《纺织复合材料》课程参考文献（Composite Materials Handbook，Volume 3）Chapter 3 QUALITY CONTROL OF PRODUCTION MATERIALS AND PROCESSES

MIL-HDBK-17-3F Volume 3,Chapter 3 Quality Control of Production Materials and Processes by Engineering.Parts may be dispositioned 1)acceptable as is,2)subjected to further rework or repair to make the part acceptable or 3).scrapped. X-ray inspection is frequently used in NDI testing to evaluate bonding of inserts in laminate panels and honeycomb core to facesheet bonds in sandwich panels.The extent of testing required is designated on the engineering drawing by type or class of inspection.The type or class is usually defined in a sepa- rate document that is referenced in the manufacturer's process specification.As with ultrasonic inspec- tion,standards with built-in defects are usually required to evaluate the radiographic film properly. 3.3.3 Destructive tests 3.3.3.1 Background Destructive tests are often used to ensure the structural integrity of a component whenever assurance cannot be gained by nondestructive techniques alone.These tests include periodic dissection of the part to examine the interior of complex structures and mechanical testing of specimens cut from excess parts of the component(Figure 3.3.3.1). Quality Assurance of Composite Parts Adequate NDI Details not adequately inspected by NDI Low Experience Higher Experience Net Trim Trim Areas No Destructive Similar Parts Testing Required First Article Full Part Trim Sections Dissection Increased Experience Critical Parts Micrographs Micrographs Mechanical Ply Verification Ply Verification Tests (Process Control) FIGURE 3.3.3.1 Use of destructive tests. 3.3.3.2 Usage Destructive tests are often used to ensure the structural integrity of a component whenever assurance cannot be gained by nondestructive techniques alone.These tests include periodic dissection of the part to examine the interior of complex structures and mechanical testing of coupons cut from excess parts of the component. 3-6

MIL-HDBK-17-3F Volume 3, Chapter 3 Quality Control of Production Materials and Processes 3-6 by Engineering. Parts may be dispositioned 1) acceptable as is, 2) subjected to further rework or repair to make the part acceptable or 3), scrapped. X-ray inspection is frequently used in NDI testing to evaluate bonding of inserts in laminate panels and honeycomb core to facesheet bonds in sandwich panels. The extent of testing required is designated on the engineering drawing by type or class of inspection. The type or class is usually defined in a sepa￾rate document that is referenced in the manufacturer's process specification. As with ultrasonic inspec￾tion, standards with built-in defects are usually required to evaluate the radiographic film properly. 3.3.3 Destructive tests 3.3.3.1 Background Destructive tests are often used to ensure the structural integrity of a component whenever assurance cannot be gained by nondestructive techniques alone. These tests include periodic dissection of the part to examine the interior of complex structures and mechanical testing of specimens cut from excess parts of the component (Figure 3.3.3.1). FIGURE 3.3.3.1 Use of destructive tests. 3.3.3.2 Usage Destructive tests are often used to ensure the structural integrity of a component whenever assurance cannot be gained by nondestructive techniques alone. These tests include periodic dissection of the part to examine the interior of complex structures and mechanical testing of coupons cut from excess parts of the component

MIL-HDBK-17-3F Volume 3.Chapter 3 Quality Control of Production Materials and Processes 3.3.3.3 Destructive test approaches There are two primary categories of destructive tests:dissection of the full part or examination of trim sections of the part.Full dissection,generally done for the first part from a new tool,gives a complete ex- amination of the part,but is expensive to perform.Examination of excess trim sections is the preferable approach whenever possible.The part is not destroyed,structural details can still be examined and me- chanical test specimens can be obtained. Full Part Dissection:Full part dissection is the approach often envisioned when the term "destructive testing"is mentioned.Since it prevents future use of the part,full part dissection should be reserved for parts that meet the following criteria: Areas cannot be adequately inspected by NDI Part is complex and there is a low experience level for working with the structural configuration or fabrication process Part is net trim;detail areas of interest cannot be examined using excess trim areas or part ex- tensions. Trim Sections:Examination and testing of trim sections offers a balance of quality assurance and cost. Trim sections can be part extensions that are intentionally designed to go beyond the trim line or can be taken from cutout areas inside the part.Section cuts from detail areas can be examined for discrepancies Test coupons can be machined from the sections and mechanically tested to ensure the structural capa- bility of the part and verify the quality of the fabrication process.Using coupons in this way can satisfy destructive testing requirements and process control requirements(Ref.Section 3.2.2). 3.3.3.4 Implementation guidelines The frequency of destructive tests are dependent on part type and experience.If the producer has significant fabrication experience,complex parts may not require periodic destructive testing,but only a first article dissection.For low experience with complex parts,periodic inspection with increasing intervals may be preferable.Critical(safety of flight)parts warrant consideration for destructive testing. Examination and testing of trim sections can be carried out on a more frequent basis and at less cost than full part dissection.Quality assurance can be enhanced by using more frequent and less elaborate trim section examinations. Destructive tests should be conducted before the part leaves the factory.Periodic destructive tests monitor the manufacturing processes to assure the quality of parts.If a problem does occur,the periodic inspections bracket the number of suspect parts.Not every part series needs to be examined.If many parts reflect the same type of configurations and complexity,they can be pooled together for sampling purposes.Parts made on tools fabricated from one master splash can also be grouped together. Sampling:A typical sampling plan might include first article full part dissection followed by periodic in- spections employing dissection of trim sections.The periodic inspection intervals can vary depending on success rate.After a few successful destructive tests,the interval can be increased.If nonconforming ar- eas are found in destructive tests,the inspection interval can be tightened up.If problems are found in service,additional components from the same production series can be dissected to assure that the prob- lem was isolated. For the trim section approach,periodic destructive tests can be conducted at smaller intervals since the cost is much less.Small intervals may be especially desirable in the case of critical parts. For first article inspection,one of the first few articles may be chosen to represent first article.Some of the reasons for not stipulating the very first structure built are:(1)it may not be as representative of the production run because of lessons learned and special handling;and(2)another part with processing problems or discrepancies may reveal far more information. 3-7

MIL-HDBK-17-3F Volume 3, Chapter 3 Quality Control of Production Materials and Processes 3-7 3.3.3.3 Destructive test approaches There are two primary categories of destructive tests: dissection of the full part or examination of trim sections of the part. Full dissection, generally done for the first part from a new tool, gives a complete ex￾amination of the part, but is expensive to perform. Examination of excess trim sections is the preferable approach whenever possible. The part is not destroyed, structural details can still be examined and me￾chanical test specimens can be obtained. Full Part Dissection: Full part dissection is the approach often envisioned when the term "destructive testing" is mentioned. Since it prevents future use of the part, full part dissection should be reserved for parts that meet the following criteria: • Areas cannot be adequately inspected by NDI • Part is complex and there is a low experience level for working with the structural configuration or fabrication process • Part is net trim; detail areas of interest cannot be examined using excess trim areas or part ex￾tensions. Trim Sections: Examination and testing of trim sections offers a balance of quality assurance and cost. Trim sections can be part extensions that are intentionally designed to go beyond the trim line or can be taken from cutout areas inside the part. Section cuts from detail areas can be examined for discrepancies. Test coupons can be machined from the sections and mechanically tested to ensure the structural capa￾bility of the part and verify the quality of the fabrication process. Using coupons in this way can satisfy destructive testing requirements and process control requirements (Ref. Section 3.2.2). 3.3.3.4 Implementation guidelines The frequency of destructive tests are dependent on part type and experience. If the producer has significant fabrication experience, complex parts may not require periodic destructive testing, but only a first article dissection. For low experience with complex parts, periodic inspection with increasing intervals may be preferable. Critical (safety of flight) parts warrant consideration for destructive testing. Examination and testing of trim sections can be carried out on a more frequent basis and at less cost than full part dissection. Quality assurance can be enhanced by using more frequent and less elaborate trim section examinations. Destructive tests should be conducted before the part leaves the factory. Periodic destructive tests monitor the manufacturing processes to assure the quality of parts. If a problem does occur, the periodic inspections bracket the number of suspect parts. Not every part series needs to be examined. If many parts reflect the same type of configurations and complexity, they can be pooled together for sampling purposes. Parts made on tools fabricated from one master splash can also be grouped together. Sampling: A typical sampling plan might include first article full part dissection followed by periodic in￾spections employing dissection of trim sections. The periodic inspection intervals can vary depending on success rate. After a few successful destructive tests, the interval can be increased. If nonconforming ar￾eas are found in destructive tests, the inspection interval can be tightened up. If problems are found in service, additional components from the same production series can be dissected to assure that the prob￾lem was isolated. For the trim section approach, periodic destructive tests can be conducted at smaller intervals since the cost is much less. Small intervals may be especially desirable in the case of critical parts. For first article inspection, one of the first few articles may be chosen to represent first article. Some of the reasons for not stipulating the very first structure built are: (1) it may not be as representative of the production run because of lessons learned and special handling; and (2) another part with processing problems or discrepancies may reveal far more information

MIL-HDBK-17-3F Volume 3,Chapter 3 Quality Control of Production Materials and Processes Potential areas:Potential areas and items to examine include: Primary load paths within the part, Areas that showed indications from non-destructive inspection, Tool markoff near cocured details, Ply drop offs at a taper, Ply wrinkles, Resin starved and resin rich areas, Corner radii and cocured details, Core to face sheet fillets, Tapered core areas. 3.3.3.5 Test types Both full part dissection and trim sections involve examination of detail areas.After machining the de- tail areas,photomicrographs can be obtained to examine the microstructure.Another type of destructive testing is ply verification.Only a small section is need to perform a deply or grind down to verify that the plies are laid up in the correct stacking sequence and orientation.For machine lay-up,this procedure should not be necessary after initial validation.To investigate items such as ply lay-up,potential ply wrin- kles and porosity,initial core plugs can be taken at fastener hole locations and photomicrographs can be developed. When mechanically testing specimens that were machined from trim sections,the coupons should be tested for the critical failure mode for that part or that area of the part.Tests addressing typical failure modes are unnotched compression,open hole compression and interlaminar tension and shear. 3.4 STATISTICAL PROCESS CONTROL 3.4.1 Introduction Since composites exhibit a strong capacity for variability,the tools used to identify,assess,and hope- fully control variability become critical.Statistical process control is a term used to tie together several different aspects of statistical and other quality methods. 3.4.2 Quality tools There are several methods which form the bulk of SPC efforts.They range from fairly simple method- ologies for gathering and evaluating data,to sophisticated statistical techniques for answering very spe- cific questions.What is described in the following sections should not be construed as a comprehensive evaluation.There are many other techniques,or variants on the techniques discussed,which can be re- viewed in the literature. 3.4.3 Gathering and plotting data One of the first concepts in evaluating data is to collect them in a rigorous manner.Once the data have been gathered,the data should almost always be plotted in some fashion.It can be very difficult to discern even moderate trends in tabular data.This can be true with even just a handful of data points.In many cases,the same data can and should be plotted in several different manners,looking for patterns and relationships between factors.or trends over time. 3.4.4 Control charts One of the specific ways that data can be plotted is as a part of a control chart.With control charts, variability in a process output is measured.The sources of variation are partitioned into chance or com- 3-8

MIL-HDBK-17-3F Volume 3, Chapter 3 Quality Control of Production Materials and Processes 3-8 Potential areas: Potential areas and items to examine include: Primary load paths within the part, Areas that showed indications from non-destructive inspection, Tool markoff near cocured details, Ply drop offs at a taper, Ply wrinkles, Resin starved and resin rich areas, Corner radii and cocured details, Core to face sheet fillets, Tapered core areas. 3.3.3.5 Test types Both full part dissection and trim sections involve examination of detail areas. After machining the de￾tail areas, photomicrographs can be obtained to examine the microstructure. Another type of destructive testing is ply verification. Only a small section is need to perform a deply or grind down to verify that the plies are laid up in the correct stacking sequence and orientation. For machine lay-up, this procedure should not be necessary after initial validation. To investigate items such as ply lay-up, potential ply wrin￾kles and porosity, initial core plugs can be taken at fastener hole locations and photomicrographs can be developed. When mechanically testing specimens that were machined from trim sections, the coupons should be tested for the critical failure mode for that part or that area of the part. Tests addressing typical failure modes are unnotched compression, open hole compression and interlaminar tension and shear. 3.4 STATISTICAL PROCESS CONTROL 3.4.1 Introduction Since composites exhibit a strong capacity for variability, the tools used to identify, assess, and hope￾fully control variability become critical. Statistical process control is a term used to tie together several different aspects of statistical and other quality methods. 3.4.2 Quality tools There are several methods which form the bulk of SPC efforts. They range from fairly simple method￾ologies for gathering and evaluating data, to sophisticated statistical techniques for answering very spe￾cific questions. What is described in the following sections should not be construed as a comprehensive evaluation. There are many other techniques, or variants on the techniques discussed, which can be re￾viewed in the literature. 3.4.3 Gathering and plotting data One of the first concepts in evaluating data is to collect them in a rigorous manner. Once the data have been gathered, the data should almost always be plotted in some fashion. It can be very difficult to discern even moderate trends in tabular data. This can be true with even just a handful of data points. In many cases, the same data can and should be plotted in several different manners, looking for patterns and relationships between factors, or trends over time. 3.4.4 Control charts One of the specific ways that data can be plotted is as a part of a control chart. With control charts, variability in a process output is measured. The sources of variation are partitioned into chance or com-

MIL-HDBK-17-3F Volume 3,Chapter 3 Quality Control of Production Materials and Processes mon cause,and assignable variation.Data are plotted as it is generated by a process,and a simple set of rules can be used to determine if an assignable cause should be pursued.With proper application, issues can be identified and addressed prior to reaching rejectable levels. 3.4.5 Process capability A fundamental question for a manufacturing process is given the variability present,what percentage of product would meet specification requirements.Numbers representing this concept are termed meas- ures of process capability.The process variability,represented by the standard deviation,is used to es- tablish tolerance limits which describe where almost all of the product should fall.For one measure of process capability the range between these limits is compared to the specification range. The lower the quantity of product produced outside the specification limits,the more capable the process.Various ratios can be used to assess process capability.An important issue is whether the proc- ess mean is centered between the specification limits,and the implications if it is not. 3.4.6 Troubleshooting and improvement Many times a new process requires characterization and development,or improvements become necessary for an established process.A process that was once in control may not be any longer for rea- sons which are not well understood.In situations such as these,tools for troubleshooting established processes,and making improvements to new or established process become valuable.Three common methods are described below. 3.4.6.1 Process feedback adjustment Introduction Process control is achieved through both process monitoring and adjustment.Process monitoring is accomplished through Statistical Process Control(SPC).including tools such as process control(or She- whart)charts and cumulative sum(Cusum)charts.These are used to interrogate the process or system to determine its stability.Process adjustment is used to bring a process back from drifting and is usually termed Engineering Process Control(EPC).SPC and EPC do not compete but can work together. They can be adapted for environments where an appreciable cost is associated with making a change to the system or taking the measurement.These look at minimizing the overall cost of controlling the system using also the cost of being off the process target.EPC can also implement bounded adjust- ment charts that will dictate both the necessity and magnitude for an adjustment to the process.Finally, the monitoring of a process that is undergoing feedback control is covered. The stable,stationary state,which is the environment under which traditional Statistical Process Con- trol(SPC)is supposed to take place,is actually very difficult to attain and maintain.While the more famil- iar technique of process monitoring through the use of control charts can help achieve this control,fre- quently processes require adjustment of parameters to attain the desired output.While some of the tools and procedures are similar to those for process monitoring.the intent and approach is actually quite dif- ferent. Process monitoring is defined as the use of control charts that are used to continuously interrogate the stability of the process being investigated.When unusual behavior is detected,assignable causes as the source of the behavior are searched for,and if possible,eliminated.This technique has been widely used in the standard parts industry as SPC. Process adjustment utilizes feedback control of some variable related to the desired output in order to keep the process as close as possible to a desired target.The origins of this procedure are in the process industry,which is termed Engineering Process Control(EPC). 3-9

MIL-HDBK-17-3F Volume 3, Chapter 3 Quality Control of Production Materials and Processes 3-9 mon cause, and assignable variation. Data are plotted as it is generated by a process, and a simple set of rules can be used to determine if an assignable cause should be pursued. With proper application, issues can be identified and addressed prior to reaching rejectable levels. 3.4.5 Process capability A fundamental question for a manufacturing process is given the variability present, what percentage of product would meet specification requirements. Numbers representing this concept are termed meas￾ures of process capability. The process variability, represented by the standard deviation, is used to es￾tablish tolerance limits which describe where almost all of the product should fall. For one measure of process capability the range between these limits is compared to the specification range. The lower the quantity of product produced outside the specification limits, the more capable the process. Various ratios can be used to assess process capability. An important issue is whether the proc￾ess mean is centered between the specification limits, and the implications if it is not. 3.4.6 Troubleshooting and improvement Many times a new process requires characterization and development, or improvements become necessary for an established process. A process that was once in control may not be any longer for rea￾sons which are not well understood. In situations such as these, tools for troubleshooting established processes, and making improvements to new or established process become valuable. Three common methods are described below. 3.4.6.1 Process feedback adjustment Introduction Process control is achieved through both process monitoring and adjustment. Process monitoring is accomplished through Statistical Process Control (SPC), including tools such as process control (or She￾whart) charts and cumulative sum (Cusum) charts. These are used to interrogate the process or system to determine its stability. Process adjustment is used to bring a process back from drifting and is usually termed Engineering Process Control (EPC). SPC and EPC do not compete but can work together. They can be adapted for environments where an appreciable cost is associated with making a change to the system or taking the measurement. These look at minimizing the overall cost of controlling the system using also the cost of being off the process target. EPC can also implement bounded adjust￾ment charts that will dictate both the necessity and magnitude for an adjustment to the process. Finally, the monitoring of a process that is undergoing feedback control is covered. The stable, stationary state, which is the environment under which traditional Statistical Process Con￾trol (SPC) is supposed to take place, is actually very difficult to attain and maintain. While the more famil￾iar technique of process monitoring through the use of control charts can help achieve this control, fre￾quently processes require adjustment of parameters to attain the desired output. While some of the tools and procedures are similar to those for process monitoring, the intent and approach is actually quite dif￾ferent. Process monitoring is defined as the use of control charts that are used to continuously interrogate the stability of the process being investigated. When unusual behavior is detected, assignable causes as the source of the behavior are searched for, and if possible, eliminated. This technique has been widely used in the standard parts industry as SPC. Process adjustment utilizes feedback control of some variable related to the desired output in order to keep the process as close as possible to a desired target. The origins of this procedure are in the process industry, which is termed Engineering Process Control (EPC)

MIL-HDBK-17-3F Volume 3.Chapter 3 Quality Control of Production Materials and Processes Control Charts Control charts that are used to observe frequencies and proportions are covered in the Section 3.4.4. Different types of charts are used for monitoring of measurement data.These look at a sample aver- age and range and are known as X bar and R charts.Some of the useful simplifications used for fre- quency and proportion data are not applicable for measurement type data.The same general terminol- ogy is used except as noted. Several rules,some of which are industry or even company specific,can be applied to these control charts.The most widely known set of rules is the Western Electric rules that are applied to the control chart to determine if a deviation warrants the search for an assignable cause to be eliminated. There are assumptions made for the application of these control charts.While some mild violation of these assumptions is not usually catastrophic,an unstable system can result in inappropriate warning and action limits Process Adjustment In the process regulation the object is not to test hypotheses about the likelihood of a set of data indi- cating special cause,but rather statistical estimation of a disturbance to the system which is then com- pensated for in various manners. As an indication of the differences between the objectives between process monitoring and process adjustment,waiting to implement a process adjustment until the process monitoring indicates a statisti- cally significant deviation as a control strategy would usually lead to excessive process output variation. For many processes acceptable control may not be achievable without process adjustment at some interval.These adjustments must not be made in an arbitrary fashion for consistent results.Important concepts in implementing process adjustment are the processes resistance to change,termed inertia, and the use of models to predict the future output of the process. A unit change in the adjustment variable will not most likely result in a unit change of the process out- put.The relationship between these factors is termed the system gain.In attempting to predict the output of the process,it is useful the split the response in the categories of the white noise,and the system drift. It is important to note that many processes,if left alone,will continuously drift away from a target value.Because of this drift,a low value is more likely to be followed by another low value,termed auto- correlation.These changes may be in the form of step changes,spikes,or changes in slope.With proc- ess adjustment,it is attempted to estimate the direction of the process,and then adjust the process, compensating to keep the process directed toward the target value. The types of disturbances that induce this drift can be environmental changes such as temperature and humidity,or changes in the composition of the input materials.Whether or not these variable have been identified,some sort of feedback control may be necessary to compensate for their effect,allowing the process output to return to the target value.These feedback adjustment procedures have a direct re- lation to the types of automatic control methods used in the process industry. While some sort of alteration of the process based on SPC input is frequently employed,a consistent methodology is seldom implemented.By using some sort of feedback adjustment scheme,the variation in the process output can be reduced several-fold in many applications. What has allowed feedback adjustment the opportunity for more widespread application outside the traditional chemical process industries is the determination that substantial errors in modeling the system, and even significant errors in applying the feedback adjustment result in minimal effects on the process output variability. 3-10

MIL-HDBK-17-3F Volume 3, Chapter 3 Quality Control of Production Materials and Processes 3-10 Control Charts Control charts that are used to observe frequencies and proportions are covered in the Section 3.4.4. Different types of charts are used for monitoring of measurement data. These look at a sample aver￾age and range and are known as X bar and R charts. Some of the useful simplifications used for fre￾quency and proportion data are not applicable for measurement type data. The same general terminol￾ogy is used except as noted. Several rules, some of which are industry or even company specific, can be applied to these control charts. The most widely known set of rules is the Western Electric rules that are applied to the control chart to determine if a deviation warrants the search for an assignable cause to be eliminated. There are assumptions made for the application of these control charts. While some mild violation of these assumptions is not usually catastrophic, an unstable system can result in inappropriate warning and action limits. Process Adjustment In the process regulation the object is not to test hypotheses about the likelihood of a set of data indi￾cating special cause, but rather statistical estimation of a disturbance to the system which is then com￾pensated for in various manners. As an indication of the differences between the objectives between process monitoring and process adjustment, waiting to implement a process adjustment until the process monitoring indicates a statisti￾cally significant deviation as a control strategy would usually lead to excessive process output variation. For many processes acceptable control may not be achievable without process adjustment at some interval. These adjustments must not be made in an arbitrary fashion for consistent results. Important concepts in implementing process adjustment are the processes resistance to change, termed inertia, and the use of models to predict the future output of the process. A unit change in the adjustment variable will not most likely result in a unit change of the process out￾put. The relationship between these factors is termed the system gain. In attempting to predict the output of the process, it is useful the split the response in the categories of the white noise, and the system drift. It is important to note that many processes, if left alone, will continuously drift away from a target value. Because of this drift, a low value is more likely to be followed by another low value, termed auto￾correlation. These changes may be in the form of step changes, spikes, or changes in slope. With proc￾ess adjustment, it is attempted to estimate the direction of the process, and then adjust the process, compensating to keep the process directed toward the target value. The types of disturbances that induce this drift can be environmental changes such as temperature and humidity, or changes in the composition of the input materials. Whether or not these variable have been identified, some sort of feedback control may be necessary to compensate for their effect, allowing the process output to return to the target value. These feedback adjustment procedures have a direct re￾lation to the types of automatic control methods used in the process industry. While some sort of alteration of the process based on SPC input is frequently employed, a consistent methodology is seldom implemented. By using some sort of feedback adjustment scheme, the variation in the process output can be reduced several-fold in many applications. What has allowed feedback adjustment the opportunity for more widespread application outside the traditional chemical process industries is the determination that substantial errors in modeling the system, and even significant errors in applying the feedback adjustment result in minimal effects on the process output variability