Automatic identification of critical design relationships

As part of the preliminary design, the designer must evaluate the benefits of many alternative design configurations, each of which may depend on a large number of design variables. Even after many alternatives are discarded using qualitative or experiential reasoning, the designer may have to further restrict his alternatives by performing a preliminary quantitative evaluation. Even very simplified design equations may be puzzling to an inexperienced designer in that a change in any one of the design variables will often influence many functional requirements. As a result, it is difficult to evaluate the merits of the design without more detailed analysis. Experienced designers, on the other hand, are often able to identify important relationships which govern or limit design performance. Identifying important relations, such as a critical ratio or difference, not only contributes to convenience and expediency, but preserves the physical reasoning associated with the design activity and helps focus the designer's creativity toward the governing or limiting aspects of the proposed solutions. A computer based system has been developed to assist die designer in identifying important design relationships. The system operates on a set of simplified design equations to produce sets of transformed equations in terms of some alternative design variables. The alternative variables are chosen for physical significance and for correspondence to functional behavior. The transformed sets of equations can be thought of as providing an alternative view of the design configuration. They are expected to enhance the physical insight of the designer, to help in identifying governing relationships among design variables and function, and to assist the designer in evaluating perfonnance limitations of alternative design configurations. University Libraries Inhaltsangabe W&hrend dcs Vorcntwurfstadimus muss dcr Konstrukteur die Vor-und Nachtcilc verschiedener Entwttfc bewcrten, jeder dicscr Entwtirfe hangt von vielen Eutwtirfsvariablcn ab. Sogar nachdcm viclc Alternativen aus qualitativen oder funktionalen Grfmden ausgcschicdcn sind, kann dcr Designer die Altcrnatiworschl&gc durch friihzcirigc zahlenmMssige Bewertung weiter beschneiden. Selbst stark vereinfachte Entwurfsrcgcln kdnnen einen unerfahrenen Designer verwiiren, dadurch dass eine einzelne Entwurfsvariablc viele funktionale Bedingungen beeinflusst Daher ist es schwierig die Von&ge eines Entwurfs ohne detaillierte Analyse abzusch&tzen. Andererseits k&nnen erfahrene Konstrukteure wichtigc Rand bedingungen erkenncn, die eine Entwurfsdurchfuhmng unterstutzen oder einschrflnken. Die wichtigen physikalischen Beziehungen, wie Gnfesenvcrhaltnissc pder physikalische Messgrttssen zu erkenncn trMgt nicht nur zur Vereinfachung und ZweckmAssigkeit bei, sondern bewahrt auch die physikalischen Beziehungen, die mit dem Entwurf vcrbundcn sind und hilft dcm Konstruktour seine Encrgic auf die wichtigen Faktorcn dcr angestrcbten Losung zu konzentricrcn. Es wurdc cin itchnergestutztes System entwickelt, dass den Konstrukteur bei dcr Eikennung von wichtigen Entwurfsbeziehungen untersttitzt. Das System arbeitet mit einem Satz vereinfachter Entwurfsrcgcln, die funktionale Abhangigkcit zu Entwurfsvariablen in Bczichung setzen, dies sind sowohl die physikalisch wichtigen als auch die direkt in Zusammenhang zu funktionalen Ausfuhrung stehenden Variables Das Ergebnis ist cin Satz von Entwurfsrcgcln. Jeder Satz ist im wesentlichen eine anderc Art einen Enwurf zu bctrachtcn. Man erwartet, dass die Regeln das physikalische VcrstSkndnis dcs Konstniktcurs vergrriisseni, indem sic wichtigc Beziehungen zwischen Entwurfsvariablen und Funktion erkenncn lassen, und den Konstrukteur in dcr Beurteilung dcr Durchfuhrbarkcit altcrnativcr Entwdlrfe unterstfltzt Introduction The objective of engineering design is the specification of a process or a product More specifically, the task is to transform a set of functional requirements for a product into a physical description of the product, including geometric, component and material specifications* The way in which the designer completes this task is a subject of considerable interest In the early stages of a design, the designer faces the task of evaluating the relative benefits and liabilities of many alternative configurations. The performance of each configuration typically dqpends cm a large number of design variables, which are not yet specified After evaluating die many alternatives using qualitative reasoning and experiential judgements, the designer eliminates all but the best alternatives from future consideration. Preliminary quantitative evaluations may then be used to further restrict the alternatives. The ease with which quantitative evaluations are made depends, in large part, cm the complexity of the design equations involved Even very simplified design equations may be puzzling to an inexperienced designer because changing the value of one of the design variables may influence many of the functional requirements. As a result, detailed analytical and optimization methods are often applied to the remaining design alternatives. The results of the analysis are used to judge the merits of the design configurations. Experienced designers, on the other hand, often shortcut the detailed analytical work by recognizing important relationships which govern the performance of the design configuration. This is accomplished by identifying important relations among functions and design variables, such as a critical ratio, a nondimcnsional parameter, or a simple difference; e.g. the column height to diameter ratio in structures, the Reynold's number in fluid mechanics, or the velocity difference across a fluid coupling. This achieves convenience and expediency in quantitative evaluations and enhances the physical reasoning associated with the design activity to better enable the designer to focus his creativity on the essential deficiencies of the proposed configuration. The discovery of such critical relationships among parameters has been made on an ad hoc basis by experienced designers and engineers. Although certain nondimensional parameters are well known and methods exist for identifying such parameters, there are not, in general, strategies which assist the designer in identifying physically significant relationships which dominate die behavior of a particular design configuration. A computer based system to aid in the identification of critical design relationships would be of value to inexperienced designers in determining better ways of looking at proposed design configurations. It would also be valuable to experienced designers and engineering analysts in determining alternative variables which are better suited for analytical manipulations, optimization, or numerical methods. The results are expected to offer insight into the relationships between design decisions and product characteristics, highlight die underlying physics, and provide increased understanding of the meaning of terms in the governing design equations. In addition to increased understanding and insight, new forms of the governing equations are expected to provide increased efficiency for numerical testing and computations. Computer Aids in Mechanical Design The current genre of mechanical CAD systems have impacted the drafting room and the use of computer based analytical methods, notably finite element programs, but have had a negligible effect on most other aspects of the designer's task. Recently a number of researchers have begun to examine design methodologies with the goal of providing additional computer based assistance to the designer. An elaborate, empirical study of human designers by Ullman, Stauffer, and Dicttrich[l] is intended to provide a basis for the development of intelligent computer based tools for mechanical designers. Ullman et al have observed that designers tend to follow a single concept in their design configuration rather than to explore alternative conceptual designs. We believe that identifying critical design relationships will encourage designers to explore more alternative configurations by helping them make quicker, more convenient and more focused evaluations of each configuration. Diettrich and Ullman [2] have also identified what they believe are several basic requirements of software tools for intelligent design aids. Two of these requirements are the ability to conduct a deep search of design spaces to evaluate alternative designs and the ability to infer consequences of a particular design decision on other components of the design. We seek to minimize this difficulty by more directly relating the goals of the design to specific design decisions. We believe that the strategies described herein may be useful, not only to the human designer, but also in computer based design assistance systems. Other programming environments which aid in the automation of design include languages such as DSPL, created by Brown and Chandrasekaran [3]. They have focused their efforts on an approach to building expert systems for routine design by creating a programming language in which to express routine design problem solving knowledge at the task level [4,5]. This language is geared toward routine design such as the design of air cylinders in which the general configuration is determined beforehand. The system does not explore alternative design configurations and is not intended to provide increased insight on many alternatives. Dixon has also described a paradigm of design and developed systems to assist designers. His model, based in part on the iterative and recursive nature of design [6], involves decomposition, specification and an iterative redesign procedure. Dixon seeks to construct programs that can produce acceptable designs from a given trial design. He has implemented some of his ideas in computer based AI strategies to solve a limited class of design