Literature

Joint Industry Foam Standards and Guidelines

APPENDIX A4.0

Published: 7/94

Interlaboratory Study of IFD Measurement Precision and Bias

A4.1 Scope
The purpose for conducting an interlaboratory study of IFD measurement is engendered by the necessity to establish acceptance range guidelines for IFD of flexible polyurethane foam for use by both the foam manufacturer and the furniture manufacturer. The test method under scrutiny is ASTM D 3574-86 Test B1 "Standard Methods of Testing Flexible Cellular Materials--Slab, Bonded, and Molded Urethane Foams".

A4.2 Procedure and Sampling
The basic outline for conducting an interlaboratory study to determine the reliability of a test method is outlined in ASTM E 691-87 "Standard Practice for Conducting an Interlaboratory Study to Determine the Precision of a Test Method".

In preparing the interlaboratory study, nineteen laboratories were chosen to conduct measurements of weight, height, 25% IFD, 65% IFD, and 25% Return IFD on flexible polyurethane foam samples. These five measurements are the fundamentals on which other parameters are calculated, i.e., density, modulus of compression, hysteresis, etc. In addition, a calibration spring was circulated among all participating laboratories to collect height and 25% compression values for comparative reference.

The flexible polyurethane foam samples were supplied by two manufacturers, each to provide a nominal 1.8 lb/ft3 density, 30 lb IFD conventional grade foam. The foam pieces were cut horizontally from the center of the foam bun and serially numbered for location identification. All foam samples were cut a nominal 24 inches square by 4 inches in thickness. Foam samples were crushed 90% of their original thickness twenty-five cycles by the use of rollers. The samples were kept segregated by supplier, thus yielding two sample sets.

A4.3 Sample screening
All foam samples were sent to a centralized screening laboratory to test for uniformity of 25% IFD measurements within the sample sets, allowing the opportunity to remove any outlying specimens from the test.

Great care was taken in establishing the interlaboratory study procedures to provide each laboratory with foam samples that were in equal condition. Previous studies circulated a great number of foam specimens from laboratory to laboratory, compounding the foam fatigue element each step along the way. In this study, each laboratory received its own set of specimens with each specimen having been subjected to exactly the same crushing and screening steps, i.e., the fatigue element was exactly the same for each specimen and therefore was eliminated as a variable.

A4.4 Data
The reported data required from each laboratory included specimen weight, height, 25% IFD, 65% IFD, 25% return IFD for three specimens from each of two sample sets. The test method used was ASTM D 3574-86 Test B1 with the preflex rate set at 10 inches per minute. Test machine type and capacity, and laboratory temperature and relative humidity readings were also reported.

A4.5 Statistical Analysis
The data returned from the nineteen laboratories was analyzed according to the consistency statistic procedures as outlined in ASTM E 691 to determine repeatability and reproducibility limits at the 95% acceptance level. Precision and bias statements were prepared by following the procedures outlined in ASTM E 177-90a "Standard Practice for Use of the Terms Precision and Bias in ASTM Test Methods." The data was also analyzed using the Kolmogorov-Smirnov test to compare distributions between sample sets and distributions between the screening lab data (used as a population distribution since there are no known true values) and the interlaboratory data.

A4.6 Conclusions
The consistency analysis results show that there was relatively consistent performance between laboratories. There was, however, some imbalance within some of the laboratories. The consistency analysis points toward equipment or measurement scale deficiencies as a possible explanation for intralaboratory inconsistences. The following tables summarize the precision statements in terms of 95% repeatability and reproducibility limits of IFD measurement for the foam sample sets A and B from the interlaboratory study. These limits demonstrate just how closely one could expect to reproduce the test values observed in this study within a single laboratory (repeatability) or among several laboratories (reproducibility).

SET A 25% IFD 65% IFD 25% Return IFD
Average test value (lbs) 29.9 65.8 24.5
95% Repeatability limits (lbs) 1.26 5.66 1.60
95% Reproducibility limits (lbs) 2.21 7.14 2.16
SET B 25% IFD 65% IFD 25% Return IFD
Average test value (lbs) 32.7 66.9 24.4
95% Repeatability limits (lbs) 2.32 4.68 2.27
95% Repeatability limits (lbs) 3.70 7.39 6.36

Using the screening laboratory values as reference values, the bias statements are:

  • The ILS 25% IFD means for sets A and B were approximately 3% lower.
  • The ILS 65% IFD mean for set A was approximately 0.2% higher.
  • The ILS 65% IFD mean for set B was approximately 1% lower.
  • The ILS 25% Return IFD means for sets A and B were approximately 0.4% higher.
  • The comparison of distributions (Kolmogorov-Smirnov test) of the interlaboratory study shows a significant difference between data sets A and B and the screening laboratory (population) data at an alpha level of 0.05, particularly on height and 25% IFD measurement (see table below).

    KOLMOGOROV-SMIRNOV TEST COMPARISON OF DISTRIBUTIONS

    DISTRIBUTIONS COMPARED HEIGHT WEIGHT 25% IFD* 65% IFD* 25% R-IFD*
    ILS vs SCREEN SETS A & B 0.0007 0.6634 0.0001 0.2115 0.0021
    ILS SET A vs ILS SET B 0.0001 0.0001 0.0001 0.0001 0.0033
    ILS vs SCREEN SET A 0.0036 0.3442 0.0001 0.4761 0.0386
    ILS vs SCREEN SET B 0.0126 0.4761 0.0001 0.0126 0.1031

     

    *p- values

    IF p-VALUE < 0.05, THEN DIFFERENCE IS SIGNIFICANT @ ALPHA OF 0.05

    One explanation for this significant difference is the degree of error associated with measuring foam height using a one pound preload. The survey of equipment used by the participating laboratories reveals that the load cells used have capacities ranging from 150 to 1000 lbs. The sensitivity of the various load cells in the extreme low end of their range is very suspect to error. The 30 pound force generated by compressing the foam samples to 25% of original height in the 25% IFD measurement is also very near the lower extremity of most of the load cell ranges.

    The conclusions drawn and the precision and bias statements presented in this interlaboratory study are only valid for 1.8 lb/ft3 density, nominal 30 lb IFD conventional flexible polyurethane foam. Extrapolation to other foam densities, IFD's, or types is highly speculative. One could expect to see similarities in inter- and intra- laboratory performance when testing different foam populations, however, the precision and bias statements will take on different values than those presented in this interlaboratory study.

    One must remember that the purpose of this study was to determine the reliability of the test method. Many of the variables that are encountered in routine IFD measurement have been controlled or eliminated within this interlaboratory study. Therefore, the data collected in this study reflects only the "good end" of the data spectrum. In actual practice, IFD measurements are victim to variables such as height measurement error, preflex speed variances, recovery time differences, flex rate differences, and numerous on-site "shortcuts" in following test methodology. Also, the physical properties of the flexible polyurethane foam that is routinely purchased in the industry are much more random than the controlled sampling obtained for this study. However, the ñ3 lb IFD acceptance range currently in practice is satisfactory and practical in this random foam supply.