《纺织复合材料》课程参考文献（Composite Materials Handbook，Volume 1）CHAPTER 2 GUIDELINES FOR PROPERTY TESTING OF COMPOSITES

MIL-HDBK-17-1F Volume 1,Chapter 2 Guidelines for Property Testing of Composites 2.2.5 Population sampling and sizing Unlike MIL-HDBK-5 for metals,MIL-HDBK-17 for composites does not require simultaneous determi- nation of B-basis values and A-basis values from the same population.This is not because of any fun- damental difference in material behavior,but due to a relative lack of need for A-basis properties,to date for composites.As a result,the composite material B-basis sample population(30+)is much smaller than the MIL-HDBK-5 A/B-basis sample population(100-300)for metals.Unfortunately,since there are usually more composite properties and directions under test,and since testing matrices for composites are often fully populated not only at room temperature but also at the environmental extremes,the total number of specimens in a B-basis composite testing program often exceeds the total number of coupons in an A/B- basis metals testing program.'However,included in and allowed by MIL-HDBK-17 are advanced statisti- cal regression techniques that offer the possibility,in specific instances and when combined with different sampling distributions,of being able to reliably determine A-basis values from a total number of composite material specimens similar in quantity to those previously needed for B-basis values (see Section 2.3.6.1) The sampling approach required for MIL-HDBK-17 B-basis nonregression data,and described in de- tail in Section 2.5.3,includes at least five batches of production material,using a minimum of 30 speci- mens distributed among the batches,and fully tests each property at each environment under considera- tion.The first five prepreg batches are each made using distinct fiber and matrix constituent lots(not re- quired of batch numbers greater than five).For each condition and property,batch replicates are sam- pled from at least two different test panels covering at least two separate processing cycles.Test panels are non-destructively evaluated using ultrasonic inspection or another suitable non-destructive inspection technique.Test coupons are not extracted from panel areas having indications of questionable quality.A test plan (or report)documents laminate design,specimen sampling details.fabrication procedures (in- cluding material traceability information),inspection methods,specimen extraction methods,labeling schemes,and test methods For general data development,sampling techniques and sample sizes may be application or qualifi- cation/certification agency dependent.A desirable goal of any sampling scheme making use of MIL-HDBK-17 statistical methods is to have multiple batches composed of uniformly-sized subpopula- tions.The five-batch minimum requirement only applies to material properties that are to be incorporated in MIL-HDBK-17.An alternate number of replicates and batches may be employed upon approval of the procuring or certifying agency.However,mechanical strength data should be evaluated by the statistical methods recommended by this handbook to ensure statistically acceptable basis values 2.2.5.1 Sample size selection Regardless of the sampling scheme,for small sample populations,the result of any basis value cal- culation is strongly dependent on the sample size.Smaller sample populations are obviously less costly to test,but there is a price of a different kind to pay since,as the population size decreases,so does the calculated basis value.Figure 2.2.5.1 shows,for a hypothetical example,the effect of sample size on the calculated B-basis value-for samples of various sizes drawn from a given infinite population normally dis- tributed.In the limit,for very large sample sizes,the B-basis(ten percentile)value for this example would be 87.2.The dotted line in the figure is the mean of all possible B-basis values for each sample size;this line can also be interpreted as the estimated B-basis value as a function of population size for a fixed sample coefficient of variation(CV)of 10%.The dashed lines represent the one-sigma limits for any given sample size(a two-sigma limit would approximately bound the 95%confidence interval). MIL-HDBK-5,the metals handbook,focuses on A-basis values and requires a minimum of 100 tensile specimens,but uses small populations of compressive shear,bearing,and non-ambient tests ratioed to the room temperature tensile properties to estimate compressive,shear,bearing and non-ambient basis values.MIL-HDBK-17 requires at least 30 specimens for each direction,for each property,and for each environment to determine B-basis values.The MIL-HDBK-17 requirement increases to 90 coupons for A-basis values.However,when using MIL-HDBK-17 advanced statistical regression techniques,the specimen populations can sometimes be spread over all of the environments under test,thus reducing the total number of test specimens needed. Any statistical calculation based on a subpopulation is only an estimate of the real value for the entire population,although the larger and more representative the sample,the better the estimate. 2-11

MIL-HDBK-17-1F Volume 1, Chapter 2 Guidelines for Property Testing of Composites 2-11 2.2.5 Population sampling and sizing Unlike MIL-HDBK-5 for metals, MIL-HDBK-17 for composites does not require simultaneous determi￾nation of B-basis values and A-basis values from the same population. This is not because of any fun￾damental difference in material behavior, but due to a relative lack of need for A-basis properties, to date, for composites. As a result, the composite material B-basis sample population (30+) is much smaller than the MIL-HDBK-5 A/B-basis sample population (100-300) for metals. Unfortunately, since there are usually more composite properties and directions under test, and since testing matrices for composites are often fully populated not only at room temperature but also at the environmental extremes, the total number of specimens in a B-basis composite testing program often exceeds the total number of coupons in an A/B￾basis metals testing program.1 However, included in and allowed by MIL-HDBK-17 are advanced statisti￾cal regression techniques that offer the possibility, in specific instances and when combined with different sampling distributions, of being able to reliably determine A-basis values from a total number of composite material specimens similar in quantity to those previously needed for B-basis values (see Section 2.3.6.1). The sampling approach required for MIL-HDBK-17 B-basis nonregression data, and described in de￾tail in Section 2.5.3, includes at least five batches of production material, using a minimum of 30 speci￾mens distributed among the batches, and fully tests each property at each environment under considera￾tion. The first five prepreg batches are each made using distinct fiber and matrix constituent lots (not re￾quired of batch numbers greater than five). For each condition and property, batch replicates are sam￾pled from at least two different test panels covering at least two separate processing cycles. Test panels are non-destructively evaluated using ultrasonic inspection or another suitable non-destructive inspection technique. Test coupons are not extracted from panel areas having indications of questionable quality. A test plan (or report) documents laminate design, specimen sampling details, fabrication procedures (in￾cluding material traceability information), inspection methods, specimen extraction methods, labeling schemes, and test methods. For general data development, sampling techniques and sample sizes may be application or qualifi￾cation/certification agency dependent. A desirable goal of any sampling scheme making use of MIL-HDBK-17 statistical methods is to have multiple batches composed of uniformly-sized subpopula￾tions. The five-batch minimum requirement only applies to material properties that are to be incorporated in MIL-HDBK-17. An alternate number of replicates and batches may be employed upon approval of the procuring or certifying agency. However, mechanical strength data should be evaluated by the statistical methods recommended by this handbook to ensure statistically acceptable basis values. 2.2.5.1 Sample size selection Regardless of the sampling scheme, for small sample populations, the result of any basis value cal￾culation is strongly dependent on the sample size. Smaller sample populations are obviously less costly to test, but there is a price of a different kind to pay since, as the population size decreases, so does the calculated basis value. Figure 2.2.5.1 shows, for a hypothetical example, the effect of sample size on the calculated B-basis value2 for samples of various sizes drawn from a given infinite population normally dis￾tributed. In the limit, for very large sample sizes, the B-basis (ten percentile) value for this example would be 87.2. The dotted line in the figure is the mean of all possible B-basis values for each sample size; this line can also be interpreted as the estimated B-basis value as a function of population size for a fixed sample coefficient of variation (CV) of 10%. The dashed lines represent the one-sigma limits for any given sample size (a two-sigma limit would approximately bound the 95% confidence interval). 1 MIL-HDBK-5, the metals handbook, focuses on A-basis values and requires a minimum of 100 tensile specimens, but uses small populations of compressive shear, bearing, and non-ambient tests ratioed to the room temperature tensile properties to estimate compressive, shear, bearing and non-ambient basis values. MIL-HDBK-17 requires at least 30 specimens for each direction, for each property, and for each environment to determine B-basis values. The MIL-HDBK-17 requirement increases to 90 coupons for A-basis values. However, when using MIL-HDBK-17 advanced statistical regression techniques, the specimen populations can sometimes be spread over all of the environments under test, thus reducing the total number of test specimens needed. 2 Any statistical calculation based on a subpopulation is only an estimate of the real value for the entire population, although the larger and more representative the sample, the better the estimate

MIL-HDBK-17-1F Volume 1,Chapter 2 Guidelines for Property Testing of Composites Not only does the estimated B-basis value increase with larger sample sizes,but,as the one-sigma limits illustrate,the expected variation in estimated B-basis value significantly decreases.The lower one- sigma limit is farther from the mean B-basis value than the upper one-sigma limit,illustrating a skew in calculated B-basis value that is particularly strong for small sample sizes.As a result of this skew,for small populations the calculated B-basis value is substantially more likely to be overly conservative than under-conservative,increasing the significant penalty in B-basis value paid by use of small populations. While similar examples for non-normal distributions would have different quantitative results the trends with sample size can be expected to be similar.Additional discussions on effects of sample size are lo- cated in Section 8.2.5. 100 80 60 ● 0 Maan=100,s.D.=10 20 5 0 15 20 25 Number Of Speclmens FIGURE 2.2.5.1 Normal B-basis values with one-sigma limits. 2.2.5.2 Batch quantity effects on ANOVA The MIL-HDBK-17 statistical methodology(Figure 8.3.1)includes a statistical test to assess the de- gree of batch-to-batch variation.If the resulting statistic indicates excessive batch-to-batch variation,the data are not conventionally pooled but are instead evaluated using an Analysis of Variance(ANOVA)ap- proach.However,the statistical methods are only as good as the quality and quantity of data that they evaluate. Small numbers of batches can cause the ANOVA approach to produce extremely conservative basis values,since it essentially treats the average of each batch as a single data point for input to a conven- tional normal distribution technique for basis value determination (Section 2.2.5.1 describes the effect of small samples on basis values).As the MIL-HDBK-17 statistical methods assume that testing variation is negligible,variation caused by testing (see related discussion in Section 2.2.4),either within or between batch,is treated as real material/process variation and can result in unrealistically low basis values. Also,the between-batch variation test becomes progressively weaker as the number of batches de- creases,or as the variation between batches decreases,or both.For example,when only a small num- ber of batches are sampled,a batch variation test result indicating no significant batch variation may be deceptive.Additional batch samples may indicate that batch variation really exists,but was masked by the small original number of batches. 2-12

MIL-HDBK-17-1F Volume 1, Chapter 2 Guidelines for Property Testing of Composites 2-12 Not only does the estimated B-basis value increase with larger sample sizes, but, as the one-sigma limits illustrate, the expected variation in estimated B-basis value significantly decreases. The lower one￾sigma limit is farther from the mean B-basis value than the upper one-sigma limit, illustrating a skew in calculated B-basis value that is particularly strong for small sample sizes. As a result of this skew, for small populations the calculated B-basis value is substantially more likely to be overly conservative than under-conservative, increasing the significant penalty in B-basis value paid by use of small populations. While similar examples for non-normal distributions would have different quantitative results the trends with sample size can be expected to be similar. Additional discussions on effects of sample size are lo￾cated in Section 8.2.5. FIGURE 2.2.5.1 Normal B-basis values with one-sigma limits. 2.2.5.2 Batch quantity effects on ANOVA The MIL-HDBK-17 statistical methodology (Figure 8.3.1) includes a statistical test to assess the de￾gree of batch-to-batch variation. If the resulting statistic indicates excessive batch-to-batch variation, the data are not conventionally pooled but are instead evaluated using an Analysis of Variance (ANOVA) ap￾proach. However, the statistical methods are only as good as the quality and quantity of data that they evaluate. Small numbers of batches can cause the ANOVA approach to produce extremely conservative basis values, since it essentially treats the average of each batch as a single data point for input to a conven￾tional normal distribution technique for basis value determination (Section 2.2.5.1 describes the effect of small samples on basis values). As the MIL-HDBK-17 statistical methods assume that testing variation is negligible, variation caused by testing (see related discussion in Section 2.2.4), either within or between batch, is treated as real material/process variation and can result in unrealistically low basis values. Also, the between-batch variation test becomes progressively weaker as the number of batches de￾creases, or as the variation between batches decreases, or both. For example, when only a small num￾ber of batches are sampled, a batch variation test result indicating no significant batch variation may be deceptive. Additional batch samples may indicate that batch variation really exists, but was masked by the small original number of batches

MIL-HDBK-17-1F Volume 1,Chapter 2 Guidelines for Property Testing of Composites The above should be understood when batch variation exists and ANOVA basis values are calculated on fewer than five batches. 2.2.6 Material and processing variation,specimen preparation and NDE In the sections of Volume 1 that follow in the handbook,the reader will find an extensive compilation of test methods for a variety of fibers,resins and composite material forms and structural elements.In most cases these materials or structural elements are the products of complex multi-step materials proc- esses.Figures 2.2.6(a)and 2.2.6(b)illustrate the nature of the processing pipeline from raw materials to composite end item.(Each rectangle in Figure 2.2.6(b)represents a process during which additional variability may be introduced into the material.)These processes may require elevated temperature, stress or pressure.They often involve evolution of volatiles,resin flow and consolidation,and readjust- ment of reinforcing fibers.If the measured properties of composite materials are to be interpreted cor- rectly and used appropriately,the variability of the properties of the materials must be understood.This variability arises during routine processing and may be increased by any of the legion of anomalies which may occur during processing. 2.2.6.1 Materials and material processing The constituents of the composite materials covered in this handbook are organic matrices (either thermosetting or thermoplastic)and organic or inorganic reinforcing fibers.Variation in the mechanical properties of the reinforcing fibers can arise from many sources,such as flaws in fiber microstructure,or variations in degree of orientation of the polymer chains in an organic fiber. Thermoplastic matrices can exhibit variations in molecular weight and molecular weight distribution as a result of processing.The melt viscosity and subsequent processability of the thermoplastic matrix may be strongly affected by such variability.Thermosetting resins are often applied to fibers in a prepregging operation and some forms partially cured to what is referred to as a B-stage.Other methods for stabiliz- ing thermoset resin systems may also be employed prior to the prepregging operation.Stability of these materials is important because there are many potential sources of variability during packaging,shipping and storage of improperly,or even properly,stabilized intermediate forms such as prepreg tape,fabrics and roving. The placement of reinforcing fibers may be accomplished through many manual or automated proc- esses.Lack of precision in fiber placement or subsequent shifting of reinforcing fibers during matrix flow and consolidation can introduce variability.Depending on the process (e.g.,pultrusion compared to RTM),cure and/or consolidation can occur simultaneously with fiber placement,or after fiber placement has occurred.This step in the process is especially vulnerable to the introduction of variability. As an example,consider the cure of a composite part from B-staged prepreg tape in an autoclave,a press or an integrally heated tool.When the resin is heated and has begun to flow,the material consists of a gas phase (volatiles or trapped air),a liquid phase (resin),and a solid (reinforcement)phase.To avoid variability in material properties due to excessive void volume,void producing gas phase material must be either removed or absorbed by the liquid phase.In order to avoid variability due to variations in fiber volume fraction,the resin must be uniformly distributed throughout the part.The fiber must maintain its selected orientation in order to avoid variability or loss of properties due to fiber misalignment. Pertinent process parameters and material effects should always be documented to aid in process control and troubleshooting.If potential processing and manufacturing pitfalls are not identified and avoided in this way,resources may be wasted in testing materials which are not representative of those which will occur in an actual part or application.In addition,heavy weight penalties may be paid to allow for avoidable material variability.A better understanding of these processing parameters and their poten- tial effect on material properties will also allow a composites manufacturer to avoid the considerable ex- penses involved in the production of materials,parts or end items with unacceptable properties. 2-13

MIL-HDBK-17-1F Volume 1, Chapter 2 Guidelines for Property Testing of Composites 2-13 The above should be understood when batch variation exists and ANOVA basis values are calculated on fewer than five batches. 2.2.6 Material and processing variation, specimen preparation and NDE In the sections of Volume 1 that follow in the handbook, the reader will find an extensive compilation of test methods for a variety of fibers, resins and composite material forms and structural elements. In most cases these materials or structural elements are the products of complex multi-step materials proc￾esses. Figures 2.2.6(a) and 2.2.6(b) illustrate the nature of the processing pipeline from raw materials to composite end item. (Each rectangle in Figure 2.2.6(b) represents a process during which additional variability may be introduced into the material.) These processes may require elevated temperature, stress or pressure. They often involve evolution of volatiles, resin flow and consolidation, and readjust￾ment of reinforcing fibers. If the measured properties of composite materials are to be interpreted cor￾rectly and used appropriately, the variability of the properties of the materials must be understood. This variability arises during routine processing and may be increased by any of the legion of anomalies which may occur during processing. 2.2.6.1 Materials and material processing The constituents of the composite materials covered in this handbook are organic matrices (either thermosetting or thermoplastic) and organic or inorganic reinforcing fibers. Variation in the mechanical properties of the reinforcing fibers can arise from many sources, such as flaws in fiber microstructure, or variations in degree of orientation of the polymer chains in an organic fiber. Thermoplastic matrices can exhibit variations in molecular weight and molecular weight distribution as a result of processing. The melt viscosity and subsequent processability of the thermoplastic matrix may be strongly affected by such variability. Thermosetting resins are often applied to fibers in a prepregging operation and some forms partially cured to what is referred to as a B-stage. Other methods for stabiliz￾ing thermoset resin systems may also be employed prior to the prepregging operation. Stability of these materials is important because there are many potential sources of variability during packaging, shipping and storage of improperly, or even properly, stabilized intermediate forms such as prepreg tape, fabrics and roving. The placement of reinforcing fibers may be accomplished through many manual or automated proc￾esses. Lack of precision in fiber placement or subsequent shifting of reinforcing fibers during matrix flow and consolidation can introduce variability. Depending on the process (e.g., pultrusion compared to RTM), cure and/or consolidation can occur simultaneously with fiber placement, or after fiber placement has occurred. This step in the process is especially vulnerable to the introduction of variability. As an example, consider the cure of a composite part from B-staged prepreg tape in an autoclave, a press or an integrally heated tool. When the resin is heated and has begun to flow, the material consists of a gas phase (volatiles or trapped air), a liquid phase (resin), and a solid (reinforcement) phase. To avoid variability in material properties due to excessive void volume, void producing gas phase material must be either removed or absorbed by the liquid phase. In order to avoid variability due to variations in fiber volume fraction, the resin must be uniformly distributed throughout the part. The fiber must maintain its selected orientation in order to avoid variability or loss of properties due to fiber misalignment. Pertinent process parameters and material effects should always be documented to aid in process control and troubleshooting. If potential processing and manufacturing pitfalls are not identified and avoided in this way, resources may be wasted in testing materials which are not representative of those which will occur in an actual part or application. In addition, heavy weight penalties may be paid to allow for avoidable material variability. A better understanding of these processing parameters and their poten￾tial effect on material properties will also allow a composites manufacturer to avoid the considerable ex￾penses involved in the production of materials, parts or end items with unacceptable properties

MIL-HDBK-17-1F Volume 1,Chapter 2 Guidelines for Property Testing of Composites ALL PRODUCTS,REGARDLESS TO STAGE OF WORK,ARE CONSID- ERED AS RAW MATERIALS.THESE MAY BE CHEMICALS FOR RAW RESINS,OR SAND TO PROCESS INTO GLASS PRODUCTS,OR PRECURSORS FOR FILAMENTS,OR WOVEN GOODS,OR FLAT MATERIAL PROCESSED SHEETS AND/OR MANY OTHER ARTICLES WHICH HAVE YET TO BE PROCESSED TO AN END ITEM. EACH RAW PRODUCT THEN IS PROCESSED,OR MIXED WITH OTHER RAW PRODUCTS,OR ALTERED TO BECOME STILL ANOTH- MANUFACTURE ER RAW ARTICLE TO EXPERIENCE YET ADDITIONAL PROCESS STEPS THROUGH THE PIPELINE.EACH AS RECEIVED RAW &/OR ARTICLE MUST BE PROCESSED IN SUCH A MANNER DURING ITS PROCESSING STEP(S)THAT VARIABILITY IS MINIMIZED AT THE PROCESSING NEXT PIPELINE FUNCTION.PROCESSING FUNCTIONS MAY BE COMPLEX;SUCH AS MATRIX IMPREGNATION,OR THEY MAY BE RELATIVELY SIMPLE;SUCH AS SHIPPING.REGARDLESS,EACH STEP MUST BE EFFECTIVE IN THAT IT DOES NOT INTRODUCE UNCONTROLLED CHANGES THAT ALTER THE PRODUCT FOR SUBSEQUENT USE OR END ITEM PERFORMANCE. FINISHED ARTICLES LEAVING ONE PROCESSING FUNCTION IS USUALLY STILL CONSIDERED A RAW PRODUCT WHEN DELIVERED FINISHED TO THE NEXT.THE ONUS FOR CONSISTENCY OF THIS PRODUCT MUST BE RECOGNIZED AND ATTENDED TO DURING THE JUST ARTICLE COMPLETED PROCESSING STEP(S).THE MATERIALS PIPELINE IS NOT COMPLETE UNTIL THE END ARTICLE IS FULLY FUNCTIONAL AS IS. FIGURE 2.2.6(a) Composite materials and processing,basic pipeline common to all materials and processes. 2-14

MIL-HDBK-17-1F Volume 1, Chapter 2 Guidelines for Property Testing of Composites 2-14 FIGURE 2.2.6(a) Composite materials and processing, basic pipeline common to all materials and processes

MIL-HDBK-17-1F Volume 1,Chapter 2 Guidelines for Property Testing of Composites RAW RESIN FIBER RAW MATERIAL MANUFACTURE MANUFACTURE MATERIAL PRODUCTS PRODUCT STRAND,TOW ROVING RAW RAW RAW MATERIAL MATERIAL MATERIAL RTM WET PREPREGER WEAVER MANUFACTURERS MANUFACTURE MANUFACTURE ARTICLE FINISHED FABRIC GOODS PRODUCTS PRODUCTS PRODUCTS FIGURE 2.2.6(b)Raw materials pipeline (example). 2-15

MIL-HDBK-17-1F Volume 1, Chapter 2 Guidelines for Property Testing of Composites 2-15 FIGURE 2.2.6(b) Raw materials pipeline (example)