Materials Science and Engineering Handbook Part 7 ppt

References 1. Committee on Application of Expert Systems to Materials Selections During Structural Design, Computer- Aided Materials Selection During Structural Design, Report No. 467, National Materials A dvisory Board, National Academy Press, Washington, DC, 1995 2. Software Showcase, Adv. Mater. Process., March 1997 Computer-Aided Materials Selection Volker Weiss, Syracuse University Expert Systems: General Description Expert systems are usually integrations of databases and knowledge bases using search and logic deduction algorithms. Once implemented, the distinction between databases and knowledge bases becomes blurred, because usually they are both treated simply as assertions, for example, "it is the case that the yield strength of 4340 steel, tempered at 400 °F, is 270 ksi." The integration of these information bases is accomplished by inference engines, typically embedded in expert system shells, or by computer languages based on some form of logic search rule (e.g., use of unification and resolution algorithms in Prolog). For all the databases and knowledge bases to be addressable by the inference engine, they have to be in an appropriate format, which is often not the case, though the situation is improving. Clearly, a fully functional computer aided materials selection system (CAMSS) will assume the character of an expert system. It has also been suggested that the CAMSS will ultimately involve not only a materials selection system, but also will be coupled, perhaps with feedback loops, to the CAD/CAM system, thus forming an integrated set of computer systems that serve the design team. Expert systems first appeared around 1965; early systems included DENDRAL, MACSYMA, MYCIN, EMYCIN and Prospector (Ref 3). They were generally designed and programmed by one or several "knowledge engineers," who obtained the information from "domain experts" and structured a user interface that was intended to be friendly to the end user, presumably not an expert in artificial intelligence programming. The classical structure of a typical expert system is shown in Fig. 2. Prolog and LISP, and, more recently, object-oriented implementations through OOPS, Smalltalk and Java, are the languages of choice for expert system building. EMYCIN, CLIPS, KEE, ART, and NEXPERT are so-called "expert system shells," that is, empty reasoning systems to which a user can add data and rules to build an expert system. MYCIN was developed for medical diagnostic purposes. EMYCIN is "empty" MYCIN, a shell that is suitable for diagnostic reasoning (such as in failure analysis). Fig. 2 Architecture of a typical expert system. Source: Ref 4 Computer-assisted materials selection, through expert or some other advanced systems, will be primarily of use for the level III and level IV design/materials selection phases (see the article "Overview of Materials Selection" in this Volume). At these stages, a design team may not be aware of, for example, all materials available in a given class (e.g., heat- resistant alloys) that also meet certain specific room-temperature requirements such as minimum fracture toughness and corrosion resistance to sea water after extended high-temperature service. References cited in this section 3. P.H. Winston and K.A. Pendergast, Ed., The AI Business, MIT Press, 1984 4. G.F. Luger and W.A. Stubblefield, Artificial Intelligence and the Design of Expert Systems, Benjamin/Cummings Publishing Co., Redwood, CA, 1989, p 294 Computer-Aided Materials Selection Volker Weiss, Syracuse University Quantitative Selection Systems Any useful quantitatively based expert system for materials selection depends heavily on available data and design rules. The databases must be accessible to the system structure (e.g., languages like Prolog, LISP, or a shell). The data requirement is primarily for well-analyzed and documented data of candidate materials. The application-specific knowledge base will contain the design rules, such as appropriate stress analyses, stress concentration factors, fracture mechanics formulas, creep and fatigue relationships, and so forth. For complicated geometries and critical applications, the application-specific knowledge base might be hybridized to a finite element system to perform the required stress analysis for the geometry under consideration. The system must have access to, or incorporate, materials properties databases to be used within the reasoning structure of the expert system. For example, a simple expert system for materials selection for a tension bar of minimum weight and a diameter not to exceed a given value and to carry a given load (Ref 5) could have mechanical properties data of materials excerpted or downloaded from a database so that yield strength and density values of a sufficient number of materials are available. Rules such as: Load-carrying capacity = Yield strength × Area × Safety factor and Weight per unit length = Area × Density and Area = (Diameter) 2 × /4 need to be invoked in the search through the database to select the materials that meet the requirement. This list can then be sorted in terms of weight per unit length, or yield strength/density, as desired by the designer. Inspection criteria, for example, for cracks, could be added to exclude materials for which the critical crack length is below the inspection limit; this would require fracture toughness values be included for all candidate materials. In addition, exclusion rules could be formulated that remove all materials from consideration that would experience corrosion damage above a specified limit in the service environment; for this purpose, corrosion data for the anticipated service environment would need to be available from the database. In addition, rules for other factors of concern (e.g. cost and appearance, "lessons-learned" rules for similar designs, and "manufacturability" rules) may be formulated and invoked. The result of a query using LOG-LISP (Ref 6) and a database of several hundred materials is shown in Table 1. Table 1 Results of LOG-LISP materials selection query The conditions wer e: (1) must safely support a load of 200 ksi, (2) must have a safety factor of 1.75, (3) maximum diameter of 1.5 in., and (4) maximum weight per unit length of 0.5 lb/in. MAT COND YS DIAM W/L ST 4340 QT-400F 262 1.35 0.40 ST 4340 QT-600F 230 1.39 0.43 ST 4130 WQT-400F 212 1.45 0.46 ST 4130 WQT-600F 200 1.49 0.49 Source: Ref 6 (a) MAT, mater ial; ST, steel; COND, condition; QT, quenched and tempered; WQT, water quenched and tempered; YS, yield strength (ksi); DIAM, diameter (in.); W/L, weight per unit length (lb/in.). This simple example is typical of a wide variety of much more elaborate materials selection problems. Database, knowledge base, and finite element data interaction would come into play for load-carrying parts having complicated geometries. Fatigue and fatigue crack growth knowledge bases and databases will be required for applications with load fluctuations and high-temperature creep data for elevated-temperature applications. For many of these applications, closed-form design formulas will not be available; therefore, the materials selection system will have to be connected to other computer programs (e.g., finite element programs). Information about the database quality is necessary to decide on the applicability of the materials selection system for preliminary or final design stages. For the former case, typical data, or supplier data, may suffice. For the latter, well- documented and preferably statistically analyzed data of the type and quality characteristic of MIL-HDBK-5 (Ref 7), for example are required. If data that include statistical information (e.g., standard deviation) are used, the knowledge base will need to have "rules" that deal with that information to determine average and (statistically defined) minimum load- carrying capacities for the structure under consideration. Iterations between design system and materials selection system should be planned for so that the designer can see if design changes also prompt material selection and other (e.g., cost) changes. It should be noted that the rules that manipulate well-defined statistical data are firm, deterministic rules that should not be confused with the application of "fuzzy logic" or "fuzzy numbers," which are addressed in the section "Qualitative and Experiential Selection Systems" in this article. Another application for a quantitative materials selection system is materials substitution. The query to be answered would be of the type: "find materials that have the same or higher tensile and yield strength as the one currently used (e.g., SAE 4340 steel, quenched and tempered at 800 °F) but contain less chromium and nickel." In addition to calling on mechanical property databases of alloys, a (nominal) chemical composition database must be available to the "inference engine." Table 2 shows the results from the same LOG-LISP prototype expert system used for the tension member design query (Table 1). Table 2 Results of LOG-LISP materials substitution query Material and condition to be replaced: ST 4340, QT-800F. Critical properties to be matched or exceeded: TS, YS. Critical all oying elements to be reduced: chromium, nickel MAT COND Cr Ni TS YS ST 4340 QT-800F 0.90 2.0 213 198 ST 8630 QT-400F 0.60 0.70 238 218 ST 8630 QT-600F 0.60 0.70 215 202 ST 8640 QT-400F 0.60 0.70 270 242 ST 8640 QT-600F 0.60 0.70 240 220 ST 8740 QT-400F 0.60 0.70 290 240 ST 8740 QT-600F 0.60 0.70 240 225 Source: Ref 6 (a) MAT, material; ST, steel; COND, condition; QT, quenched and tempered; TS, tensile strength (ksi); YS, yield strength (ksi). The prototype for a materials selection system to meet the special needs of the aerospace industry, the Intelligent Knowledge System for Selection of Materials for Critical Aerospace Applications (IKSMAT), was developed in the late 1980s (Ref 8). The plans were for an elaborate, mostly quantitative system with a front-end program for data entry and a diagnostic program to ensure that all of the important properties and characteristics of all the candidate materials are considered. The model for such a materials selection system is illustrated in Fig. 3. This model was implemented using the system architecture illustrated in Fig. 4. A "master" database facility, containing well documented and evaluated data, mostly of the type found in MIL-HDBK-5 (Ref 7), provided the input through MESSENGER mainframe software. The user interface was planned not only to allow the user to work within the existing knowledge base, but also to add data, rules, new criteria or other information, and to redefine priorities. IKSMAT was turned over to the U.S. Air Force in 1993, and it may have been discontinued due to the expense of maintaining it. Fig. 3 Model of the IKSMAT system for materials selection in the aerospace industries. MPD, materials properties database. Source: Ref 8 Fig. 4 Specific system architecture for the prototype IKSMAT materials selection system. Source: Ref 8 References cited in this section 5. V. Weiss and K.J. Green, Expert Systems for Materials Selection, Artificial Intelligence in Minerals and Materials Technology, U.S. Department of the Interior, Bureau of Min es, Tuscaloosa Research Center, Oct 1987 6. J.A. Robinson and E.E. Sibert, "The LOGLISP User's Manual," Technical Report, School of Computer and Information Sciences, Syracuse University, 1981 7. Metallic Materials and Elements for Aerospace Vehicle Structures, MIL-HDBK-5E, Military Standardization Handbook, MIL-HDBK-5, Department of Defense, 1986 8. J.G. Kaufman, "An Intelligent Knowledge System for Selection of Materials for Critical Aerospace Applications (IKSMAT)" Final Technical Report, U.S. Air Force Contract F33615-87-C- 5305, Department of Defense, 1988 Computer-Aided Materials Selection Volker Weiss, Syracuse University Qualitative and Experiential Selection Systems Qualitative and experiential materials selection systems are needed for applications that involve considerable uncertainty. In such systems, the computer code seeks to assist or even substitute for design experts. Typical examples are materials selection requirements for service in corrosive environments, for cumulative fatigue damage, and for service under thermal and load cycling. Uncertainties in such applications involve, for example, the effects of combinations of corrosive agents, effects of stress amplitude sequence in cumulative fatigue, and effects of in-phase and out-of-phase load and temperature cycling. Systems that address these problems must be able to deal with approximations, and "fuzzy numbers," and they must compare and manipulate them and present the results of such manipulations with qualifiers, which distinguish them from results obtained from "hard-number" quantitative systems. The system must allow data to be expressed as minima, maxima, ranges, and unknowns, in addition to hard values. Different systems may, based on the underlying "fuzzy-logic" reasoning structure, come to different conclusions. For example, if the corrosion rate for material 1 in polluted harbor water is between 0.04 and 0.38 mm/yr, and that for material 2 between 0.18 and 0.28 mm/yr, which material is more corrosion resistant? Selection based on a "worst-case scenario" would lead to the selection of material 2, while a selection based on average rates would lead to the selection of material 1. For such systems it will be necessary to include the "desired scenario" in the query. Example: Heat Exchanger Materials Selection System. A selection system for materials subjected to service in corrosive environments is typical for this class. Specifically, Syracuse University developed a prototype materials selection system for tube (heat exchanger) applications where corrosion and erosion results from a combined effect of water chemistry and flow rate. The prototype system was implemented in Prolog on a personal computer. The knowledge base for the system consists of six components: • A corrosion property knowledge base for candidate materials • A fuzzy number knowledge base for reasoning with uncertain data • An experiential knowledge base stating conditions for excluding materials • A location knowledge base containing water analyses and derivable characteris tics (e.g., Langelier and Ryzner indices) • An installation knowledge base providing information on experience of prior installations at the same site • A control knowledge base containing the rules that govern the selection process The structure of this prototype heat exchanger materials selection system (HEMS) is shown in Fig. 5. The output of the system, once a location is specified, is the estimated lifetime of heat exchanger tubes, that is, time to first leak. This time is then expressed in "design life fraction" using a company-specified design life, for example, 20 years. The 12 candidate materials are then ranked in order of decreasing life expectancy. A query for tube materials for a hypothetical location (with extremely polluted, sulfide-containing harbor water) resulted in the ranking given in Table 3. It should be noted that there is considerable uncertainty about corrosion rates in the database, especially the effects of combinations of corrosive media. However, a "rule correction module" was incorporated in the knowledge base, which enabled correction of corrosion rate data on the basis of field reports. For the given location, the materials selection system indicates that only titanium tubes are suitable. For other locations, several materials may be adequate, perhaps ranging from design life fractions of 2 to 6. In such a case, the customer must weigh the generally increased cost with increasing design life of the materials, for example, copper, Cu-10Ni, and Cu-30Ni. Such cost information can readily be added to the knowledge base and incorporated in the output. Table 3 Results of heat exchanger materials selection system query Query: tube materials suitable for service in extremely polluted, sulfide-containing harbor water Rank Material Design life fraction 1 Titanium >5.00 2 Type 316 stainless steel 0.35 3 Aluminum bronze 0.26 4 Type 304 stainless steel 0.19 5 Aluminum brass 0.18 6 AISI/SAE 1010 steel 0.13 7 Admiralty brass 0.13 8 Red brass 0.12 9 Copper 0.11 10 Cu-10Ni 0.10 11 Cu-30Ni 0.08 12 Monel alloy 0.07 Fig. 5 Structure envisioned for the prototype heat exchanger materials selection (HEMS). Source: Ref 5 The development of a user-friendly interface is of great importance when designing expert systems. For HEMS, in the prototype version, developed in Prolog, the user-friendliness was achieved by employing a hypertext link with "buttons" on keywords that activate the appropriate Prolog queries. The following is a short excerpt from the hypertext version of the prototype HEMS system. The words in bold type are "buttons" that activate actions, either display of explanatory text or initiation of a query process, usually through query-specific windows: The HEMS Expert System provides help in the selection of tube materials for centrifugal chillers, reciprocating chillers, and other heat exchangers. The system consists of two major components, an information base and a material selection assistant expert system . The two components can be used independently. In most instances the user will chose the interactive mode, making use of the local water chemistry database, at least for a start, and of the water chemistry calculation program. The materials ranking progra m, partially implemented here as a prototype system, allows you to select tube materials for typical water chemistries. If the user selects material selection assistant or materials ranking, windows appear that ask for location, flow rate, and updates on water chemistry and initiates the appropriate queries that result in a ranking of the type given in Table 3. Selecting other buttons provides either information about the subject or connects to databases that provide information such as water chemistries by location and corrosion rates for the candidate materials in various environments. The system was also linked to already existing FORTRAN programs that calculate Langelier and Ryzner indices from the water chemistry data. Access was provided on two levels, general user and materials specialist. The latter group also had the capability of modifying the system, adding data and programs, and modifying corrosion rate calculations on the basis of new field data. For the materials specialist, who had some familiarity with Prolog programming, ad hoc querying of the knowledge base of the system, beyond the standard prescribed query actions offered through the windows in the general system, could readily be accomplished. To the author's knowledge, the prototype HEMS system was not used commercially. Reference cited in this section 5. V. Weiss and K.J. Green, Expert Systems for Materials Selection, Artificial Intelligence in Minerals and Materials Technology, U.S. Department of the Interior, Bureau of Mines, Tuscaloosa Research Center, Oct 1987 Computer-Aided Materials Selection Volker Weiss, Syracuse University Object-Oriented Systems Object-oriented computer languages, such as C++, Java, and so on, gained in popularity in the 1990s, primarily because of the way these languages allow information to be handled in a hierarchical way. Figure 6 shows how the object Al 2024-T3 is treated as a special case of various classes and subclasses. Each instance of an object or class can contain data and rules, or methods. Information from general classes is "inherited" to less-general classes and finally to the object. This hierarchy of the knowledge base increases its range of applicability and compactness as compared to a rule-based system operating on relational databases. Design information can be stored in a similar fashion as illustrated in Fig. 7, where the hierarchy for a specific "rod" as a subclass of a solid is presented. Fig. 6 Materials hierarchy in an object-oriented system. Source: Ref 9 Fig. 7 Geometrical solid hierarchy for a specific rod in an object-oriented system. Source: Ref 9 Krishnamurthy and Smith (Ref 10) have developed an experimental quantitative problem solver (QPS) that is applicable to materials selection problems. Instead of asking for materials that satisfy a range of property requirements, one initiates materials selection with QPS by specifying the characteristics of the item being designed and allowing the system to specify the material query. With increasing computing speed, object-oriented programming has become increasingly practical, even for large systems. Many of the AI languages and expert system shells are now available in object-oriented versions. Interesting new developments are the languages or programs that can interface with the Internet technologies, including the World Wide Web (WWW); two examples are Java and the EXSYS Web runtime engine (EXSYS WREN) (Ref 11). Java is an object-oriented programming language that allows the writing of small programs, called "applets," that can be embedded into a hypertext markup language (HTML) Web page. The HTML can be used to create hypertext documents that can be moved from one platform to another. Embedded applets execute on the local client workstation. Thus it is possible to use Java, or some other similarly capable program, to interact with information on the WWW to provide answers to queries subject to specified rules and constraints. EXSYS WREN (Ref 11) is an expert system development tool that, in connection with rules that can be produced by the compatible EXSYS Rule Book, acts on WWW information to respond to questions. With Java, the rules and other aspects of the applicable knowledge base have to be embedded in the applets. A simple applet may retrieve data on yield strength, modulus, and density from the Web or a database to compute the deflection of a beam of specified dimensions under its own weight. Newer versions of Java offer a new interface and greater database connectivity. Agents. "Agents" and "agencies," (i.e. collections of "agents") are other developments worthy of consideration for use in constructing systems for materials selection. Agents are "mini expert systems" that have specific limited functions, for example, search-specified databases, or WWW files, for data on a specific subject (such as yield strength and plane strain fracture toughness) to enhance the database of the larger expert system. Agents are usually object-oriented language applications. Data mining programs also may be useful when building knowledge bases for materials selection expert systems. These programs some of them using object-oriented languages, some neural networks (Ref 13) are an extended development of agents. They not only import data of interest, but also are capable of performing statistical and other operations on the data that lets the user make a judgment about their quality or even lead to predictions (perhaps via fuzzy numbers) of outcomes, for example, product demand. They generally can be applied to databases that support open database connectivity (ODBC). Typical commercially available data mining programs are "Alice" (from Isoft), "Profiler" (from Attar Software UK), "Model Quest Miner" (from AbTech Corporation), and "Predict" (from NeuralWare Inc.). References cited in this section 9. F.J. Smith, M.V. Krishnamurthy, S.R. Tripathy, and P. Sage, An Intelligent Object- Oriented Database for Materials Information, Computerization and Networking of Materials Databases, C.P. Sturrock and E.F. Begley, Ed., STP 1257, ASTM, 1995 10. M.V. Krishnamurthy and F.J. Smith, Integration of Scientific Data and Formulas in Object- Oriented Knowledge Based Systems, Knowl. Based Syst. J., 1995 11. D. Huntington, Web-Based AI-Expert Systems on the WWW, PC AI, March 1997, p 20 13. H.M.G. Smets and W.F.L. Bogaerts, Neural Networks for Materi als Data Analysis: Development Guidelines, Computerization and Networking for Materials Databases, C.P. Sturrock and E.F. Begley, Ed., STP 1257, ASTM, 1995 Computer-Aided Materials Selection Volker Weiss, Syracuse University Current Status and Outlook At present, many components exist that can make up viable computer-based materials selection systems, though no integrated system such as envisioned by the 1995 National Materials Advisory Board study (Ref 1) has been implemented. Several user-specific, smaller-scale, special-purpose materials selector systems have been developed and are in use at various companies and institutions. Rockwell International developed a "Materials and Process Design Advisor" based on EXSYS. It reportedly consists of expert systems for corrosion protection, adhesive selection, encapsulant selection, conformal coating selection, heat treatment of metallic materials, and selecting soldering processes. The systems are made available to the Rockwell design engineers. Granta Design Limited is marketing the "Cambridge Materials Selector" (CMS 2.0), which was developed by the Cambridge University Engineering Department. A generic database, containing representatives of virtually all groups of engineering materials, forms the core of the system. After identifying a group of candidate materials, the user can refine the selection with the help of additional databases in the materials group of the candidate materials. The type of use that is made of these and other emerging systems most critically depends on the quality of data available to the rule base (including design features, environment, etc.) and the inference engine. Material selection for final design of critical components will certainly only be possible if the database consulted consists of well-documented and statistically evaluated data. For materials selection for preliminary or conceptual design stages, databases containing "typical" properties data or information from the material supplier will suffice. Expert systems for materials selection may even become useful for cases where no commercially available material can satisfy all the performance requirements specified. In such cases, it might be possible to obtain guidance toward potential materials classes by relaxing the requirements until available materials appear as possible candidates. This may lead to identification of materials in research or under development that might eventually help solve the design problem. Ultimately, it is hoped that computing programs will emerge that allow atomic modeling of new materials and their properties. A great number of combinations could be explored inexpensively and in a relatively short time. Leading [...]... and K.J Green, Expert Systems for Materials Selection, Artificial Intelligence in Minerals and Materials Technology, U.S Department of the Interior, Bureau of Mines, Tuscaloosa Research Center, Oct 19 87 6 J.A Robinson and E.E Sibert, "The LOGLISP User's Manual," Technical Report, School of Computer and Information Sciences, Syracuse University, 1981 7 Metallic Materials and Elements for Aerospace Vehicle... and Networking of Materials Databases, C.P Sturrock and E.F Begley, Ed., STP 12 57, ASTM, 1995 10 M.V Krishnamurthy and F.J Smith, Integration of Scientific Data and Formulas in Object-Oriented Knowledge Based Systems, Knowl Based Syst J., 1995 11 D Huntington, Web-Based AI-Expert Systems on the WWW, PC AI, March 19 97, p 20 12 D.W Rasmus, Agents A New Agency, PC AI, March 19 97, p 37 13 H.M.G Smets and. .. H.M.G Smets and W.F.L Bogaerts, Neural Networks for Materials Data Analysis: Development Guidelines, Computerization and Networking for Materials Databases, C.P Sturrock and E.F Begley, Ed., STP 12 57, ASTM, 1995 14 E.F Codd, A Relational Model for Large Shared Databanks, Commun ACM, Vol 13 (No 6), 1 970 , p 377 Value Analysis in Materials Selection and Design Theodore C Fowler, Fowler & Whitestone Introduction... cost, and its potential to produce other problems (which can happen) One option may be the selection of a different material References cited in this section 7 G.F Vander Voort, Conducting the Failure Examination, Met Eng Q., Vol 15 (No 2), May 1 975 , p 31-36 8 Failure Analysis and Prevention, Vol 11, ASM Handbook (formerly Metals Handbook, 9th ed.), American Society for Metals, 1986 9 V.J Colangelo and. .. & Sons, 1 974 10 R.D Barer and B.F Peters, Why Metals Fail, Gordon and Breach Science Publishers, 1 970 11 D.J Wulpi, Understanding How Components Fail, American Society for Metals, 1985 12 C.R Brooks and A Choudhury, Metallurgical Failure Analysis, McGraw-Hill, 1993 Use of Failure Analysis in Materials Selection George F Vander Voort, Buehler Ltd Methods for Analyzing Failures to Improve Materials Selection... Military Standardization Handbook, MIL-HDBK-5, Department of Defense, 1986 8 J.G Kaufman, "An Intelligent Knowledge System for Selection of Materials for Critical Aerospace Applications (IKSMAT)" Final Technical Report, U.S Air Force Contract F33615- 87- C-5305, Department of Defense, 1988 9 F.J Smith, M.V Krishnamurthy, S.R Tripathy, and P Sage, An Intelligent Object-Oriented Database for Materials Information,... better and sales of your part are dropping Or, operating conditions may have changed and your part is no longer performing acceptably Of course, it is possible that parts have failed and the failure study recommended upgrading the material being used, perhaps also with other suggestions regarding the design and manufacturing processes Use of Failure Analysis in Materials Selection George F Vander Voort,... Design, Report No 4 67, National Materials Advisory Board, National Academy Press, Washington, DC, 1995 14 E.F Codd, A Relational Model for Large Shared Databanks, Commun ACM, Vol 13 (No 6), 1 970 , p 377 Computer-Aided Materials Selection Volker Weiss, Syracuse University References 1 Committee on Application of Expert Systems to Materials Selections During Structural Design, ComputerAided Materials Selection... and strain in structures and to predict fatigue life This information is useful when selecting materials, but it is only part of the overall picture Materials properties databases are another useful tool in considering alternate materials, although mechanical properties are only part of the overall picture The failure analysis methodology generally centers on cause and effect and ignores prevention Failure... in Materials Selection and Design Theodore C Fowler, Fowler & Whitestone Selected References • • • • • • • A.J Dell'Isola, Value Engineering in the Construction Industry, Construction Publishing, 1 974 C Fallon, Value Analysis, Prentice Hall, 1980 T.C Fowler, Value Analysis in Design, Van Nostrand Reinhold, 1990 L.D Miles, Techniques of Value Analysis and Value Engineering, 3rd ed., McGraw-Hill, 1 972 . 0 .70 238 218 ST 8630 QT-600F 0.60 0 .70 215 202 ST 8640 QT-400F 0.60 0 .70 270 242 ST 8640 QT-600F 0.60 0 .70 240 220 ST 874 0 . 19 87 6. J.A. Robinson and E.E. Sibert, "The LOGLISP User's Manual," Technical Report, School of Computer and Information Sciences, Syracuse University, 1981 7. Metallic Materials. Database for Materials Information, Computerization and Networking of Materials Databases, C.P. Sturrock and E.F. Begley, Ed., STP 12 57, ASTM, 1995 10. M.V. Krishnamurthy and F.J. Smith,