and delivers the majority of the spring training courses that the IST offers globally. Readers are encouraged to contact him with comments about this cautionary tale, and with subjects that they would like to be addressed in future tales. Contact Hayes at (011) 44 114 252 7984, fax (011) 44 114 2527997, or e-mail email@example.com.