Chapter13 - Data sources 13.1 Introduction and synopsis The...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Data sources 13.1 Introduction and synopsis The engineer, in selecting a material for a developing design, needs data for its properties. Engineers are often conservative in their choice, reluctant to consider material with which they are unfamiliar. One reason is this: that data for the old, well-tried materials are reliable, familiar, easily found; data for newer, more exciting, materials may not exist or, if they do, may not inspire confidence. Yet innovation is often made possible by new materials. So it is important to know where to find material data and how far it can be trusted. This chapter gives information about data sources. Chapter 14, which follows, describes case studies which illustrate data retrieval. As a design progresses from concept to detail, the data needs evolve in two ways (Figure 13.1). At the start the need is for low-precision data for all materials and processes, structured to facilitate screening. At the end the need is for accurate data for one or a few of them, but with the richness of detail which assists with the difficult aspects of the selection: corrosion, wear, cost estimation and the like. The data sources which help with the first are inappropriate for the second. The chapter surveys data sources from the perspective of the designer seeking information at each stage of the design process. Long-establisihed materials are well documented; less-common materials may be less so, posing problems of checking and, sometimes, of estimation. The chapter proper ends with a discussion of how this can be done. So much for the text. Half the chapter is contained in the Appendix, Section 13A. It is a catalogue of data sources, with brief commentary. It is intended for reference. When you really need data, this is the section you want. 13.2 Data needs for design Data breadth versus data precision Data needs evolve as a design develops (Figure 13.1). In the conceptual stage, the designer requires approximate data for the widest possible range of materials. At this stage all options are open: a polymer could be the best choice for one concept, a metal for another, even though the function is the same. Breadth is important; precision is less so. Data for this first-level screening is found in wide-spectrum compilations like the charts of this book, the Materials Engineering `Materials Selector' (1997), and the Chapman and Hall Materials Selector (1997).* More effective is software based on these data sources such as the CMS and CPS (1992, 1998) selection system. The easy access gives the designer the greatest freedom in considering alternatives. * Details in Further reading. 304 Materials Selection in Mechanical Design Fig. 13.1 Data needs and data structure for screening and for further information. The calculations involved in deciding on the scale and lay-out of the design (the embodiment stage) require more complete information than before, but for fewer candidates. Data allowing this second-level screening are found in the specialized compilations which include handbooks and computer databases, and the data books published by associations or federations of material producers. They list, plot and compare properties of closely related materials, and provide data at a level of precision not usually available in the broad, level 1, compilations. And, if they are doing their job properly, they provide further information about processability and possible manufacturing routes. But, because they contain much more detail, their breadth (the range of materials and processes they cover) is restricted, and access is more cumbersome. The final, detailed design, stage requires data at a still higher level of precision and with as much depth as possible, but for only one or a few materials. They are best found in the data sheets issued by the producers themselves. A given material (low-density polyethylene, for instance) has a range of properties which derive from differences in the way different producers make it. At the detailed-design stage, a supplier should be identified, and the properties of his product used in the design calculation. But sometimes even this is not good enough. If the component is a critical one (meaning that its failure could be disastrous) then it is prudent to conduct in-house tests, measuring the critical property on a ~ample of the material that will be used to make the component itself. Parts of power-generating equipment (the turbine disc for instance), or aircraft (the wing spar, the landing gear) and nuclear reactors (the pressure vessel) are like this; for Data sources 305 Table 13.1 Material data types Data type Numeric point data Numeric range data Boolean (yesho) data Ranked data Text Images Example Atomic number of magnesium: N , = 12 Thermal conductivity of polyethylene: A = 0.28 to 0.31 W/mK . 304 stainless steel can be welded: Yes Corrosion resistance of alumina in tap water (scale A to E): A Supplier for aluminium alloys: Alcan, Canada.. . these, every new batch of material is tested, and the batch is accepted or rejected on the basis of the test. Properties are not all described in the same way. Some, like the atomic number, are described by a single number (`the atomic number of copper = 29'); others, like the modulus or the thermal conductivity are characterized by a range (`Young's modulus for low-density polyethylene = 0.1-0.25 GPa', for instance). Still others can only be described in a qualitative way, or as images. Corrosion resistance is a property too complicated to characterize by a single number; for screening purposes it is ranked on a simple scale: A (very good) to E (very poor), but with further information stored as text files or graphs. The forming characteristics, similarly, are attributes best described by a list (`mild steel can be rolled, forged, or machined'; `zirconia can be formed by powder methods') with case studies, guidelines and warnings to illustrate how it should be done. The best way to store information about microstructures, or the applications of a material, or the functioning of a process, may be as an image - another data type. Table 13.1 sets out the data types which are typically required for the selection of materials and processes. 13.3 Screening: data structure and sources Data structure for screening and ranking To `select' means: `to choose'. But from what? Behind the concept of selection lies that of a kingdom of entities from which the choice is to be made. The kingdom of materials means: all 306 Materials Selection in Mechanical Design metals, all polymers, all ceramics and glasses, all composites as in Figure 5.2. If it is materials we mean to select, then the kingdom is all of these; leave out part, and the selection is no longer one of materials but of some subset of them. If, from the start, the choice is limited to polymers, then the kingdom becomes a single class of materials, that of polymers. A similar statement holds for processes, based on the kingdom of Figure 11.26. There is a second implication to the concept of selection; it is that all members of the kingdom must be regarded as candidates -they are, after all, there -until (by a series of selection stages) they are shown to be otherwise. From this arises the requirement of a data structure which is comprehensive (it includes all members of the kingdom) and the need for characterizing attributes which are universal (they apply to all members of the kingdom) and discriminating (they have recognizably different values for different members of the kingdom). Similar considerations apply to any selection exercise. We shall use it, in a later chapter, to explore the selection of manufacturing processes. In the kingdom of materials, many attributes are universal and discriminating: modulus and thermal conductivity are examples. Universal attributes can be used for ranking, the initial stage of any selection exercise (Figure 13.2, upper half). But if one or more screening attributes are grossly inaccurate or missing, that material is density, bulk screening and the values of eliminated by Fig. 13.2 Summary of the selection search for further information. strategy. The upper box describes screening, the lower one the Data sources 307 default. It is important, therefore, that the database be complete and be of high quality, meaning that the data in it can be trusted. This creates the need for data checking and estimation, tackled by methods described later in this chapter. The attribute-limits and index methods introduced in Chapters 5 and 11 are examples of the use of attributes to screen, based on design requirements. They provide an efficient way of reducing the vast number of materials in the materials kingdom to a small manageable subset for which further information can be sought. Data sources for screening (see also the Appendix, Section 13A) The traditional source of materials data is the handbook. The more courageous of them span all material classes, providing raw data for generic screening. More specialized handbooks and tradeassociation publications contain data suitable for second-level screening (Figure 13.2) as well as text and figures which help with further information. They are the primary sources, but they are clumsy to use because their data structure is not well suited to screening. Comparison of materials of different classes is possible but difficult because data are seldom reported in comparable formats; there is too much unstructured information, requiring the user to filter out what he needs; and the data tables are almost always full of holes. Electronic sources for generic screening can overcome these problems. If properly structured, they allow direct comparison across classes and selection by multiple criteria, and it is possible (using methods described in this chapter) to arrange that they have no holes. Screening, as we have seen, identifies a set of viable candidates. We now need their family history. That is the purpose of the `further information' step. 13.4 Further information: data structure and sources Data structure for further information The data requirements in the further information step differ greatly from those for screening (Figure 13.2, lower half). Here we seek additional details about the few candidates that have already been identified by the screening and ranking step. Typically, this is information about availability and pricing; exact values for key properties of the particular version of the material made by one manufacturer; case studies and examples of uses with cautions about unexpected difficulties (e.g. `liable to pitting corrosion in dilute acetic acid' or `material Y is preferred to material X for operation in industrial environments'). It is on this basis that the initial shortlist of candidates is narrowed down to one or a few prime choices. Sources of further information typically contain specialist information about a relatively narrow range of materials or processes. The information may be in the form of text, tables, graphs, photographs, computer programs, even video clips. The data can be large in quantity, detailed and precise in nature, but there is no requirement that it be comprehensive or that the attributes it contains be universal. The most common media are handbooks, trade association publications and manufacturers' leaflets and catalogues. Increasingly such information is becoming available in electronic form on CD-ROMs and on the Internet. Because the data is in `free' format, the search strategies differ completely from the numerical optimization procedures used for the screening step. The simplest approach is to use an index (as in a printed book), or a keyword list, or a computerized full text search, as implemented in many hyper-media systems. 308 Materials Selection in Mechanical Design Data sources for further information (see also the Appendix, Section 13A) By `further information' we mean data sources which, potentially, can contain everything that is known about a material or a process, with some sort of search procedure allowing the user to find and extract the particular details that he seeks. The handbooks and software that are the best sources for screening also contain further information, but because they are edited only infrequently, they are seldom up to date. Trade organizations, listed in the Appendix, Section 13A, do better, providing their members with frequent updates and reports. The larger materials suppliers (Dow Chemical, Ciba-Geigy, Inco, Corning Glass, etc.) publish design guides and compilations of case studies, and all suppliers have data sheets describing their products. There is an immense resource here. The problem is one of access. It is overcome by capturing the documents on CD-ROM, keyworded and with built-in `hot-links' to related information, addressed through a search-engine which allows full-text searching on topic strings (`aluminium bronze and corrosion and sea water', for example). Expert systems The main drawback of the simple, common-or-garden, database is the lack of qualification. Some data are valid under all conditions, others are properly used only under certain circumstances. The qualification can be as important as the data itself. Sometimes the question asked of the database is imprecise. The question: `What is the strength of a steel?' could be asking for yield strength or tensile strength or fatigue strength, or perhaps the least of all three. If the question were put to a materials expert as part of a larger consultation, he would know from the context which was wanted, would have a shrewd idea of the precision and range of validity of the value, and would warn of its limitations. An ordinary database can do none of this. Expert systems can. They have the potential to solve problems which require reasoning, provided it is based on rules that can be clearly defined: using a set of geometries to select the best welding technique, for instance; or using information about environmental conditions to choose the most corrosion-resistant alloy. It might be argued that a simple checklist or a table in a supplier's data sheet could do most of these things, but the expert system combines qualitative and quantitative information using its rules (the `expertise'), in a way which only someone with experience can. It does more than merely look up data; it qualifies it as well, allowing context-dependent selection of material or process. In the ponderous words of the British Computer Society: `Expert systems offer intelligent advice or take intelligent decisions by embodying in a computer the knowledge-based component of an expert's skill. They must, on demand, justify their line of reasoning in a manner intelligible to the user.' This context-dependent scheme for retrieving data sounds just what we want, but things are not so simple. An expert system is much more complex than a simple database: it is a major task to elicit the `knowledge' from the expert; it can require massive programming effort and computer power; and it is difficult to update. A full expert system for materials selection is decades away. Success has been achieved in specialized, highly focused applications: guidance in selecting adhesives from a limited set, in choosing a welding technique, or in designing against certain sorts of corrosion. It is only a question of time before more fully developed systems become available. They are something about which to keep informed. Data sources on the Internet And today we have the Internet. It contains an expanding spectrum of information sources. Some, particularly those on the World-Wide Web, contain information for materials, placed there by Data sources 309 standards organizations, trade associations, material suppliers, learned societies, universities, and individuals - some rational, some eccentric - who have something to say. There is no control over the contents of Web pages, so the nature of the information ranges from useful to baffling, and the quality from good to appalling. The Appendix, Section 13A includes a list of WWW sites which contain materials information, but the rate of change here is so rapid that it cannot be seen as comprehensive. 13.5 Ways of checking and estimating data The value of a database of material properties depends on its precision and its completeness - in short, on its quality. One way of maintaining or enhancing quality is to subject data to validating procedures. The property ranges and dimensionless correlations, described below, provide powerful tools for doing this. The same procedures fill a second function: that of providing estimates for missing data, essential when no direct measurements are available. Property ranges Each property of a given class of materials has a characteristic range. A convenient way of presenting the information is as a table in which a low ( L ) and a high ( H ) value are stored, identified by the material class. An example listing Young's modulus, E , for the generic material classes is shown in Table 13.2, in which EI, is the lower limit and EH the upper one. All properties have characteristic ranges like these. The range becomes narrower if the classes are made more restrictive. For purposes of checking and estimation, described in a moment, it is helpful to break down the class of metals into cast irons, steels, aluminium alloys, magnesium alloys, titanium alloys, copper alloys and so on. Similar subdivisions for polymers (thermoplastics, thermosets, elastomers) and for ceramics and glasses (engineering ceramics, whiteware, silicate glasses, minerals) increases resolution here also. Table 13.2 Ranges of Young's modulus E for broad material classes All solids Classes of solid Metals: ferrous Metals: non-ferrous Fine ceramics* Glasses Polymers: thermoplastic Polymers: thermosets Polymers: elastomers Polymeric foams Composites: metal-matrix Composites: polymer-matrix Woods: parallel to grain Woods: perpendicular to grain 0.00001 70 4.6 91 47 0.1 2.5 0.0005 0.0000 1 81 2.5 1.8 0.1 1000 220 570 1000 83 4.1 10 0.1 180 240 34 18 *Fine ceramics are dense, monolithic ceramics such as Sic, A1203, ZrO2, etc. 310 Materials Selection in Mechanical Design Correlations between material properties Materials which are stiff have high melting points. Solids with low densities have high specific heats. Metals with high thermal conductivities have high electrical conductivities. These rules-ofthumb describe correlations between two or more material properties which can be expressed more quantitatively as limits for the values of dimensionless property groups. They take the form CL < PIP; < C H (13.1) (13.2) or CL < PjP;Py < C H (or larger groupings) where P I , P2, P3 are material properties, n and m are simple powers (usually - 1, - 1/2, 1/2 or l), and CL and C N are dimensionless constants - the lower and upper limits between which the values of the property-group lies. The correlations exert tight constraints on the data, giving the `patterns' of property envelopes which appear on the material selection charts. An example is the relationship between expansion coefficient, a (units: K-I), and the melting point, T , (units: IC) or, for amorphous materials, the glass temperature T g : CL 5 f f T m 5 CH CL 5 f f T g 5 CH (13.3a) (13.3b) - a correlation with the form of equation (13.1). Values for the dimensionless limits C L and C H for this group are listed in Table 13.3 for a number of material classes. The values span a factor to 2 to 10 rather than the factor 10 to 100 of the property ranges. There are many such correlations. They form the basis of a hierarchical data checking and estimating scheme (one used in preparing the charts in this book), described next. Data checking The method is shown in Figure 13.3. Each datum is associated with a material class, or, at a higher level of checking, with a sub-class. It is first compared with the range limits L and H for that class and property. If it lies within the range limits, it is accepted; if it does not, it is flagged for checking. Table 13.3 Limits for the group aTm and aT' for broad material classes* Correlation* C L < aT, < CH ~~(x10-3) ~~(~10-3) All solids Classes of solid Metals: ferrous Metals: non-ferrous Fine ceramics* Glasses Polymers: thermoplastic Polymers: thermosets Polymers: elastomers Polymeric foams Composites: metal-matrix Composites: polymer-matrix Woods: parallel to grain Woods: perpendicular to grain 0. I 13 2 6 0.3 18 11 35 16 10 0.1 2 6 56 27 21 24 3 35 41 56 37 20 10 4 17 *For amorphous solids the melting point T , is replaced by the glass temperature T , Data sources 3 1 1 Input Data Assign Class Range Test Physical Limits Output Data Fig. 13.3 The checking procedure. Range checks catch gross errors in all properties. Checks using dimensionless groups can catch subtler errors in certain properties. The estimating procedure uses the same steps, but in reverse order. Why bother with such low-level stuff? It is because in compilations of material or process properties, the commonest error is that of a property value which is expressed in the wrong units, or is, for less obvious reasons, in error by one or more orders of magnitude (slipped decimal point, for instance). Range checks catch errors of this sort. If a demonstration of this is needed, it can be found by applying them to the contents of almost any standard reference data books; none among those we have tried has passed without errors. In the second stage, each of the dimensionless groups of properties like that of Table 13.3 is formed in turn, and compared with the range bracketed by the limits C L and C H . If the value lies within its correlation limits, it is accepted; if not, it is checked. Correlation checks are more discerning than range checks and catch subtler errors, allowing the quality of data to be enhanced further. Data estimation The relationships have another, equally useful, function. There remain gaps in our knowledge of material properties. The fracture toughness of many materials has not yet been measured, nor has the electric breakdown potential; even moduli are not always known. The absence of a datum for a material would falsely eliminate it from a selection which used that property, even though the material might be a viable candidate. This difficulty is avoided by using the correlation and range limits to estimate a value for the missing datum, adding a flag to alert the user that they are estimates. In estimating property values, the procedure used for checking is reversed: the dimensionless groups are used first because they are the more accurate. They can be surprisingly good. As an example, consider estimating the expansion coefficient, a, of polycarbonate from its glass temperature T,. Inverting equation (13.3) gives the estimation rule: CL - < a < - CH - T, T, (13.4) 312 Materials Selection in Mechanical Design Inserting values of CL and C H from Table 13.3, and the value T , = 420K for a particular sample of polycarbonate gives the mean estimate Z = 63 x K-' (13.5) The reported value for polycarbonate is (Y = 54 - 62 x K-' The estimate is within 9% of the mean of the measured values, perfectly adequate for screening purposes. That it is an estimate must not be forgotten, however: if thermal expansion is crucial to the design, better data or direct measurements are essential. Only when the potential of the correlations is exhausted are the property ranges invoked. They provide a crude first estimate of the value of the missing property, far less accurate than that of the correlations, but still useful in providing guide-values for screening. 13.6 Summary and conclusions The systematic way to select materials or processes (or anything else, for that matter) is this. (a) Identify the taxonomy of the kingdom from which the selection is to be made; its classes, subclasses and members. (b) Identify the attributes of the members, remembering that they should be universal and discriminating within this kingdom; resolution is increased by defining second-level `sub-kingdoms' allowing an expanded set of attributes, universal within the sub-kingdom. (c) Assess the quality and completeness of the data sources for the attributes; both can be increased by techniques of checking and estimation described in the previous section. (d) Reduce the large population of the kingdom to a shortlist of potential candidates by screening on attributes in the first and second-level kingdoms. (e) Identify sources of further information for the candidates: texts, design guides, case studies, suppliers' data sheets or (better) searchable electronic versions of these, including the Internet. (0Compare full character profiles of the candidates with requirements of the design, taking into account local constraints (preferences, experience, compatibility with other activities, etc.). To do all this YOU need to know where to find data, and you need it at three levels of breadth and precision. Conceptual design requires a broad survey at the low accuracy offered by the selection charts of Chapters 4 and 11, and by other broad-spectrum data tabulations. Embodiment design needs more detail and precision, of the kind found in the handbooks and computer databases listed in the Appendix, Section 13A. The final, detailed, phase of design relies on the yet more precise (and attributable) information contained in material suppliers' data sheets. The falling cost and rising speed of computing makes databases increasingly attractive. They allow fast retrieval of data for a material or a process, and the selection of the subset of them which have attributes within a specified range. Commercially available databases already help enormously in selection, and are growing every year. Some of those currently available are reviewed in the Appendix, Section 13A. Expert systems lurk somewhere in the future. They combine a database with a set of rules for reasoning to permit simple, logical deductions to be made by the computer itself, allowing it to Next Page Data sources 313 retrieve relevant information which the operator did not know or forgot to ask for. They combine the data of a handbook with some of the expertise of a materials consultant. They are difficult to create and demand much computer power, but the selection process lends itself well to expert-systems programming; they will, sooner or later, be with us. Don't leave this chapter without at least glancing at the compilation of data sources in the next section. It is probably the most useful bit. 13.7 Further reading Ashby, M.F. (1998) `Checks and estimates for material properties', Cambridge University Engineering Department, Proc. Roy. Soc. A 454, 1301-1321. Bassetti, D., Brechet, Y. and Ashby, M.F. (1998) `Estimates for material properties: the method of multiple correlations', Proc. Roy. Soc. A 454, 1323- 1336. Cebon, D. and Ashby, M.F. (1992) `Computer-aided selection for mechanical design', Metals and Materials, January, 25-30. Cebon, D. and Ashby, M.F. (1996) `Electronic material information systems', I. Mech. E. Conference on Electronic Delivery of Design Information, October, 1996, London, UK. CMS (Cambridge Materials Selector) (1992), Granta Design, Trumpington Mews, 40B High Street, Trumpington, Cambridge CB2 2LS, UK. CPS (Cambridge Process Selector) ( 1 998), Granta Design, Trumpington Mews, 40B High Street, Trumpington, Cambridge CB2 2LS, UK. The Copper Development Association (1994) Megabytes on Coppers, Orchard House, Mutton Lane, Potters Bar, Herts EN6 3AP, UK; and Granta Design Limited, 20 Trumpington St., Cambridge CB2 IPZ, UK, 1994. 13A Appendix: Data sources for material and process attributes 13A.l Introduction Background This appendix tells you where to look to find material property data. The sources, broadly speaking, are of three sorts: hard copy, software and the Internet. The hard copy documents listed below will be found in most engineering libraries. The computer databases are harder to find: the supplier is listed, with address and contact number, as well as the hardware required to run the database. Internet sites are easy to find but can be frustrating to use. Section 13A.2 lists sources of information about database structure and functionality. Sections 13A.3 catalogues hard-copy data sources for various classes of material, with a brief commentary where appropriate. Selection of material is often linked to that of processing; Section 13A.4 provides a starting point for reading on processes. Section 13A.5 gives information about the rapidly growing portfolio of software for materials and process data and information. Section 13A.6 - the last - lists World-wide Web sites on which materials information can be found. 13A.2 General references on databases Waterman, N.A., Waterman, M. and Poole, M.E. (1992) `Computer based materials selection systems', Metals and Materials 8. 19-24. ...
View Full Document

This note was uploaded on 04/26/2008 for the course M&AE 212 taught by Professor Miller during the Spring '07 term at Cornell University (Engineering School).

Page1 / 11

Chapter13 - Data sources 13.1 Introduction and synopsis The...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online