Notes

1. Introduction
   1. Engineering & Design
      
      Engineers are primarily seen as those who use fundamental and developing scientific and technical knowledge to create products for the use, need and benefit of society. Engineering Design is the activity of engineers directed toward that goal. The direct product of the designer's activity is not always a finished item ready for use but rather is more likely to be a plan, a process, a model, a drawing, a report etc., commonly lumped under one term and referred to as a design.
      Goal-directed activities are generally referred to as problem solving. We can extend that idea further and observe that almost every activity of human involvement has a goal and constitutes problem solving. Understanding, or at least appreciating, how the human mind handles everyday problems then is useful, if not essential, to formulating a methodology suitable for engineering design. Development of that methodology is the goal of this course, and is the goal of the problem we wish to solve.
      The statement of what engineers do stresses that they provide a service to society, and do so to benefit society. This means of course that they are under ethical obligation as professionals to do so causing no harm to individuals, groups or natural resources.

      Engineering Design & Manufacturing
      

      For many years (throughout most of this century) engineering design activities have been carried out independently of the tooling, manufacturing and testing of the final product. The activity was (or is often, still)part of a sequential process, with the design being passed "over the wall" to the next group involved. When inconsistencies arose, a new requirement had to be passed back through the complete loop again.

      Figure A1 shows the Design/Manufacturing/Test loop for the design of a complex system, a new aircraft or a modification to an existing aircraft. Notice the number of loops back through the process that are possible before final certification. Each of these loop paths, particularly those that occur latest are very costly in both time and cost. It has been suggested that it is ten times more costly to make changes at the next level of the design process instead of making certain that everything is correct and complete at the current level.

      Modern Trends
      Concurrent Engineering (E&J p40)
      

      In today's world-wide competitive marketplace the product cycle must be compressed considerably. Market shares will be lost if the product is not available for distribution rapidly in a changing market place. This has led to a move away from the Design/Manufacturing Loop to what is called Concurrent Engineering.

      Concurrent Engineering implies that the processes and product are developed together. It is a structured process, focusing on penetration of a market. This requires working together through cross-disciplinary teams, a process which has been enhanced by the explosion of electronic communication and information technology. Included in the teams may be tool designers, marketing representatives, customers, distributors etc.

      Computer-Aided Design/Manufacturing (E&J p41)
      

      The computer has not only aided in the communication among those involved in creating and delivering a product; its earliest uses in engineering were (and still are) applied to repetitive simulation and analysis calculations (Computer Aided Engineering) and to visual aids, replacing the tedious drawing required previously. Boeing engineers through electronic-based design of the 777 were able to "assemble" and check parts before manufacture. CAD/CAM procedures have been extended to the point of actually supplying input tapes for digitally-controlled machines which create parts and tools.

      Knowledge-based Systems (E&J p117)
      

      The computer has opened up an exciting and important new direction to Engineering design. The knowledge of experts in a particular engineering area is stored and then retrieved and used when it is pertinent to a particlar problem. Because of this future importance of such systems, the methodology we develop here will address knowledge levels as an implicit part of the process.

2. Problem Solving
   
   1. Essential parts of a problem

      
      In almost everything we do, we are confronted with a problem to solve;

      getting to work in the morning
      walking across the room
      finding a job
      etc.
      

      The common characteristics of these problems are that we are at some place or condition, we want to go to another place or condition and the problem confronting us is: How do we get from here to there? What path should we follow?
      Thus we can categorize a problem as being composed of 3 parts.

      Initial or current condition Goal Solution

      (here) (there) (path )

      In terms of states of knowledge these can be shown equivalently as a continuous change in states as we progress toward the goal. (See also Figure B1).

      Initial State Solution States . . . . Goal State

      (i)Initial
      The initial state of a problem is most frequently the "given" of the problem and expresses where we are as we begin. It includes some of the requirements (called initial constraints below) that are placed upon us initially.

      (ii) Goal

      The goal is the final state and could be characterized as the "purpose" of the problem (e.g. as in a laboratory experiment) or the "required" of a mathematical problem (e.g. the statement of a theorem to be proven). When achieved, the problem is solved.

      (iii)Solution

      This is the "procedure", scheme, action or process to be followed to solve the problem. It might be a direct path, but more often it is a set of steps to go through to get to the solution. Each step becomes the goal of the previous step and once achieved, the current step becomes the initial state for the following step.
      For example, suppose you have picked up a car in downtown Seattle (initial state) and wish to drive to the space needle (goal state) which you can see in the distance. You have no map, a full tank of gas and are meeting a friend there in one hour (initial state constraints). What (solution) path should you take to get there? You must come up with a scheme, or a plan to follow. You are constrained to stay on the road of course (called physical constraints below) and must obey traffic laws and drive safely (called implied constraints below).
      If you are lucky there will be a sign with an arrow saying "Follow This Sign to the Space Needle". Your solution plan would then be to follow the signs. i.e. you have the direct path solution mentioned above.
      You might give some structure to your plan by a path which follows ever-increasing numbers of city blocks, again under the above constraints, until you achieve success (the direct path here is a square search used by search and rescue aircraft).
      You might just start driving, hoping that you will come across your goal, performing what is called a blind search or you might keep track of those attempts that have failed and thereby perform a systemic search. Blind search and systemic search are two forms of what is termed "trial an error" search.
      The chances of eventual success using a trial & error search may be good but it is very time-consuming. Obviously these approaches are very inefficient.
      Since you can see your goal you might decide that your scheme will be to turn toward the space needle whenever you see it (hills blocking your view are now additional constraints) and whenever the previous constraints allow you to turn.
      Each time you turn, if you indeed get closer, you enter a new solution state, which from your point of view can be considered a brand new starting point or initial state. The important characteristic here is that you ask yourself at each step "what shall I do next?" and you are applying a "proximity search". The form of solution, where one attempts, as a scheme, to reduce one variable (distance to the goal) is termed a hill-climing method. When more than one variable is involved, the form is called means-end analysis. Hill-climbing and means-end analysis are two forms of proximity search methods.
      A further classification for searches for solutions are "knowledge-based" searches. These use knowledge the solver has in memory or knowledge searched out to aid in the solution. That knowlege may not be directly related to the problem to be solved but may be experience with an analogous problem or a similar class of problem.
      We classify Trial & Error and Proximity searches as heuristic procedures in that they are valuable for finding a solution but do not guarantee that one will be found. Algorithmic procedures on the other hand, do guarantee a solution if they are used properly. The "direct path" and the "square search" mentioned above apply algorithms, as well, of course, as a mathematical equation or set of equations which express the goal state in terms of the intitial state. Knowledge-based searches can be either heuristic or algithmic.
      In summary:
      Searches for Solution

      1. Trial and Error
         1. blind search
         2. systematic search
      2. Proximity
         1. hill-climbing
         2. means-end
      3. Knowledge-based

      Solution Procedures

      1. heuristic
      2. algorithmic

      The search methods discussed above can all be applied to engineering problems, but no one is suitable for all. We would like to develop a well-defined strategy to follow for such problems but we must accept the likelihood that the best strategies are dependent upon the type of problem to be solved. It will help if we develop strategies that are natural and consistent ones for human beings; for that we need some understanding of how we think, of how the mind works. This, along with the importance of human thinking processes in knowledge-based systems, warrant us taking a brief look at the working of the human mind.

      The human mind

      The tool for everyday problem solving and one that is always an essential part of complex problems, is the human mind. The mind is our center for reason and for the storage of all our knowledge. That knowledge, all that we know, comes into the mind only through our senses; some is stored, some is discarded without storing.

      How we individually use knowledge to solve problems and go about everyday activities is the subject of a great deal of study by psychologists. Much of the impetus for this activity comes from a desire to use computers to perform some of the reasoning tasks that only humans can do now ( e.g. artificial intelligence applications to knowledge-based design).

      Computers do not operate under the same principles as does the human mind, but there are some similarities which at least allow us to draw analogies that may help in the understanding of what occurs in the latter.

      1. The computer & the mind
         Information (knowledge) comes into the computer through its "senses", the keyboard, tapes, discs, network connections, digitizers etc. and is placed into a temporary working register, and then, if required, into its internal memory locations. If the information is to be kept permanently it is stored in permanent memory (e.g. a hard disc).
         This is analogous to information coming into the mind through, for example the eye and placed as an image in its temporary sensory memory. If the information is important it will be stored in a temporary location, analogous to the internal memory location of the computer, that is in "short term memory" (STM) in the brain. If the image is to be kept permanently it is stored in "long term memory" (LTM), the brain's "disc" storage.
         A key difference between the mind and a computer is that objects in computer memory can be located using known addresses of the objects in memory. Images are stored in a hierarchial form throughout the brain, not at a single convenient location. The mechanisms for retrieving information in the brain is very complex and for most of us often very frustrating. A very brief summary of our memory mechanisms as suggested by some psychologists might help us at this point. Remember though that our understanding is primitive at best.

         Memory structure

         (i)Sensory

         Images from the senses are generally fleeting, probably held for less than a second, and are not all are processed. As we drive along in a car, the landscape passes by and we continually monitor it with our eyes. Yet we certainly do not retain all the detail of what we see. Only if we make a conscious effort, or if there is an unusual occurrence, do we pass this information into STM.

         (ii)STM

         Short term memory is a location where we can store a limited amount of information for a short time for immediate processing. It holds about a half-dozen "chunks" of information which decays in the order of 10 to 20 seconds if it is not consciously refreshed ( e.g. repeating a phone number over and over). A chunk may be as small as a single digit, as in part of a phone number (hence the 7-digit limit on phone numbers) or large enough to hold the information on location of pieces on a chess board in the case of a master chess player.

         (iii)LTM

         Long term memory is essentially permanent. It includes learned factual information processed in STM and then consciously memorized. It also includes learned procedural information. The latter become automatic skills such as walking, typing, guitar fingering, etc.

         Retrieving factual information from memory as mentioned above may be difficult; we have all experienced the sudden loss of a word or a name. The simple addressing system of a computer is not available to us, so that many "memory enhancement" techniques have been proposed and marketed. See for example Hayes

         Senses

         All external information enters the human mind through several sensors imbedded in the following organs.

         skin: touch

         tongue: sweet/sour taste

         nose : smell

         eyes : sight

         ears : hearing

         Information is returned to our surroundings through

         voice

         activation of nervous and muscular system of the body

         Thinking, Images, Stories, Analogies, Reasoning, etc.
         For those who are interested I recommend R. Sylwester

3. Taxonomy of Engineering Design
   1. Knowledge Types
      If we now follow some of the general considerations of problem solving as outlined above, we can view engineering design as a transformation from an initial knowledge state to a final knowledge state. The particular initial and final states depend upon on the problem type and we transform from only one initial state to one final state. Knowledge at a higher level state implies that all lower level states are known. The two states involved may of course be intermediate states or design tasks with sub-goals along the solution path as described earlier.
      C. L. Dym has reviewed taxonomies of Ullman and Dixon et al and suggests the following 6 types of knowledge for initial and final states of design problems. Note that these proceed from most abstract to most specific. The levels must be known or specified at an initial state, but are generally to be determined in the final state. See Figure C1

      1. perceived need
         This is the most abstract level of knowledge about a design. It is in general a simple word statement sometimes referred to as a primitive need. It does not, at this level suggest an artifact but in an engineering context the need statement would imply that it would be satisfied by a scientific means,
         e.g., the range of your production aircraft must be increased to be competitive.

         function

         The function statement states what must be accomplished, but not how to do it. In an engineering context it would include design requirements,
         e.g. range requirements, payload requirements.

         physical phenomena

         These are the physical principles that are prescribed or determined.
         For example, total drag reduction techniques must be used.

         embodiment or concept

         The generalized shape, form or approach consistent with the function and physical phenomena is specified or is to be determined.
         e.g., re-wing to reduce induced and parasite drag.

         artifact type

         The concept is further specified or detailed as to its type, (no dimensions or only partially dimensioned).
         e.g., a NACA airfoil may be specified in a wing design, but not its thickness or chord.

         artifact instance

         Actual values are to be determined,
         e.g., the airfoil, its thickness and its chord are all to be determined.

      Criteria

      Engineering problems may be classified as either analysis problems or synthesis problems. analysis problems and synthesis problems.

      1. Analysis is a straightforward determination of the behavior of a well-defined system. The initial state is specified.
      2. Synthesis involves the determination of a system that will satisfy a defined behavior or requirement. The initial state is in terms of design or performance requirements.

      Most design problems are of the synthesis type, and as such do not have a unique solution as do analysis problems. Therefore some sort of criteria must be established for selecting among many syntheses.

      The criteria may be somewhat qualitative in nature and judged on a basis of reasonableness or may involve an optimal solution, if mathematical functions (see criterion function) can be developed to express the criterion or criteria to be optimized.

      Problem Types

      1. Problems in engineering design can be classified to some extent by their initial and final knowledge states. They may be further classified according to the criterion being applied.In principle, a problem can be from any lower initial state to any higher final state. The most common types, and those of most interest to us are as follows:

         Design Type	Initial State	Final State
         conceptual design	perceived need function	concept concept
         preliminary design	concept artifact instance	artifact type artifact type
         detailed design	artifact type	artifact instance

         Notice that in the second case under preliminary design in the table above, the knowledge level of the initial state appears higher than that of the goal state. This will occur where an existing artifact is being modified for a new purpose or application. The initial state instance obviously cannot be expected to satisfy the goal state. The initial state in this case is sometimes referred to as a "baseline". Our re-wing example of the last section is a problem of this type.
         Conceptual design processes are heavily dependent upon creativity and experience of the designer and really require a separate methodology. This is discussed briefly below. From this point on we will be concerned only a methodology for preliminary design.
      2. Some common classifications according to criterion are presented below. A design type may of course belong to both this table and the previous one.

         Design Type	Criterion
         feasibility study	any feasible synthesis i.e.,satisfies all constraints
         optimal design	optimal solution i.e., the "best" synthesis

      3. Classifications may also be made according to limitations or characteristics of the parameters or problem variables. Again a design type may belong to both this table and the previous ones.

         Design Type	Parameter Limitation
         probabilistic design	uncertainties in material properties
         experimental design	uncertainties in processes

      Constraints

      Engineering designers must consider many constraints that are implicitly or explicitly applied to their efforts. These range from those that are implied by the profession (responsibility to society) to those that are implied by natural, scientific laws, to those that are explicitly stated as requirements by a corporation or federal agency. The types are categorized partially as follows:

      1. Implied
         1. Ethical
         2. Legal
         3. Environmental
         4. Safety
      2. Practical
         1. Need
         2. Cost
         3. Profitability
         4. Manufacturing
         5. Marketing
         6. Reliability
         7. Ergonomics
      3. Initial
         1. Specified Initial State
         2. Performance Requirements
      4. Physical
         1. Geometrical
         2. Material Allowables
         3. Failure Criteria
         4. Natural Laws
         5. Functional

      Solution Process
      As mentioned earler, working through the solution process involves working through sub-states of knowledge from the initial state to the goal state. The process we will consider in what follows are applicable only to those where the initial state is at the conceptual level or higher as discussed above. Further, the implication is that we consider engineering design to be synthesis rather than analysis problems.

      Ullman suggests four components of a design process as being essential to characterizing a design:

      1. Plans
         Planning the design, deciding on the approach.
      2. Processing Actions
         Putting the plans (above)into effect
      3. Effects
         How the processing affects the designs
      4. Failure Actions
         What to do when the goal criterion is not met.

      Note the similarity in these 4 parts of a design process to the parts of a management process in E&J (p60). Also these closely parallel the steps in quality improvement processes by Deming, namely his PDCA process Plan-Do-Check-Act. All of these are loop processes being done over and over, until the goal criterion is met.
      The selected portion of Ullman's taxonomy can then be summarized as in Figure C2. The following sections will expand the taxonomy to give more specific process and problem descriptions suitable for this course .

4. Design Process
   1. Problem Statement
      Whatever the type of design problem being addressed or if the problem is only a task within larger effort, a clear and unambiguous statement of the problem must be established and understood by all involved, before any other action is taken. Based upon our previous discussions a Problem Statement should completely define the Initial State and will include any initial state constraints or design requirements.
      

      Management

      Once the problem description is established and understood a planning or management phase is essential, whether one or many designers are involved. Specific tasks here would include

      1. Information collection
      2. Objective analysis
      3. Task establishment
      4. Resource management
      5. Team assignments
      6. Scheduling

      A more detailed approach is summarized under "Plan" and "Direct" in Figure 2.3 of E&J.

      Solution

      The search for a solution can begin transforming our knowledge from the initial state to the goal state through a set of intermediate levels.

      1. Approach (Plans)
      2. Modeling (Processing Actions)
         1. Physical Model
         2. Criteria Functions
         3. Mathematical Model
      3. Nominal Solution (Effects)
         1. Analysis
         2. Feasibility Assessment
      4. Evaluation & Decision (Failure Actions)
         1. Design Space Studies
         2. Trade-off Studies
         3. Iteration
         4. Assess Concepts
         5. Parameters Choice
         6. Reassess Criteria

      Presentation

      Throughout the design process communication is essential to an efficient operation which minimizes misunderstandings. This normally includes written status reports and of course always a final report. If there is no reporting, there is no design. Communications are generally one or more of the following types and depend upon management requirements and the end artifact being produced.

      1. Graphical
      2. Text
      3. Numerical
      4. Verbal
      5. Physical model

5. A Design Methodology
   We are now prepared to set up a methodology which can serve as a guide in proceeding with an engineering design problem. It is not our intent to stifle creativity by forcing an approach but we do need uniformity for communicating with each other in a course like this one. The approach leaves quite a bit of flexibility and will aid you in understanding the language of design as used by modern designers.
   It is important to understand that we have not delineated between a large multi-tasked project and a small project which might be a task or sub-task of a major project. So whether you are the Chief Engineer, a Team Leader, a Team Member, or a Consulting Engineer, this methodology is applicable. Some of the parts will be modified of course, depending upon the task level.
   The steps of the methodology are (loosely) related to the problem knowledge levels and the taxology in Figure E1. An outline is given in the table below.

6. Conceptual Design(E&J p10,11)
   Conceptual design is normally performed by very experienced design engineers in a company. They are creative people with a well-developed background of knowledge, but willing to be open-minded. As mentioned above (Section C.3) this process should take the problem from and initial state of recognized need or specified function to one or more concepts which potentially should satisfy the need.
   The 4 parts of the process of Section D above are still followed as in preliminary design. The Problem Statement is the Need Statement (D1) and Management is much as in D2.
   The solution process (D3)for conceptual design is usually a group effort and should be unstructured so as not to restrict ideas and creativity. A commonly used process is "brainstorming". The purpose of brainstorming is to make sure all possibilities are considered. Even if ideas have no apparent merit they may lead others to a new productive train of thought.
   Brainstorming goals and rules ( see also E&J p 11) might be as follows:

   OBJECTIVE:

   • generate many ideas and concepts
     
   • minimize personal bias and expand viewpoints by working in a group

   RULES:

   • criticism and judgement are not allowed
   • the more the better
   • think far out

   ORGANIZATION:

   • No leader
   • No recording secretary preferably, use tape recorder

   How do you think up new concepts? You must of course rely on things you already know, and attempt to apply this knowledge to a new situation. The way most of us would do this is by developing in our minds, images of analogous (similarly behaving things) that we already understand, and apply them to the problem at hand.

   Finally (D4), the presentation will be in the form of a Problem Statement and sketch for each concept to be studied for feasibility. The problem statement should include a discussion of important goals and criteria.

   Decision Making

   The job of an engineering problem solver is always involved with making decisions among various alternative courses of action.

   Variables (or parameters) used in design may be grouped into variables associated with the three parts of a problem: input variables (initial state), solution variables and output variables (final state). If we begin a problem with the input variables specified or fixed, and the criterion for solution has been established, then a fixed solution process will lead to the final values for the output variables. However, many variables are not under the control of the designer e.g., material variability, the weather, commodity prices, political unrest, etc.These latter "uncertain" variables must be handled as statistical variables.

   There are usually many alternative paths which might be taken to get to the final state, or there may be many ways to manage a multi-tasked effort (or many sub-goals in the process) where each task in some way depends upon completion or partial completion of some of the others.

   A single criterion or Criterion Function is seldom the case and it become necessary to decide on the relative importance of each in forming a final solution.

   All of these alternative choices require decisions to be made be the designer. The designer needs tools to aid in making these decisions and we will discuss some here grouped under the headings Decision Matrix, Decision Tree, Networks.

   3.1 Decision Matrix

   The decision matrix is a table of alternative strategies vs decision variable. The values in the matrix are some measure (called the utility) of gain, or loss, or goodness specified by the designer. For example suppose we have undertaken a conceptual design study and have established 4 concepts. Let us further suppose that a study of the need statement has resulted in 5 desireable attributes for concepts. If we rate each of these attributes on a scale of 1 to 5 might have a utility table as follows:

   assessment	value
   excellent	4
   very good	3
   acceptable	2
   barely acceptable	1
   unacceptable	0

   Then a decision matrix might take the following form:

   Concept	safety	convenience	weight	score
   concept 1	4	2	2	8
   concept 2	3	2	1	6
   concept 3	2	3	4	9
   concept 4	1	3	3	7

   The score column then gives an overall ranking of the various concepts. However some of the attrbutes are surely more important than others. For example if people are to be in contact with this device, safety may be more important than convenience. A weighting factor is therefor generally applied to each attribute to indicate its relative importance.

   3.2 Decision Tree

   3.3 Networks