Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 153
--> 13 Process Precision And Metrology The ability to produce quality products hinges on four key competencies: modeling of process form and precision levels, design tolerancing of parts and products, selecting production processes that match part specifications, and applying quantitative measurement methods for inspection and process control. The first two—process modeling and design tolerancing—are of primary importance and drive the second two; however, both are surprisingly ill-understood in a scientific sense. Mathematical models for predicting process precision, and quantitative precision and inspection data for actual processes, are scarce and often proprietary. Tolerancing today is based on informal definitions and on tolerance-assignment and inspection procedures of limited generality and validity. As a result, tolerancing; process selection and control; and, to some extent, metrology and nondestructive evaluation still rely largely on tradition. Process modeling is discussed in Chapter 10. This chapter deals with process precision, a crucial but often overlooked component of quality technology.1 Spatial precision (i.e., issues of form and fit) is particularly important, since it pervades almost all discrete goods production. This chapter reviews the current research status and needs of process precision and metrology. It concludes with recommended research opportunities. 1 Quality technology encompasses all those technologies necessary to assure that a product can be/is produced to meet the design specification. It can be divided into two areas: (1) precision and metrology and (2) nondestructive evaluation.
OCR for page 154
--> Research Status and Needs The emphasis thus far in this report has been on processes and on the knowledge and technologies needed to implement them. An alternative approach is to shift the focus to parts and products and view unit manufacturing processes merely as the means to make quality parts and products. This approach exposes new issues that strongly influence the usefulness of unit processes and an overall ability to make quality goods. These issues arise, because manufacturing and assembly processes produce parts and products that vary. Variations in part geometry, as graphically illustrated in Figure 13-1, could result from inherently imprecise processes, or from variations in process control. Control variations could be due to a lack of knowledge concerning the process variables, inadequate means of process control, indifference to process control, etc. Distinguishing between the imprecise execution of a process and the execution of an imprecise process is at the heart of precision engineering. Most processes underlying Figure 13-1 Tolerance as a function of components metalworking processes (adapted from Kalpakjian, 1992).
OCR for page 155
--> manufacturing and assembly are quite precise. Consequently, mechanisms that accommodate and control variability are woven throughout the entire manufacturing system. When parts and products are designed, dimensional tolerances are assigned to specify allowable variations. Parts are then manufactured and products assembled by selecting processes that are repeatable and precise enough to meet the specified tolerances. Thus three key producibility themes emerge: design to accommodate variability arising from the control process design, design to ensure that the realized process variations do not exceed the design tolerance, and design to minimize the dispersion of variations within the allowed ranges of variability through careful control of manufacturing and assembly processes. The nominal design of products, subassemblies, and parts is driven mainly by functionalism—that is, by what the items must do. The main tools are parametric modeling (e.g., dimensioned drawings, computer-aided design models) and analytical procedures (e.g., finite-element analysis). The next stage of design, detailed design, supplies the details that were ignored in nominal design and accommodates manufacturing and assembly variability by specifying allowable variations in spatial forms and relations. Interchangeable assembly usually becomes the dominant constraint. The main working tool is tolerancing (i.e., a set of standardized practices for determining and specifying variations). Current tolerancing standards prohibit specification by process and by reference to other artifacts. As a result, parts must be specified as free-standing geometric entities, rather than by procedures for making them or by requirements that they mate with other parts. (These restrictions were motivated by procurement problems; they have the intent of preserving full manufacturing freedom and facilitating competitive ''out-sourcing.'') Manufacturing and assembly planning can be simplistically viewed as a mix-and-match exercise in which processes of adequate precision are selected to produce the various features of a part or to mate parts in assembly and then are sequenced to meet process, functional, and cost constraints. In physical manufacturing, parts are made using unit processes. The processes must be controlled passively or actively for predictable results, and every form of control uses some form of process model. Physical assembly is analogous to physical manufacturing in that unit assembly processes (e.g., align, insert, screw, cement, etc.) are used to produce subassemblies and, finally, complete products. Assembly processes must be controlled passively or actively for predictable results, and every form of control uses some form of assembly process model. The conformance testing (i.e., inspection) phase determines whether parts meet their specifications. The main techniques are conventional parametric measurements and binary (i.e., accept or reject) inspection using special gages.
OCR for page 156
--> Testing strategies vary from 100 percent inspection of all toleranced features of all parts through statistical sampling of small lots of parts to no inspection at all (when the manufacturing processes are very tightly controlled). Statistical design of experiments during the process development phase could guide the establishment of a statistical process control system that will lead to the minimum inspection program required to assure high quality (Taguchi et al., 1989). Performance testing of the final product is the analog to part conformance testing.2 Thus, as one moves downstream from nominal design, the control of variability becomes the major production concern. Variability arises from the physical processes used to make and assemble parts. Four central factors are involved: tolerancing: the means used to specify allowable variations in parts and products; metrology: the means used to determine whether artifacts meet their design specifications; process form and precision modeling: the means used to specify, for a particular process, the expected accuracy based on knowledge of the sources of imprecision; and process planning: the means used to select and sequence processes to make or assemble parts. Tolerancing and process modeling dominate, because they influence, or provide critical data to, the other two factors. Some of the current topics in process precision and metrology are discussed below. They include issues in dimensional scale and precision in manufacturing, dimensional tolerances and metrology, process planning, and process modeling. Dimensional Scale and Precision in Manufacturing Table 13-1 indicates that typical manufactured products vary greatly in scale and in their requirements for precision. In the table, dimension, D, is a normal size parameter and tolerance, T, is a typical limit on the allowable variation in D. The T/D precision ratio (i.e., fractional linear variability) is one measure of 2 The tools and methods used for performance testing are highly dependent on the nature of the product.
OCR for page 157
--> precision. The scale of the items, as well as their T/D precision ratios, spans a range of 106. However, most components of conventional products (e.g., motor vehicles, aircraft, appliances, and machinery of many kinds) vary in scale by a factor of about 102 within a larger absolute range extending from about 1 mm to 10 m and have precision ratios in the range 10-3 to 10-5 (again, a factor of 102). Conventional parts and products are a main focus, since they constitute well over half of all discrete goods by dollar value and contribute close to 10 percent of the gross national product, and they are produced using the unit processes discussed earlier. Unit processes for the most part have been designed to operate in the scale and precision ranges spanned by these products. Figure 13-2 distinguishes "precision" and "ultraprecision" machining from "normal" machining in terms of dimensional scale and tolerance. Precise and ultraprecise manufacturing and measurement processes are quite specialized and limited in applicability, but the volume, value, and technical importance of the products requiring processing in these regimes are growing. Obviously, this requires improvements in existing processes, either by implementing the practices of the next higher quality level or by improving the existing process. In either case, a cultural change is often necessary to institutionalize the higher quality level. A similar plot can be made for other unit processes, such as those listed in Table 13-1; the trends for those unit processes are directly analogous to the case of machining. Dimensional Tolerances and Metrology Parts are specified in terms of their nominal (ideal) shapes and nominal material properties, with allowable variations on both. Assemblies are specified in terms of part associations and performance specifications, again with allowable variations on both. The trend toward tighter tolerances is being motivated by a desire for longer life, faster but quieter operation, greater efficiency, and simplified assembly operations. For example, sorting piston pins for proper match into the piston was once a common practice. It is now considered obsolete, because accurate machining is currently inexpensive enough to enable all parts to match. By contrast, the fit required for diesel fuel injector plungers is so critical that current technology does not allow for economical manufacture of parts for universal assembly.
OCR for page 158
--> Table 13-1 Dimensional Scale and Precision for a Range of Manufactured Items (Swyt, 1992) Part or Subassembly Geometric Attribute Process Dimension D Tolerance T Ratio T/D Auto door assembly Panel size/position Die stamping 1 m 1 mm 10-3 Auto engine piston Diameter/ cylindricity Machining 100 mm 7-8 mm 10-4 Magnetic read/write heads Cut-face position Diamond slicing 125 mm 2.5 mm 2 ×10-5 Wafer micropattern Pattern position Lithography 250 mm 0.2 µm 10-6 Optical Fibers Fiber diameter Die drawing 125 µm 0.2 mm 10-3 Integrated circuit Line width Lithography 0.5 µm 50 nm 10-1 The conformance of parts and assemblies to geometrical specifications is assessed by physical measurements. The term "dimensional metrology" covers the various instruments and techniques used for making such measurements. There
OCR for page 159
--> Figure 13-2 Three relatively distinct manufacturing regimes (adapted from Wirtz, 1991). are three main classes of instruments and techniques in use today, all of which require physical contact with the specimen:3 Manual instruments and so-called open setup, or manual, methods . These include rules, calipers, micrometers, dial indicators, and surface plates. These are the classic measurement instruments used in manufacturing. They are low cost and widely available. However, their use is time consuming and subject to human error. Functional (hard) gages. These range from simple plug gages to custom fixtures designed to simulate features of mating parts. They effectively pick up maximum or minimum features with assurance and are made either fully mechanical (i.e., go/no-go) or with an analog transducer (i.e., dial indicator) that provides readout of deviations. Custom fixtures are expensive and difficult or 3 Contact technologies are relatively precise and robust, but they are inherently slow and therefore expensive. They are likely to be replaced gradually with faster noncontacting technologies based on wave phenomena.
OCR for page 160
--> impossible to modify and hence are justified only for critical or large-volume production. Coordinate measuring machines (CMMs). These are very precise devices that measure the positions of points on the surface of a part by probing ("touching"), usually under computer control. CMMs resemble machine tools but are more precise and are often temperature-compensated. The digital data collected by a CMM must be processed by algorithms to assess whether the part conforms to its tolerance specifications, and this raises issues that cut to the core of tolerancing and metrology. There is growing interest in simulating hard gages with properly programmed CMMs for low-volume production. The current American tolerancing standard, ANSI Y14.5M-1982 (referred to in this report as "Y14.5" for brevity), evolved from shop and drafting-room practice; its roots lie in gaging technology, although it is not a gaging standard (ANSI, 1982). The standard is best viewed as a collection of sensible principles, defined mainly through examples cast in prose and graphics. There is no companion standard to specify how pans are to be measured to assess conformance to the definitions in Y14.5. The informal definitions of Y145.5, and the absence of a measurement standard, caused few problems in the pre-CMM era, when inspection was done manually or with automatic gages. When CMMs arrived, however, inspectors found that CMM results did not always agree with traditional inspection results. This is called "methods divergence," and its recognition triggered increasingly strident warnings in the 1980s about a "metrology crisis." A second problem, "specification ambiguity," was exposed when CMM and computer-aided design programmers found that some of the prose and graphic definitions in Y14.5 are ambiguous. The American Society of Mechanical Engineers and the National Science Foundation convened a workshop in 1988 to discuss these matters. The major recommendations are as follows (Tipnis, 1989). The Y14.5 standard should be mathematized and generalized, so that provably correct measurement procedures can be designed to assess conformance. Education and research in tolerancing and metrology should be expanded markedly. The standards community responded vigorously to the 1988 recommendations by establishing new committees in 1989 to mathematize Y14.5 and to address measurement methods. A new standard, "Y14.5.1: Mathematical Definition of Dimensioning and Tolerancing Principles," will be issued in late 1994 to accompany a new edition of Y14.5, and a second new standard,
OCR for page 161
--> "B89.3.2: Dimensional Measurement Methods," is expected to be issued in 1996 (Srinivasan and Voelcker, 1993). The advent of these new standards, in which mathematics is the defining medium, probably marks the end of the 200-year era in which tolerancing and metrology evolved as industrial practices without strong theoretical underpinnings. It is important to note that while a new era with a precise language (mathematics) for defining tolerances has begun, there is no proper mathematical theory to govern the use and interaction of tolerances. A theory can be expected, however, because research in tolerancing and metrology is growing (Menon and Voelcker, 1993; Menon and Robinson, 1993). Unfortunately there has been little progress in introducing these topics into engineering curricula. It is worth noting that there is a movement in Europe advocating the adoption of vectorial tolerancing as a replacement for, or at least a co-equal alternative to, the International Standards Organization brand of geometric tolerancing. Vectorial tolerancing was formulated by Adolph Wirtz of Switzerland (Wirtz, 1991); Figure 13-3 conveys some of its essential elements. The basic concept is to cast tolerances in terms of parameters that are important in manufacturing and inspection, with the current formulation oriented toward machine tools and CMMs. This has the advantage of removing the language mismatch noted later in this chapter, but it does so only for one or two families of processes. It is a retrograde step in the sense that it re-establishes the coupling between design specifications and manufacturing methods that was deemed harmful in the early days of geometric tolerancing. Process Planning Process planning directly links manufacturing to design. Process planning can be described using the simple example of machining the bracket described in Figure 13-4. Observe that the plan shown as output in Figure 13-5 is influenced strongly by the hole tolerances. Specifically, the central hole D should be generated first, because it serves as a position datum for the four-hole pattern.4 Further, at least two processes—drilling, followed by boring—are required to meet the position and perpendicularity tolerances on D, and a final light reaming might be needed to meet the cylindricity tolerance.5 The four holes 4 This will not matter on numerically controlled machines if D is spot-drilled. 5 Drilling is an imprecise hole-making process. Boring and reaming are hole-finishing operations that have different form and positional accuracies.
OCR for page 162
--> Figure 13-3 An illustration of (a) vectoring tolerancing and (b) its potential convenience (adapted from Wirtz, 1991).
OCR for page 163
--> Figure 13-4 Example bracket.
OCR for page 164
--> Figure 13-5 Planning the machining of the holes of the bracket in Figure 13-4. in the pattern do not require special finishing, because their size tolerances are loose, but they do require spot-drilling for positional accuracy. Finally, note that the plan shown in the figure does not require special tooling. However, if a different process family had been used—molding, for example—then tooling (e.g., mold-system) design would have been a major activity. Experienced machinists can easily construct plans such as that shown in Figure 13-5, because they possess a wealth of experiential data and subtle reasoning powers. To date, researchers have tried to codify, or to replace, this data and logic with automated process planning systems with little success.6 Twenty years of research have failed to produce automatic machining planners 6 An experienced machinist knows semiquantitatively, for example, that boring is positionally and orientationally more accurate then reaming but less accurate in terms of cylindrical form.
OCR for page 165
--> that have a reasonable degree of generality. To understand why, observe that the study of process planning raises two basic issues: what knowledge (e.g., models and data) is needed to do planning, and how is that knowledge used to produce plans? The first issue will receive focus here, because it is the sine qua non for understanding and automating planning (and deficiencies in this area almost surely are responsible for the lack of progress just noted). Readers interested in the second issue, planning logic, are referred to the research literature cited in Manufacturing Intelligence (Wright and Bourne, 1988) and Computer-Aided Manufacturing (Chang et al., 1991). The main elements of process planning are selecting particular processes and operating parameters that can produce the specified part, either as a whole (e.g., by molding or casting) or as a collection of distinct features (e.g., machined features, forged features), and sequencing the selected processes appropriately. The first step requires that part geometry, both ideal form and allowable variations, be matched with process capabilities. Thus the central knowledge needed for process planning is a set of process models that prescribe for each process the nominal forms it can produce and the variability associated with each form. Process Modeling A complete process model covers more than form and form precision; it also deals with the bulk and surface properties of the processed material, with the energy needed to apply the process, with the scaling laws of the process, and so forth. This section, however, focuses on form and fit and does not discuss material-property and kinetic modeling. The entire area of process-induced form modeling requires a significant increase in research, because there are few process models that provide explicit representations of the forms that can be produced, and even fewer that provide the precision data useful for manufacturing planning and tooling design. Machining has been studied longer (for more than 100 years) and more intensively than other processes and therefore is a logical first place to look for form models. The obvious first sources, engineering texts on machining, yield little. However, machining handbooks and broad-scale manufacturing-process tests (Kalpakjian, 1992) have tabulations of form capabilities (such as Table 13-2) and linear precision graphs (Figure 13-6 has conservative tolerances; some applications demand much greater precision). Note that bounding-surface equations can be associated easily with most of the forms in Table 13-2.
OCR for page 166
--> The data summarized in Table 13-2 and Figure 13-6 illustrate the level of well-known current machining form and precision models.7 The process models cannot cope effectively with 3-, 4-, and 5-axis machining, and they are virtually useless for selecting processes to meet given geometric tolerances. Recent and current research, however, are now providing models that, while still experimental, are more powerful and promising. For example, mathematical and computational tools offered by solid modeling enable the nominal form-modifying effects of numerical control machining to be modeled mathematically and simulated computationally with increasing fidelity (Menon and Robinson, 1993; Menon and Voelcker, 1993). Research on improved precision models for machining is more scattered, but several references provide useful introductory discussions and references (See Chang et al., 1991; Wood, 1993, for a different approach based on fractal theory). Much of this work is based on a three-step paradigm: (1) postulate a generic relation between form error and parameters of the process, (2) construct a phenomenological model for a particular class of errors to particularize the generic relation, and (3) use the result to guide the collection of experimental data to create a specific model. Thus far the discussion of process modeling has focused on machining. Nominal-form and form-precision modeling seem to change character when bulk dynamic processes are addressed. Consider nominal form. In machining, form-feature process models (see Table 13-2) are needed for localized matching with part features. In casting and molding, nominal forms for entire parts are defined by molds, and a major component of process planning is designing mold-filling systems that will yield good parts (i.e., no voids or internal chill fronts, low residual stress, etc.). Elaborate computer programs to do the requisite hydrodynamical and thermodynamical analyses are under continuous development to aid such work. In deformation processes such as forging, nominal forms are defined by (final) dies, and the design of progressive die sets and preforms are major activities in process planning. Again, elaborate computer programs to analyze highly nonlinear plastic deformation phenomena are under continuous development to support such work. 7 More-extensive precision data are often available within companies for processes important to each company, but such information is usually proprietary.
OCR for page 167
--> Table 13-2 Forms Produced by Selected Classical Unit Machining Processes Process Form Drilling Straight cylindrical hole Boring Straight cylindrical hole Reaming Straight cylindrical hole Face milling Plane surface Profile milling Ruled surface normal to a plane End milling The above plus approximations to general curved surfaces Shaping, planing Plane surface Turning Surface rotationally symmetric about an axis The development of process-precision models for bulk dynamical processes involves such factors as shrinkage, creep, and residual stress prediction. Much of this work is in its infancy. Research in process modeling is diffuse, because it is distributed over the many families (and associated communities of researchers) discussed in Chapter 11. With the exception of machining, there does not appear to be much work aimed at generating the types of process-precision models needed for process planning, although some of the sensitivity analyses done for some processes may be convertible into precision models. The situation in machining is different. A group devoted to precision has a long history in the annals of machining and has paced the improvements summarized in Figures 13-2 and 13-7. Precision engineering is nearing recognition as a distinct subdiscipline within mechanical engineering and applied physics, and its new domestic professional society, the American Society for Precision Engineering, is growing and has a well-regarded journal, Precision Engineering, and strong ties to sister societies in other nations. The historical focus of precision engineers has been the construction of ultraprecise (for the era) artifacts and of the machinery to produce such artifacts. To pursue this work, precision engineers have had to probe deeply into the sources of process and machine variability, and thus process engineers should participate, or even take the lead, in building process-precision models (Slocum, 1992). Process planning, as a distinct activity and as an area for formal study, has a much shorter history. Most of its research roots lie in computer science and
OCR for page 168
--> Figure 13-6 Tolerance versus dimension data for various machining processes (adapted from Kalpakjian, 1992). artificial intelligence. These fields have a large literature centered on automatic planning, but little of it is devoted specifically to planning for manufacturing. It is probably safe to predict that automatic process planning for manufacturing will not progress very far until better process and process-precision models are available.
OCR for page 169
--> Figure 13-7 Precision machining domains (adapted from Taniguchi, 1983). Research Opportunities Manufacturing and assembly processes are inherently imprecise, and mechanisms to accommodate and control variability are woven through the entire manufacturing system. In design, the primary mechanisms are dimensional, and surface tolerances assigned to ensure interchangeable assembly while preserving function. Dimensional tolerances are governed by national and international standards and are required to be manufacturing-process independent. In manufacturing, process variability is kept within acceptable bounds by process control. Models for characterizing process variability (or process precision) are not standardized, are not well developed for many important processes, and usually are cast in terms of parameters natural to the process rather than for consistency with part tolerances. Thus there is a pervasive language mismatch between the mechanisms used to limit variability in design (i.e., tolerances) and the mechanisms used to describe the variability of processes (i.e., process precision models). This mismatch complicates process planning considerably and is a major inhibitor of further manufacturing automation. Based on the preceding discussion, the following areas emerge as strong candidates for future research emphasis. A major effort is needed to devise models and metrics for characterizing the precision of unit manufacturing processes in ways useful to process planners and designers. A two-pronged effort is likely to be needed, in
OCR for page 170
--> which models and metrics are first sought in the natural parameters of each process, and then means are sought to translate or interpret these parameters in terms of design tolerances. The models and metrics developed above should be concurrently applied to unit processes of interest, particularly regarding their scalability; their intrinsic or ultimate precision, as set by the underlying physics or chemistry of the process; and the precision currently available in typical industrial implementations. The efforts now under way to mathematize and generalize Y14.5, the American national tolerancing standard, and to define systematic measurement procedures based on a rigorous tolerance standard, should be encouraged. The harmonization of the American National Standards Institute and the International Standards Organization standards should also be encouraged. University researchers should be encouraged to participate in all of this work, including that of the standards committees, and to take the lead for some of it. One of the most important contributions these researchers can make is the systematic codifying of the scattered knowledge in these important fields. Academia should be encouraged to teach the essential principles of tolerancing, metrology, and process modeling in engineering design and manufacturing courses. At present, only process modeling receives more than cursory attention.
OCR for page 171
--> References ANSI. 1982. Dimensioning and Tolerancing. American National Standards Institute Standard Y14.5M 1982. New York: American Society of Mechanical Engineers. Chang, T-C, R.A. Wysk, and H-P. Wang. 1991. Computer-Aided Manufacturing. Englewood Cliffs, New Jersey: Prentice Hall Kalpakjian, S. 1992. Manufacturing Engineering and Technology. Reading, Massachusetts: Addison-Wesley. Menon, J.P., and H.B. Voelcker. 1993. Toward a comprehensive formulation of NC verification as a mathematical and computational problem. The Journal of Design and Manufacturing 3(4):263-277. Menon, J.P., and D.M. Robinson. 1993. High performance NC verification via massively parallel raycasting: Extensions to new phenomena and geometric domains. American Society of Mechanical Engineers Manufacturing Review 6(2): 141-154. Slocum, A.H. 1992. Precision Machine Design. Englewood Cliffs, New Jersey: Prentice Hall. Srinivasan, V., and H.B. Voelcker. Proceedings of the 1993 International Forum on Dimensional Tolerancing and Metrology, held by the American Society of Mechanical Engineers (ASME), in Dearborn, Michigan from June 17-19, 1993. CRTD-27. New York: American Society of Mechanical Engineers. Swyt, D.A., 1992. Challenges to NIST in dimensional metrology; The impact of Tightening Tolerances in the U.S. Discrete-Part Manufacturing Industry. JNISTIR 4757, 1992. Gaithersburg, Maryland: National Institute of Standards and Technology. Taguchi, G., E.A. Elsayed, and T. Hsiang. 1989. Quality Engineering in Production Systems. New York: McGraw-Hill. Taniguchi, H. 1983. Current status and future trends of ultraprecision machining and ultrafine materials processing. Annals of the CIRP (International Institute of Production Research) 2(2):611-622
OCR for page 172
--> Tipnis, V.A. 1989. Research needs and technological opportunities in mechanical tolerancing. National Science Foundation Workshop held September 28-October 2, 1988 in Orlando, Florida. New York: American Society of Mechanical Engineers. Wirtz, A. 1991. Vectorial tolerancing for production quality control and functional analysis in design. Pp. 77-84 in Proceedings of the International Working Seminar on Computer-Aided Tolerancing. University Park, Pennsylvania: Pennsylvania State University. Wood, K.L. 1993. Fractal-based tolerancing: Theory, dynamic process modeling, test bed development and experiments. Pp. 731-740 in the Proceedings of the 1993 National Science Foundation Design and Manufacturing Systems Conference held 6-8 January 1993 at the University of North Carolina at Charlotte. Dearborn, Michigan: Society of Manufacturing Engineers. Wright, P.K., and D.A. Bourne. 1988. Manufacturing Intelligence. Reading, Massachusetts: Addison-Wesley.
Representative terms from entire chapter: