Design and Optimization of Thermal Systems Episode 3 Part 6 pdf

These considerations are of critical importance, since a design that is technically feasible may not be acceptable because of the excessive cost or because it violates regulations regard

and for the others

Determine the optimum path that leads to the minimum amount of

time in going from A to D.

C B

4 4

3 3

2 2

1 1

FIGURE P10.20

in practice for a successful design have also been outlined These include nomic, safety environmental, regulatory, legal, and other issues that may be tech-nical or nontechnical in nature These considerations are of critical importance, since a design that is technically feasible may not be acceptable because of the excessive cost or because it violates regulations regarding safety or the environ-mental impact.

eco-In this chapter, we will consider knowledge-based design, which is a

non-tra-ditional design methodology based on experience, informal approaches or ristics, information on existing systems, and current practice The main elements

heu-of this method and the overall scheme are outlined, followed by examples heu-of a few thermal systems to demonstrate the power and usefulness of this approach Also considered are some additional important issues with respect to the design

of thermal systems, such as professional ethics and other constraints Also cussed in this chapter are the sources of information that may be employed to provide inputs for design Some of the important sources for information on material property data, characteristics of components, economic variables, opti-mization techniques, computer software, etc., are given An overview of the design and optimization of thermal systems is also presented This overview serves to put the entire design and optimization process in perspective Several design projects are included as problems at the end of the chapter to cover the entire process for typical thermal systems Groups of students may use these as projects in design courses that involve design and optimization undertaken over the period of a semester

dis-11.1 KNOWLEDGE-BASED SYSTEMS

With the extensive growth in computer-based design, considerable effort has been directed at streamlining the design process, improving the design methodology, automating the use of existing information, and developing strategies for rapid con-vergence to the final design Many of these techniques are discussed in the literature

(Suh, 1990) A particularly important approach that is finding increasing use as a

component in the design process is that of knowledge-based design The

develop-ment and use of this tool is based on the premise that the more the machine or computer knows or learns as it proceeds, the more effective and efficient this process will be Therefore, an attempt is made to include relevant information on the system, process, and current practice, adding to this information with time and employing the information base to guide the design The experience gained by a designer over time and various so-called “rules-of-thumb” or heuristics are also included Recent advancements in computer science in areas such as information storage and retrieval, artificial intelligence, and symbolic languages are used in developing knowledge-based design aids, which are then used to provide inputs to the design process

11.1.1 I NTRODUCTION

Storing the knowledge and experience of an expert in a particular area and using

these to make logical decisions for selection, diagnostics, and design is the basic concept behind knowledge-based systems Therefore, knowledge-based systems are

also known as expert systems and involve artificial intelligence features such as

1 Stored expert knowledge and experience

of software currently in use include MYSIN, which diagnoses diseases; PECTOR, which evaluates potential ore deposits; MACSYMA, which solves problems in calculus by using symbolic manipulation; and DENDRAL, which finds structures of complex organic compounds Several expert systems have been developed for the design of different types of systems, including thermal systems, and are discussed later in this chapter

PROS-The knowledge-based methodology requires efficient storage of expert edge so that repetition is avoided, minimum space is taken, and rapid retrieval of information is possible A common arrangement used for the storage of informa-

knowl-tion is a tree structure, in which objects are organized in a hierarchical scheme with certain objects taken as subclasses of other objects These subclasses inherit

the common features from the objects above it Therefore, a relationship similar

to that of a parent and child is established with respect to inheritance of teristics and properties Figure 11.1 shows a tree structure for storing information

charac-on animals, with charac-only two choices at each step So if we ccharac-onsider a cat, its chy indicates that it is a nonvegetarian, nonflying land animal Only information specific to cats needs to be placed at the particular location, with more general features being derived from its hierarchy

hierar-Similar tree structures can be developed for thermal processes, as given in Figure 11.2 for cooling systems for electronic equipment Different types of cool-ing arrangements, fluids, and transport mechanisms are included Figure 2.7 gave

a similar tree structure for forced convection cooling, considering different types

Fan Other Horizontal

Vertical

Horizontal Vertical Other

Liquid N 2 Forced Natural

Liquid

Cooling

of electronic equipment

FIGURE 11.2 An example of a tree structure for storing data on cooling of electronic

equipment.

of systems Similarly, Figure 2.32 gives a tree structure that can be used to store information on different types of materials Again, the use of subclasses helps in information storage and retrieval The types of information that may be stored are knowledge and experience available with an expert, material characteristics, design rules, empirical data, and other inputs that may be used for design In many practical cases, intuitive ideas, heuristics, and general features are used to guide the design These may also be built into the system to obtain an acceptable

The user interacts with the front end, which interfaces with the other

com-ponents of the system Numerical, symbolic, or graphical inputs are provided by the user to the front end The geometry, configuration, dimensions, materials, and operating conditions are entered The front end then obtains material property data and supplies these to the computational modules to obtain the simulation results needed for design Empirical data, correlations, component characteris-tics, etc., may be included in the computational modules to complete the modeling and the simulation These are linked with the knowledge base The given design rules are then used to obtain the final design, which is then communicated as graphical or tabulated results

Material database

Computational modules

Graphics output Front end

Knowledge base

User

FIGURE 11.3 Components of a knowledge-based system for design.

Front End and Knowledge Base

The front end contains the design rules, constraints, requirements, design ables, and other aspects pertaining to the given system Some of these, partic-ularly constraints due to material limitations, are obtained from the databases associated with the system The knowledge base contains the expert knowledge, which includes

vari-1 Information from previous designs

2 Rules of thumb

3 Heuristics based on informal methods

4 Safety and environmental regulations

5 Information on existing and similar systems

6 Current engineering practice

7 Other information that constitutes the experience of a designer

All this knowledge may be used in the development of a realistic and ful design Therefore, the knowledge base is a very important component of this design methodology It helps a designer avoid mistakes made in the past and use earlier design efforts for accelerating the iterative design process It is worth not-ing that many of these aspects are typically employed in the design process even

success-if the systematic approach given here is not followed

analyz-1 Gauss-Jordan method for matrix inversion

2 Least squares method for best fit

3 Numerical differentiation and integration

4 Successive over relaxation (SOR) method for linear algebraic equations

5 Runge-Kutta method for ordinary differential equations

6 Finite difference and finite element methods for partial differential equations

Separate modules may be developed for a given problem, such as a glass nace, air-conditioning system, diesel engine, etc Information on the discretiza-tion methodology, convergence criteria, data storage for graphics, etc., is provided

fur-to enable accurate results fur-to be obtained and linked with the other parts of the tem Programming languages such as Fortran and C or software like MATLAB and Mathcad are used for carrying out rapid computations Parallel computing, with a large number of processors, may also be employed for faster response from these modules Empirical data, usually in the form of correlations, are also included here.

sys-Material Databases

The material databases contain information on various materials that are of est for the types of systems under consideration Important items that may be included are

inter-1 Thermal properties

2 Allowable ranges of temperature and temperature gradient

3 Strength data, hardness, malleability, and other physical characteristics

4 Cost per unit mass or volume

5 Availability, including import considerations

6 Manufacturability or ease of fabrication

Thermal properties, such as thermal conductivity, diffusivity, specific heat, density, and latent heat, are stored for thermal systems, usually at different temperatures or as functions of temperature In order to avoid damaging them, constraints on temperature and temperature gradient are given for the various materials Damage may occur, for instance, due to the melting or charring of the material, thermal stresses, deformation at high temperatures, etc Cost, avail-ability, manufacturability, strength, and other relevant properties are important in material selection and should also be included The information stored is usually strongly dependent on the application Again, the information is stored in terms

of classes and subclasses of materials, as shown in Figure 2.32, to facilitate sion of additional property data and information retrieval

Therefore, the outputs can be fine-tuned to a given application For example,

if a plastic screw extruder is being designed, the pressure and temperature rise in the extruder as the material flows from the hopper to the die may be displayed Color graphics or contour plots may be used to indicate hot and cold regions in the flow As an example, Figure 11.4 shows the temperature distribution in the channel of a plastic extruder in terms of isotherms The temperature distributions across the channel at four down-channel locations are also shown This figure

shows how the plastic heats up as it moves from the hopper at z* 0 to the die at the other end of the channel In iterative design, the results may be displayed after each iteration, allowing the user to observe the convergence to the final design and to intervene if the iterative process appears to be diverging or if the design emerging from the design process is not satisfactory

Graphical inputs to the front end are also important in many applications since the geometry, boundary conditions, and dimensions are most conveniently entered on a graphical display An example of casting is seen in Figure 11.5, where the mold, cast cavity, runner, and thermal conditions at the outer surfaces are shown Such a schematic may be displayed and the user may interactively enter the appropriate quantities and parameters such as dimensions, heat trans-fer coefficients, and materials Many available programming languages, such as Visual Basic, are particularly suited for such graphical inputs

Languages

The programming language employed in the knowledge-based design system forms another important consideration Symbolic languages, such as LISP, PROLOG, and SMALLTALK, which allow the use of symbols rather than just numbers for manipulation and control of the software, are particularly useful for the front end and the knowledge base For instance, descriptions of a surface as

“smooth,” viscosity as “high,” and disturbances as “small” are all symbolic in form and digital values may or may not be associated with these This is similar

to the concept of fuzzy logic discussed in Chapter 7 In the storage and use of knowledge, we need symbolic representations for

1 Symbolic manipulation of objects

2 Rules of thumb and heuristic arguments in symbolic form

3 Inputs/outputs given in symbolic form

4 Use of symbolic notation for storage of data

5 Symbolic representation of design rules

A symbolic environment allows the versatility and flexibility needed for specifying design rules, constraints, expert knowledge, and other pertinent infor-mation LISP is a commonly used language and variations of LISP are often used

to develop expert system shells in which the rules and expert knowledge for a given application can be easily entered (Winston and Horn, 1989) PROLOG and its various versions, such as Sigma PROLOG, have a variety of other features that

may be appropriate for certain types of applications (Clark and McCabe, 1984) One of these features is easy link with computational modules, which are often based on languages such as FORTRAN and C, and this makes PROLOG attrac-tive for the design of thermal systems SMALLTALK is a more powerful lan-guage, but it is more complicated and more difficult to implement.

As previously mentioned, computations are generally performed in tific programming languages Parallel computing is particularly attractive for the design and control of thermal systems since very fast computations can be achieved, allowing real-time simulation of many systems and providing appro-priate graphical outputs Parallel machines such as Hypercube, Ncube, Convex, and computer clusters employ algorithms that make effective use of the multiple processors to speed up the computations

scien-11.1.3 E XPERT K NOWLEDGE

Expert systems are based on the knowledge, or expertise, of an expert, so that the

logical decisions made by an expert in a given area can be made by the computer itself Empirical data, heuristic arguments, and rules for making decisions are all part of this knowledge-based methodology The expert knowledge is obvi-ously specific to a given application and represents the knowledge and experience acquired by the expert over a long period of work in the area of interest (Jackson, 1999; Giarratano and Riley, 2005)

For instance, if an expert system is to be developed for the solution of ential equations, an expert will tell us that the solution depends on the nature of the equation, some of the important characteristics being

differ-1 Ordinary or partial differential equation

2 Linear or nonlinear

3 Order of the equation

4 Characteristics: elliptic, parabolic, hyperbolic

FIGURE 11.5 Geometrical configuration of an ingot casting process.

Using mathematics, we can develop rules to determine the nature of the tions A database of analytical solutions can be built and expert knowledge can

equa-be used to determine if a particular equation can equa-be solved analytically Then a search in the database may yield the desired solution If a numerical solution is necessary, the knowledge base may again be used to choose the following:

1 Numerical scheme

2 Grid size and discretization

3 Appropriate time step so that numerical stability is ensured

4 Convergence criterion to terminate iteration

5 Initial guessed values, if needed

A computational expert uses his or her experience and knowledge in ing many issues such as the method, grid for desired accuracy, termination of scheme, obtaining an analytical solution if possible, accuracy of the numerical results, and so on Therefore, one who is presumably not an expert in this area can use this expertise effectively to guide the solution A database of analytical solutions, different numerical methods, stability criteria, and other constraints and rules are built into the expert system (Russo et al., 1987)

decid-Similarly, expert knowledge on other applications may be developed and included in the design process, as seen from the examples presented later in this chapter A very important element in the development and use of knowledge-

based methodology is object-oriented programming, developed using a

program-ming language such as C or Java In a non-object-oriented, or procedural,environment, a programmer begins with an initial state and, through a set of prescribed procedures, arrives at the goal For example, to invert a matrix in procedural programming, one must prescribe every step of the method In an

object-oriented environment, a message would be sent to an object called matrix inversion, which is like a subroutine and which would invert the matrix The

object would already have all the necessary information on the procedures for inverting a matrix Thus, an object is the housing used to store information This

information comes in the form of procedures known as methods, which can act

on the given data, being a matrix in this example, upon receiving a message to

do so Objects are organized in a hierarchical scheme with certain objects taken

as subclasses of other objects, as discussed earlier These subclasses, therefore, inherit the common features from the methods of the object above it The three components of the object-oriented system described here are encapsulation, mes- sage passing, and inheritance (Cox, 1986; Budd, 2002).

All the information needed to use an object is stored in the object itself The methods embedded or encapsulated in each of the objects are unique to that object Therefore, encapsulation makes different objects reusable and reduces duplication within a program Message passing is independent of the methods and

is like a call statement, making it possible to execute the methods stored in the object Inheritance gives each object access to the features of the methods from the class above it on a tree structure This aspect allows any system to be easily