Joints ........................................................................................ B

Structural Sizing ...................................................................... C

Presentations ...................................................................... D

Repairs ............................................................................... E

Mass Properties ...................................................................... F

Design with Composites ............................................................. G

Motherhood, Truisms, and Catch Phrases .................................. H

Cost Estimating ...................................................................... J

Technical Writing ...................................................................... K

Computer Programming ............................................................. L

Cognitive Human Factors .................................................... M

Reference Books for Aerospace Structures Design

Index

DESIGN PRACTICE

A1) If a design looks bad, it probably is bad...better find out what your subconscious is trying to warn you about. (Confidence rating = 8)

A2) Design is an iterative process, and the designer should expect to cycle any given concept at least three times; first to get the concept out on paper where it can be studied, the second to clean up the obvious shortcomings of the first iteration, and the third to transform the crude but workable second iteration into a partially optimized and somewhat polished design. (Confidence rating = 9)

A3) Flight structure must withstand the entire environment imposed upon it, but designs are usually driven by a single aspect of that environment which can be determined by inspection. For example, compact structures, such as brackets, are usually sized by strength considerations. Large or extended structures, such as cradles or pallets, tend to be sized to meet a target stiffness. Thin panels may be sized by acoustic loading, particularly if they are located near an acoustic energy source, such as a rocket engine. (Confidence rating = 8)

A4) The stiffness and buckling behavior of thin panels tend to be governed more by the short dimension of the panel than by the panel's long dimension. For this reason, a stiffener pattern that breaks a large sheet up into long narrow bays is frequently preferable to a pattern that produces approximately square bays. (Confidence rating = 7)

A5) When comparing two sections for torsional capability, the section that will contain the larger inscribed circle will generally be the stiffest. ( This is an approximate version of the membrane analogy described, for example, in Theory of Elasticity, by S.P. Timoshenko.) (Confidence rating = 9)

A6) Always double check calculations. An engineering computation is worthless until it has been done twice. (Confidence rating = 10)

A7) Accurate results can often be obtained from crude calculations by bracketing a problem with conservative and non-conservative approaches then making a considered interpolation between the two results. (Confidence rating = 9)

A8) There is at least one bug remaining in all computer programs. All computer output should be subjected to a sanity check before being put to use. Examples of check procedures are hand analysis, use of a different piece of software or different model, or comparison of results to similar existing hardware.

(Confidence rating = 10)

A9) It is generally desirable to separate primary and secondary structural functions, i.e. don't drill the main spar full of holes to support black boxes. (Confidence rating = 8)

A10) While it shouldn't need stating, a straight line really is the most efficient load path between two points. Load paths containing abrupt angles generally indicate poor design. (Confidence rating = 10)

A11) For a truss or space frame type beam, the optimum orientation for diagonal members is at about a 45 degree angle from the primary load direction. Greater angles drive up the load in the members to excessive levels, while smaller angles require an overly long load path. (Confidence rating = 8)

A12) As an aid in concept generation, form the habit of rotating your point of perception. For example, when designing a joint that seems to have a "natural" parting plane in the X-Y plane, new insights might be obtained from considering solutions with parting planes in the X-Z or Y-Z planes. (Confidence rating = 9)

A13) A space frame is generally the optimum structure for carrying discrete loads from point to point. Beaming or shearing point loads from place to place should only be done as a last resort. Beams and shear webs are normally associated with distributed loads or awkwardly constrained envelopes. (Confidence rating = 9)

A14) Good layout practice is to first make an accurate print or drawing of all given information for a design problem. This may include interfaces, stay-out envelopes, geometry of nearby parts, etc. Then use onionskin to overlay this skeletal layout and sketch in a wide variety of conceptual design solutions far more rapidly than such ideas could be looked at using Unigraphics or other computer aided drawing tools. The most promising ideas can then be rendered on Unigraphics to impose the discipline of scale on the designers thinking. (Confidence rating = 10)

A15) Avoid leaving sharp edges with less than 90 degrees included angle in load carrying structure. Such edges will become damaged in transit and service and act as crack starters. (Confidence rating = 10)

A16) Load paths should be continuous. For example, a stringer should not end abruptly at a door or in the middle of a panel. (Confidence rating = 10)

A17) To gain an understanding of how a structure works and of probable failure modes, it is helpful to draw deflected geometry sketches. (Confidence rating = 10)

A18) An effective approach for trade studies is:

a) Establish design requirements.

b) Find the boundaries of the "solution space", i.e. try to identify all possible design approaches.

c) Select specific configurations representative of all identified approaches.

d) Eliminate all configurations that do not meet design requirements.

e) Compare all remaining configurations on the basis of total system costs.

f) Recommend lowest cost design that meets the customer’s requirements.

(Confidence rating = 9)

A19) One frequently sees trade studies done by:

a) randomly selecting some group of concepts,

b) ranking these by arbitrarily selected criteria,

c) then subjectively weighting these ranking numbers to produce a numerical score which is used to select a concept for recommendation.

This approach rarely proves convincing to anyone except the person who did the study. (Confidence rating = 9)

A20) In performing trade studies, it is often desirable to assign a dollar value to potential weight savings. Ideally, the customer will establish a "value of a pound" number. Lacking this, for expendable upper stages and payloads, it is conventional to use the launch cost of a pound to the orbit of interest, devalued by 20% to reflect the difficulty of full utilization of such weight savings. For first stage structures, divide these "value of a pound" numbers by five. For reusable spacecraft structures, the single launch "value of a pound" can be multiplied by the expected number of missions. Table A1 gives approximate launch costs per pound for commonly used launch vehicles in 1992 dollars. (Confidence rating = 7)

A21) Make full scale prints from the design geometric database to calibrate your designer's eye. Continually working with subscale views on the workstation monitor can result in poorly proportioned designs; for example, flanges made overly thick because they look so fragile at quarter scale. (Confidence rating = 10)

A22) Casting may be an appropriate fabrication technique for making parts in quantity when low ductility can be tolerated. The following table provides a general comparison of metal casting processes.

(Confidence rating = 8)

A23) Dr. Bollard, of the University of Washington, used to say that nature has only three numbers; zero, one, and infinity. His point was that when confronted with a physical problem, simple reasoning can often be used to show that some key parameter tends toward an extreme such as zero or infinity. In engineering practice, it is often helpful to identify such parameters or trends before beginning more elaborate optimization analysis. (Confidence rating = 9)

A24) Perhaps the single most common cause of spacecraft failures is "off-the-shelf-syndrome". A solution to an engineering problem is selected because it has been used successfully for some other application, without due consideration of whether the current problem is sufficiently similar to the problem solved by the borrowed solution. If a component has been selected because it is "already space qualified", or if it is planned to qualify a component "by similarity", the new environment must be compared to the original solution's environment with care and a healthy degree of skepticism. (Confidence rating = 9)

B1) In thin sheet metal, allow a minimum fastener edge distance of 2D + .06 inch, where D is the diameter of the fastener. For unreinforced quasi-isotropic graphite/epoxy sheets, a 3D edge distance is recommended. (Confidence rating = 8)

B2) In thin sheet metal, space fasteners at least 4D apart, where D is the fastener diameter. (Confidence rating = 8)

B3) Design joints to carry their full design load with any one fastener of the pattern out. (Confidence rating = 8)

B4) When designing joints, allow adequate space and edge distance for the installation of fasteners one size larger than specified. (Confidence rating = 8)

B5) Fasteners should be no smaller in diameter than the thickness of the thickest material being joined. Fastener diameter should be no larger than three times the thickness of the thinnest material being joined. (Confidence rating = 7)

B6) Avoid using rivets in tension applications. The tensile strength of a rivet is only about ten percent of its shear strength. (Confidence rating = 7)

B7) Try to keep joints between parts planar. Two dimensional interfaces are far easier to design and fabricate than three dimensional interfaces.

(Confidence rating = 8)

B8) A commonly used value for joint stiffness ratio ( compressive stiffness of joint material compared to that of the joining fastener) is 3. This is appropriate for a typical joint geometry, aluminum structure, and steel fasteners.

(Confidence rating = 8)

B9) For structural applications, use at least a #10 (.190 dia.) bolt. Do not use #8 (.164 dia.) bolts due to the danger of them being mistakenly used for #10's and threaded into a #10 nut. (Confidence rating = 9)

B10) Preloaded joints should be compressed to a value equal to at least 1.25 times the ultimate joint tensile load. Gapping joints are prone to failure.

(Confidence rating = 9)

B11) Bonded joints should be sized to provide an adhesive shear capability which exceeds the adherend strength by at least fifty percent. (Confidence rating = 9)

B12) Bolted joints should initially be designed to use 160 KSI (or lower) heat treat bolts. The higher heat treat bolts are more brittle and should be kept as a last resort to bail out a design that has suffered "load creep". (Confidence rating = 9)

B13) When designing a joint with a seal, the bolt spacing should be less than six times the bolt diameter. Flange thickness should be at least equal to the bolt diameter for steel structure or 1.5 times the bolt diameter for aluminum or magnesium structure. (Confidence rating = 9)

B14) In general, avoid threading steel bolts directly into aluminum parts. Use inserts for a stronger joint that is less susceptible to damage. When special circumstances (such as inadequate edge distance for insert installation) lead to considering threaded aluminum holes, the thread engagement in the aluminum part should be greater than 1.5 times the bolt diameter to develop the load capability of the bolt. (Confidence rating = 8)

B15) Joints with a conical faying surface, such as shear cone or clampband joints, having an included angle of 30° or greater can be expected to separate easily. Surfaces with an included angle of 16° or less can be expected to frictionally lock up after loading. Included angles as low as 20° have been used for clampband joints where a clean separation was not critical. (Confidence rating = 8)

B16) Faying surfaces in structural joints near shock sensitive components should not be allowed to chatter under flight loads. Typically, joints near electronic components have fasteners spaced less than 15 fastener diameters apart over the contacting length of the joint. (Confidence rating = 7)

B17) For adhesively bonded joints in applications with cyclic loads, keep limit shear stresses in the adhesive below one quarter of the ultimate static shear strength of the adhesive for untoughened epoxies. Toughened epoxies may be cyclically loaded to half the ultimate static strength of the joint.

(Confidence rating = 7)

B18) Bolted joints in an aluminum structure will attenuate shock transmitted across the joint by approximately 40% (60% of peak amplitude is transmitted). This rule should be applied to a maximum of three joints in series. This joint attenuation factor is not applied after the peak amplitude remaining drops below 2000 G's. Lockbolted, bonded, or welded joints will not attenuate shock as well as a bolted joint. (Confidence rating = 7)

B19) Helicoil inserts should not be used for structural applications. Because this type of insert is made from a coil of wire, it is more susceptible to failure due to improper preparation of its threaded mounting hole than a conventional insert with a solid wall that can carry load away from an imperfection. (Confidence rating = 9)

B20) Spherical or "Monoball" bearings are often used at the end of column members in an attempt to get "pin-ended" boundary conditions. This design approach has often failed for the following reasons:

a) The loose fit of bolt to bearing produces a high frequency chatter on launch that can fail adjacent electronics. An expanding sleeve bolt can be used through the Monoball to eliminate this problem if free rotation is not required about the bolt axis.

b) When the column member is carrying load, friction tends to lock up the bearing, resulting in significant moment being carried across the joint and imposed on the column. If this loading has not been planned for when sizing the column, the column may buckle in use.

c) The bearing assembly itself is weak normal to the plane of the housing. Sufficient friction loads may be developed in this direction to pop the bearing out of its race.

To avoid these problems, consider using a flexure (a short rod or web) built into the end of a column to limit the moment that can be applied. (Confidence rating = 8)

B21) The effects of moisture, temperature, and fatigue on the peel and shear strengths of adhesive in a bonded joint can be accounted for by the following approximate relation:

Sa/Sao = ((Tgw-T)/(Tgd-To))0.5 - 0.1 log N

Where:

Sa = adhesive strength corrected for environment

Sao = adhesive strength at reference conditions, usually room temperature/dry

Tgw = glass transition temperature of adhesive when wet, given by:

Tgw = (0.005 M2 - 0.1 M + 1.0) Tgd

M = moisture in adhesive in percent by weight

T = temperature of the loading environment

Tgd = glass transition temperature of adhesive when dry

To = reference temperature at which Sao was determined

N = number of loading cycles joint must sustain (Confidence rating = 6)

C1) As an aid in estimating the ability of a structure to distribute a point loading, visualize the load path as fanning out from the concentrated load at an angle no greater than 45 degrees from the centerline of the load path. Treat only the material within this fan as being effective in carrying the load.

(Confidence rating = 7)

C2) Flanges with a free edge should be no wider than ten times their thickness. (Confidence rating = 7)

C3) There is an optimum depth to length ratio for any beam type member, with shallow beams being inefficient due to inadequate separation of the caps and extremely deep beams being dominated by web weight. For uniformly loaded aluminum I beams, simply supported at each end, this optimum length to depth ratio is commonly around 9. For truss construction beams, simply supported at each end, the optimum span to depth ratio is in the vicinity of 7.

(Confidence rating = 7)

C4) For columns, there will also be an optimum geometry which, for aluminum structure, will generally be found in the vicinity of L*SQR(A/I) = 80, where L is column length, A is cross sectional area, and I is minimum section moment of inertia. For circular cross sections this works out to an optimal length to diameter ratio of about 28. Beam columns, such as a typical space frame member, should have a length to depth ratio that is intermediate between the pure beam of Rule C3 and the pure column discussed above. (Confidence rating = 7)

C5) Panel stiffeners should be sized to provide simple support along their length. This can be accomplished by making the stiffener at least an order of magnitude stiffer than that portion of the skin that would be considered as effectively working with the stiffener. (Confidence rating = 8)

C6) There are practical minimums to the gauge of sheet metals that can be used in a real world environment. For aluminum sheet, .020 is about the minimum that can be successfully formed and fastened into bracket type structures. For clips and stiffeners in a workspace environment, where they will likely be bumped, snagged with sleeves, and so forth, .062 is the minimum recommended thickness. (Confidence rating = 8)

C7) When designing doublers, (reinforcements around openings in a shell) an approximate sizing guide is to replace the removed material. In highly loaded structures or where fatigue is a concern, more reinforcement may be required. I've been told that replacing three times the removed material is the norm for commercial aircraft.

(Confidence rating = 6)

C8) For a cylinder of skin and stringer construction, optimum frame spacing is around L/D = .12 where L = frame spacing and D = cylinder diameter.

(Confidence rating = 6)

C9) For aluminum skin and stringer construction, the effective width of skin, of thickness T, working with a single flanged stiffener is approximately 20T. (Confidence rating = 8)

C10) Hollow aluminum tubes of circular cross section should have a wall thickness no less than 1/50 of the tube diameter. Below this thickness the tube wall will tend to fail by crippling. (Confidence rating = 7)

C11) When designing unpressurized semi-monocoque shells, a conservative first cut at sizing can be made by designing the stringers (with effective skin) to carry the full axial load on the shell and selecting a skin thickness that will carry the shear load. To a first approximation, the weight of the skin will equal the weight of the stringers. If the shell is pressurized, the optimized structure will have a lower percentage of weight in stiffeners. (Confidence rating = 8)

C12) When selecting an end fixity for column calculations, it is generally advisable to use C = 1 for conservatism. In practice, aerospace structures are rarely able to provide end fixities greater than 2. (Confidence rating = 9)

C13) Structures often fail due to secondary loadings such as: kick loads, prying loads, and friction loads. These loadings must be considered in the design and sizing effort. (Confidence rating = 10)

C14) Be alert for material anisotropies. Commonly encountered examples are poor short transverse properties in composite laminates and forgings.

(Confidence rating = 10)

C15) While a 0% margin is acceptable for simple load paths and well understood failure mechanisms, the prudent designer will allow additional margin on complex or poorly understood structures. Recommended "designer's factors":

joints ....................................................... 1.50

buckling critical structure ............. 1.30

(Confidence rating = 8)

C16) A well proportioned beam will have roughly two thirds of its mass in tension and compression members and one third in the connecting shear structure.

(Confidence rating = 6)

C17) Given a ring of centroidal diameter D, axially loaded at n hard points equally spaced around one edge of the ring, the minimum ring height required to evenly distribute the load over the opposite face of the ring is 1.6 D/n .

(Confidence rating = 7)

C18) When computing the vibration response of structures to transient and cyclical disturbances, an estimate must be made of the damping coefficient (as a percentage of critical damping). 1% to 2 % of critical damping is often used for predicting launch induced dynamic response of typical bolted aluminum aerospace structures.

Structures with less energy dissipation, such as welded structures, will have lower damping coefficients. Damping coefficients will also be lower for low amplitude and low frequency vibrations. Damping measured during modal survey testing normally indicates low damping due to the low vibration levels used for testing - normally 1/10 - 1/100 of the maximum predicted flight response. The Payload Assist Module - S (PAM-S) structure modal test indicated 0.27% damping at the first fundamental frequency of 9.1 Hz. The same structure had 0.58 % damping for an axial vibration mode at 24 Hz. The Delta II Graphite Epoxy Motorcase (GEM) structure supported on high dissipation mechanical links had 0.47% damping associated with a 4.5 Hz lateral bending mode and 2.1% damping associated with a 46.5 Hz lateral bending mode at modal survey excitation levels. (Confidence rating = 7)

C19) Hollow shear pins should have a wall thickness greater than 1/7 of the pin diameter to avoid local bearing failures. (Confidence rating = 7)

C20) An estimate of effective bearing stress in material around a pin may be made by dividing the shear load in the pin, P, by an area corresponding to the chord of a ninety degree included angle on the pin. Effective bearing stress = approximately P/(0.707Dt) where D is the pin diameter and t the thickness of the material on which the pin bears. (Confidence rating = 8)

D1) Begin planning a viewfoil presentation by listing the points you wish to make. Next prepare a storyboard to work out how many foils and what art are needed to make these points. Make the viewfoils as the last step. (Confidence rating = 9)

D2) A viewfoil should contain no more than five "bullets" of no more than five words each. (Confidence rating = 8)

D3) When preparing a viewfoil presentation, allow three minutes per viewfoil for the pitch, less for a high level overview, and more for a presentation to a technically oriented audience. (Confidence rating = 7)

D4) Viewfoils should be "pleasing to the eye" to avoid making a bad impression on the audience. (Confidence rating = 10)

D5) On your own copy of your pitch, highlight a path across each foil. This will remind you of points to discuss and help you identify entrance and exit points to insure a smooth flow of ideas from foil to foil. (Confidence rating = 10)

D6) Never leave a viewfoil on the projector if you are not actively using it to make a point. Abandoned foils distract the audience from your presentation.

(Confidence rating = 9)

D7) Don't flip back and forth between foils or conduct extensive searches for backup foils. Keep the pitch flowing. A ragged presentation brings out the predatory instincts in an audience. (Confidence rating = 9)

D8) Every foil should have something more than text on it. If you don't need a picture, chart, or table to make a point, you don't need that viewfoil.

(Confidence rating = 8)

D9) Don't sail through your foils so fast that the audience doesn't have a chance to read them. You will frustrate your customer and leave the impression that you are trying to slip something by him. (Confidence rating = 10)

D10) When responding to questions or criticism, try to keep cool. A calm response is more convincing than an emotional one. (Confidence rating = 9)

D11) If the original of a viewfoil can be read at a range of ten feet, the foil will be legible when projected. (Confidence rating = 7)

D12) Budget 90 minutes of preparation time for every minute of presentation. (Confidence rating = 7)

D13) In the interest of legibility, use upper case text for brief labels, but mixed case text for any string over a few words long. (Confidence rating = 8)

D14) When preparing for an important presentation by means of a dry run, include at least one person in the audience who is not a member of the project team. This will encourage the presenter to format his pitch more nearly as it will be given to the target audience. A person who is not familiar with the program and who has not bought into the assumptions that have become second nature for team members will more readily spot holes in the presentation. (Confidence rating = 7)

D15) Vary the format of your foils. A series of foils with identical format is boring and makes it more difficult for the audience to retain the points you are trying to make. Cognitive human factors research indicates that varied appearance in foils induces structure in memory for containing associated facts.

(Confidence rating = 8)

E1) When repairing a scratch or gouge caused by an impact type event, always etch and do a dye penetrant (or equivalent) inspection of the area of the defect (which will commonly be blended out to a ten to one width to depth ratio). Even a dropped wrench can start and propagate a crack in light structure.

(Confidence rating = 10)

E2) When investigating a discrepancy report, always go see the structure for yourself. Human vision occurs more in the mind than the eye and there is no substitute for informed observation. (Confidence rating = 10)

F1) For purposes of converting a known weight of an electronics package to a volume or vice versa, use an approximate density of 0.03 lbs /cubic inch. For NiCd batteries, use 0.08 lbs /cubic inch. (Confidence rating = 8)

F2) Conventional aluminum spacecraft bracketry generally weighs from 15% to 25% of the weight of the supported equipment. As an example of how far such an approximate rule can be extended, historical data indicate that cradle type structures flown on the STS Orbiter roughly follow the rule:

Ws = 1025 + .154 Wp

where Ws is structural weight in pounds and Wp is supported payload weight in pounds. (Confidence rating = 7)

F3) Table F1 is a compilation of historical data on weight estimate growth. The prudent designer will make allowance for component weight growth when designing supporting structure. (Confidence rating = 6)

F4) For composite spaceframes, joint weights can be estimated using joint penalties from the table below.

Joint Penalty Table for Launch Condition Driven Composite Spaceframes

Weight of structure less joints

(Confidence rating = 8)

F5) The following table summarizes spacecraft subsystem weight data from an Aerospace Corporation database of 16 space vehicles. (Confidence rating = 8)

F6) Pressure vessel weight can be estimated using the performance factor, K, defined as the product of tank burst pressure (psi) and volume (cu. in) divided by tank weight (lbs); K = PV/W. The table below lists theoretical performance factors as well as data from as-built hardware with weld lands, penetration bosses, and attach flanges. Values are given both for pressure dominated designs and for applications in which flight loads are relatively significant. The heavier tank walls needed to carry flight loads significantly affect the realizable performance factor so that the optimum tank material may well be different for flight load dominated vs. pressure dominated designs. (Confidence rating = 9)

F7) For seventy-five (75) spacecraft launched from 1975 to 1984, the average wet stowed density for a complete spacecraft was 0.0028 lbs/cubic inch. Extreme values for this set of spacecraft were 0.00072 lbs/cubic in. and 0.0062 lbs/cubic inch. (Confidence rating = 8)

G1) In a layup designed for production by the pultrusion process, at least 20% of the fibers must be oriented axially to allow drawing the material through the die. (Confidence rating = 9)

G2) "A basis" material properties, such as are listed in MIL-HDBK-5 for commonly used aerospace metals, are rarely available for composites. Due to the wide scatter in material properties from one composite specimen to another, it is mandatory to apply some kind of knock down factor to the average material properties values commonly given for composite materials. In the absence of better data, reduce average strength values by 30% and average stiffness values by 20% to obtain usable design properties. (Confidence rating = 7)

G3) Quasi-isotropic fiber reinforced epoxy layups have stiffness and strength on the order of 30% to 40% of the axial properties of a unidirectional layup, with the difference being more pronounced the more the fiber properties exceed those of the matrix. (Confidence rating = 8)

G4) When selecting a cross-ply layup for a high modulus graphite/epoxy composite for use in "room temperature" environments, the maximum change in fiber angle between plies should be less than 60 degrees to reduce the probability of microcracking. For example, the layup [0/+45/0/-45/0//]s is preferable to [02/+45/-45/0//]s For applications with large temperature variations, the maximum change in fiber angle should be less than 30 degrees.

(Confidence rating = 7)

G5) To insure good load transfer between plies of a laminate, try to avoid having over four layers with the same fiber orientation. (Confidence rating = 6)

G6) For components made by filament winding, fifteen degrees is the lowest off-axis angle that can be wound without extensive development. With development, and perhaps special tooling, angles as low as ten degrees may be obtained. (Confidence rating = 7)

G7) The following rules apply to the design of mechanically fastened joints through unreinforced composites.

a) The best bolted joints can barely exceed half the strength of unnotched

laminates.

b) Optimized joints with a single row of fasteners have approximately three

quarters of the strength of an optimized four row joint.

c) Fastener diameter should be selected to develop bearing strength of

laminate rather than by fastener rated shear strength.

d) Bolt bending is more significant for joints in composites than for joints in

metals because composite joints tend to be thicker for a given load and

because composites are more sensitive to nonuniform bearing stresses.

e) Optimum fastener spacing for a joint with a single row of fasteners in

composite material is about three fastener diameters. This is less than the

4D usually recommended for metals, due to the greater tendency for the

composite to fail in bearing if the spacing is so large as to put too high a

load per individual fastener.

f) Optimum fastener spacing for a joint with three rows of fasteners in

composite material is about five fastener diameters in the first row (to

minimize load transfer in this row nearest the joint), then 4D for the

second row, and 3D for the final row.

g) Best fiber patterns for composites in the area of bolted joints have at least

12 % of the plies in each of the four directions, 0°, +45°, -45°, and 90°. No

more than 38% of the plies should be in any one direction.

(Confidence rating = 8)

G8) The following rules apply to the design of adhesive bonded joints in laminated composites.

a) Bonded joints should be designed to be stronger than the adjacent

structure. A failure in a weak bonded joint can propagate catastrophically

from a local defect.

b) Bonding works best for thin structures.

c) Thick bonded structures need complex stepped-lap joints to develop

adequate efficiency. Analysis of stepped-lap joints in thick structures

requires non-linear analysis.

d) Thick structures cannot be practically repaired by bonding. If damage is

likely to be sustained during the service life of a thick composite structure,

bolted field joints should be provided to allow segment replacement.

e) Proper faying area surface preparation is a must. Beware of "cleaning"

solvents and peel plies. Mechanical abrasion is more reliable. Grit blasting is preferable to sanding for irregular surfaces such as cloth layups.

f) Laminates must be dry before bonding.

g) The key to durability of bonded joints is that some of the adhesive must

not carry sustained high loads. Creep will occur in joints where all of the

adhesive is kept under high load.

h) Bonded joints are sensitive to environmental conditions. Bonded overlaps are commonly sized by the hot/wet environment. If exposed to cold

conditions, the adhesive becomes brittle and especially susceptible to

failure in peel.

i) In a lap joint, adherend ends should be tapered to a thickness of

approximately 0.020 inches with about a ten to one (6°) slope.

j) Adherend ends should be chamfered to provide a local thickening of the

bondline around the periphery of the joint. This will reduce peak stresses

and help prevent the initiation of peeling.

k) Assuming that the primary load in a bonded joint is in the 0° direction, do not use 90° plies at joint faying surfaces. If thermal excursions are large, joint strength may be increased by using off angle plies (i.e. 45°) nearest

the bondline, rather than 0° plies, in order to reduce thermal stresses.

l) For joints between equal thickness quasi-isotropic laminates, near

optimum splice geometries are: 80t overlap for single lap joints, 30t

overlap for double lap joints, and 1/50 slope for a scarf joint, where

"t" is the thickness of the laminates being joined.

m) When designing a laminate with bonded joints, avoid layups with large

Poisson's ratio mismatches between adherends.

(Confidence rating = 8)

G9) Tight control of the thickness of a laminate typically requires special tooling, such as matched die molds. Without special tooling, thickness variations up to ten percent of the nominal part thickness can be expected. (Confidence rating = 8)

G10) Since the outer ply of a laminate is the most likely to be damaged, its orientation should be chosen to allow some damage tolerance. For example, a column member, which will tend to have its 0° fibers most heavily loaded, should use a shear carrying off axis ply (such as a 45° ply) as its outer layer. (Confidence rating = 7)

G11) The "Ten Percent Rule": When designing a lay-up, use a minimum of ten percent of the plies in each direction. This is based on the assumption that you are already following Rule G4, so that the maximum arc between reinforcement directions is about 60 degrees. (Confidence rating = 6)

G12) It is common knowledge that to avoid coupling of extensional strains with bending and twisting in a composite laminate, the layup should be both symmetrical about its midplane and balanced (each +X° ply matched with a -X° ply). An implication of this basic guidance combined with rules G4 and G11 above, is that a well designed laminate must be at least seven plies thick. (For what it's worth, there are a lot of four ply laminates in service, so for the seven ply rule, Confidence rating = 4)

G13) When adding or dropping off plies at a local reinforcement in a laminate, the step spacing should be chosen so that the surface slope angle with the shell midplane does not exceed ten degrees. (Confidence rating = 7)

G14) When doing trade studies to select material for a structure, do not neglect to include tooling costs. For short runs, the tooling cost for some types of composite construction can exceed the cost of flight units made of aluminum. (Confidence rating = 9)

G15) Stiffness and strength of a laminate made from eight harness satin cloth are about 30% lower than the equivalent values for a tape plied laminate of the same fiber and matrix. Unidirectional fabrics or pre-plied broadgoods are alternate ways of avoiding the property degradation associated with fiber kinking in woven products. (Confidence rating = 7)

H1) "The bitterness of poor quality lasts long after the sweetness of meeting the schedule has gone." (Confidence rating = 10)

H2) A design can meet any two of the following criteria at the expense of the third: 1) Good, 2) Fast, and 3) Cheap. To produce a satisfactory product, the designer should understand the relative valuation his customer places on performance, schedule, and cost. (Confidence rating = 10)

H3) "Don't let your education get in the way of your common sense." (Confidence rating = 10)

H4) The given design requirement is never the real design requirement. It is the responsibility of the designer to determine the real requirements and produce a design to meet them. (Confidence rating = 10)

H5) "If you cannot be clever, you can at least be careful." (Confidence rating = 10)

H6) "An engineer is someone who can accomplish with a dollar what any fool can do for three dollars and fifty cents." Cost is a factor in all design decisions. (Confidence rating = 10)

H7) Never do business with a company you haven't visited. (If a subcontractor has contractual freedom to sub out your work, you may well end up doing business with a company you haven't visited.) (Confidence rating = 10)

H8) In the aftermath of the Three Mile Island reactor radiation release, Admiral H. G. Rickover prepared a report recommending nuclear utility management objectives. With slight editing, these apply as well to any engineering effort:

A) Require rising standards of adequacy.

B) Be technically self-sufficient.

C) Face facts.

D) Respect even small problems.

E) Require adherence to the concept of total responsibility.

F) Develop the capacity to learn from experience.

(Confidence rating = 10)

H9) The Mountaineer's club of Seattle publishes rules for climbers, one of which is:

"Never let judgment be overcome by desire when selecting a route or deciding whether to turn back."

Suitably paraphrased, this rule is applicable to any engineering endeavor. (Confidence rating = 10)

H10) "Hope is not a method." - General Gordon Sullivan (Confidence rating = 10)

H11) "You learn something every day - if you aren't careful."

(Confidence rating = 10)

H12) To paraphrase biographer James Gleick in describing an insight of Nobel prize winning physicist Richard Feynman:

To avoid errors requires an intimate acquaintanceship with the rules of the engineer's game. It also requires not just honesty, but a sense that honesty requires exertion.

(Confidence rating = 10)

H13) Kelly Johnson, of Lockheed Skunk Works fame, offered the following "all you need to know to run a company":

A. It's more important to listen than to talk.

B. Be decisive; even a timely wrong decision is better than no decision.

C. Don't halfheartedly wound problems - kill them dead.

(Confidence rating = 10)

H14) Carl Printz often had occasion to point out that there's no point in our making the same mistakes over and over again when we could be making new and exciting mistakes! (Confidence rating = 10)

H15) "I pay you to make me look good and when I tell you to do something stupid, you’re supposed to be smart enough not to do it!"

- Chief engineer to lead structures designer immediately after design change ordered by said chief engineer caused failure resulting in destruction of structural qualification test article (Confidence rating = 10)

H16) Not-designed-here-syndrome is a common ailment amongst designers of all stripes. The best designer I’ve ever known offered the following preventative measures:

a) Circulate your ideas widely among known good designers. If they point out fatal defects in your brainchild, you can just ditch the ugly baby quietly. Most people have short memories...

b) Listen with open mind to all ideas proffered to you, especially if they come up repeatedly from various plausible sources. Meticulously attribute the good ideas to their originators the first time you cite them. Fold them into your work.

c) As time passes and people comment that your designs work uncannily well, accept their praise gracefully. Your glowing reputation will be truly deserved, because nurturing the humility to be able to accept the other guy’s idea when it is better is, in fact, one of the keys to becoming a great designer. (Confidence rating = 10)

H17) In laying off a young fellow from his lab, Thomas Alva Edison gave the following reason: "I don’t mind the fact that you don’t know much yet. The trouble is that you don’t even suspect." The good engineer aims to be knowledgeable, but even the budding engineer should be intelligently suspicious.

(Confidence rating = 10)

J1) The United States Air Force Unmanned Spacecraft Cost Model, Version 5.0, gives the following cost estimation relationships for structural/mechanical subsystems with a total weight ranging from 16 to 942 lbs. These formulas are inappropriate for estimating component costs. The formulas have been expressed in 1990 kilodollars (K$).

Nonrecurring Cost = 2468 + (232*W0.66) K$

Recurring Cost = 48.5*W0.65 K$

where W is the total structural/mechanical subsystem weight.

(Confidence rating = 3)

J2) Jerry Fish, of Cost Estimating, provided the formulas in Table J1 for estimating the costs of large structural assemblies for the Space Station program. All formulas have been corrected to 1990 dollars. (Confidence rating = 4)

J3) The estimated cost of a spacecraft usually doubles between initial contract award and delivery of the first article. (Confidence rating = 2)

J4) The average engineer underestimates by a factor of two the manhours it will take him to accomplish a design job of the order of complexity of a Delta payload attach fitting. This assumes that the job has been broken down to some level of detail and each portion of the job estimated separately. An offhand guess will generally produce an even more optimistic estimate. Estimation accuracy will be better for a smaller job and worse for a larger one. (Confidence rating = 4)

J5) The cost of a finished composite spaceframe strut with metal end fittings is approximately three times the cost of the raw materials. (Confidence rating = 2)

E-glass/epoxy 10 $/lb

S-glass/epoxy 20 $/lb

Kevlar/epoxy 30 $/lb

T300 graphite/epoxy 60 $/lb

T300 graphite/toughened epoxy 100 $/lb

IM7 graphite/epoxy 75 $/lb

IM7 graphite/toughened epoxy 125 $/lb

P75 graphite/toughened epoxy 350 $/lb

FT700 graphite/toughened epoxy 550 $/lb

P100 graphite/toughened epoxy 800 $/lb

P120 graphite/toughened epoxy 850 $/lb

J7) On a typical aerospace project, the costs for engineering, planning, and quality assurance will be roughly equal. The manufacturing cost will be 250% of the engineering cost. (Confidence rating = 4)

J8) On a typical missile design effort, the hours charged by various groups as a percentage of Structures Design group hours are:

Structural Analysis group - 75%

Structural Dynamics group - 40%

Mass Properties group - 25%

Obviously, these proportions can vary considerably with the nature of the project. Composite structures typically require a higher proportion of analysis support. (Confidence rating = 4)

J9) Below is an estimate of the average hours allotted to various structures design tasks per drawing (twelve zone "J" size). The time period covered is from contract award through completion of a first flight item. This estimate was prepared in 1991 and is based on the use of Unigraphics Version 7.0 for drawing preparation. The concurrent engineering estimate includes coordination with analysis and other engineering groups as well as Operations disciplines. (Confidence rating = 3)

Task Hours/drawing

Concurrent engineering ................ 20

Layout & trade studies ............... 180

Detail design ...................................... 30

Drawing preparation .......................45

Technical check ................................ 25

Drawing maintenance .................... 30

Manufacturing support ................. 20

Total .................................................... 350

J10) Total program costs for one-of-a-kind mid-sized satellites are known to vary over the range of $8K/lb for modest modifications of an existing design to $36K/lb for technically ambitious projects. The value $20K/lb in 1991 dollars has been used as something of a standard for preliminary estimates. (Confidence rating = 5)

J11) Each level of security classification increases program cost by a factor of 1.25 times the proportion of the program data classified at that level. Example: If a program's data is 20% Secret and 5% Top Secret, the program cost factor will be approximately (0.8)(1.0)+(0.2)(1.25)+(0.05)(1.25)2 = 1.13; that is the program will be about 13% more expensive than if it were a completely unclassified program to produce the equivalent hardware. (Confidence rating = 3)

J12) Per a list published in "Space Mission Analysis and Design", small satellites (below 150 lbs) range in cost between $6.5K/lb and $20.2K/lb in 1990 dollars.

(Confidence rating = 8)

K1) The Fog Index, F, of a piece of writing is given by the formula:

F = 0.4(W+S)

where:

W is the average number of words per sentence

S is the number of words per hundred having three or more syllables

The Fog Index can be thought of as the reading age required to comprehend the text. Try to keep the Fog Index of your reports below 15.

Examples:

"Big Two-Hearted River" - Hemingway ....................................................5.9

"Microsoft Excel Reference Manual" ........................................................ 14.3

"Rules of Thumb for Structural Design" ................................................. 16.8 (Hmmm...)

"Tools and Approaches for Total Quality Management" ................. 19.7

(Confidence rating = 6)

K2) In engineering analysis documentation it is desirable to:

• Explicitly state assumptions

• Define symbols used in formulae

• Use sketches and figures where required for clarity

• Highlight results and conclusions by shadowboxing them as they occur in the text or by putting them in a summary paragraph.

(Confidence rating = 8)

K3) Minimize references to other documents, especially ones that may be difficult for the reader to obtain. Information vital to making your point is worth quoting or paraphrasing. (Confidence rating = 8)

K4) Minimize your use of acronyms. Always define an acronym at first use in a document. In a large document with many acronyms, provide an acronym table.

(Confidence rating = 9)

K5) If you have to look up the definition of a word, then you probably shouldn't use it in your writing. (Confidence rating = 9)

K6) If you feel that a point is too obvious to state, better state it anyway. Your readers may not share your assumptions or your familiarity with your subject. (Confidence rating = 8)

K7) Budget four hours per page for writing a technical article. Of this, one hour is to produce the rough draft and the remaining three for editing and rewriting.

(Confidence rating = 6)

L1) Write the documentation first, then write the program to match the documentation. This provides a clear requirements statement to serve as a guide during the actual coding, and incidentally guarantees that at least some form of documentation will be available to go with the software. (Confidence rating = 6)

L2) When estimating time required to write a program, figure on about a five to one ratio for debugging vs. coding. (Confidence rating = 7)

M1) Salience Bias Asked to estimate the proportion of events in a sample, people make good estimates on mid-range proportions, say from 10% frequency to 90% frequency, but will over report very infrequent events (less than 10% frequency) and underreport very common events (greater than 90% frequency). The net effect is to overemphasize the occurrence of rare events or to bias reporting to emphasize exceptional (and therefore salient) events. By a similar mental process people tend to estimate the average of a set of numbers as about half way between the highest and lowest values noted, regardless of the dispersion of intervening values. Extremes and unusual occurrences stick in memory, whereas common experiences don't. In another example, on a control panel, data that is brightly lit, displayed at high contrast, changes rapidly, or is at the top center of the display is more likely to be taken into account in decision making than equally important data that does not do as good a job of catching the eye. Personal experience is more salient and likely to influence decision making than equally relevant data that was obtained second-hand.

Prior to making a decision every effort should be made to review the underlying data in a format that puts facts on an even footing.

(Confidence rating = 9)

M2) Linearization Bias Asked to extrapolate trends from past data, people tend to err towards a linear extrapolation of the trend curve taken from the present moment.

Predictive software can be used to provide more sophisticated trend extrapolation data. (Confidence rating = 8)

M3) Anchoring and Recency People tend to form working conclusions based on the first few pieces of data they receive and then gradually modify these initial impressions based on subsequent data. People also tend to overweight recently received data in comparison with older data. Thus conclusions reached can be dependent upon the temporal sequence in which data is presented.

Decision forming data that is available simultaneously should be presented simultaneously. (Confidence rating = 9)

M4) As-If Heuristic People tend to treat all information as if it were equally reliable. People are quite capable of estimating the reliability of information sources directly; however, even if information is known to be of poor quality, if it is available in great quantity it will tend to drive decisions. Five dubious arguments tend to be more convincing than two solid and sound ones.

Decision support software is becoming available to allow weighting of information on the basis of operator provided reliability judgments. Programs like these offer the promise of more reasonable decision making than can be accomplished by the unaided human brain. (Confidence rating = 9)

M5) Confirmation Bias Once a person has formed a hypothesis, he will tend to interpret the available data to support this hypothesis and ignore disconfirming evidence. Data will be sought that tends to confirm the initial hypothesis but data which might disprove the hypothesis will be subconsciously avoided. This tendency towards "cognitive tunneling" is exacerbated by stress and schedule pressure. Even a partial solution of a problem may lead to protectiveness and inability to consider alternatives that may lead to a better solution of the overall problem.

Some work has been done with expert system architectures to combat confirmation bias, but ready availability of this type of aid is probably far in the future. Good trade study methodology, which stresses keeping options open and comparing options quantitatively and objectively, will help avoid this trap. Other precautions include good design teamwork and conscious adoption of alternating roles of advocacy and criticism for each design option considered.

(Confidence rating = 9)

M6) Positive Evidence Bias Explicitly presented information (data-on-a-platter) is more readily used to test hypotheses than equally available data of a negative nature. In a famous literary example, Sherlock Holmes solved a murder case by noting, as the inept constabulary had not, that the victim's dog had not barked. Absence of anomalies may be useful in tracing the branches of a fault tree. Another aspect of the same bias is the "puzzle trap". People tend to look for a set of rules to govern their search for problem solutions even if these rules are not imposed by nature. This leads to consideration of an unnecessarily limited subset of available problem solutions while neglecting part of the available solution space.

In a laboratory situation, training involving practice in the application of non-cued data has been shown to increase the trainee's tendency to search for and utilize this type of information. (Confidence rating = 8)

M7) Framing Sensitivity People will choose different problem solutions to identical problems depending on how the problem is stated. If the problem is presented as a choice between potential gains or benefits, people tend to be risk adverse and will select the low gain, but sure-thing, option. If the same situation is viewed as a choice between potential losses or costs, people tend to be risk-seeking and will select the low probability, high penalty option, attempting to avoid the more certain loss.

The decision maker should seek a balanced presentation of cost vs. benefit to avoid biased framing. (Confidence rating = 7)

M8) Miscategorization Given a large number of alternatives to choose from, people tend to prune the number of choices by focusing on a single selection criterion and eliminating all but the few choices that score best on that one criterion. This approach may lead to early elimination of options that may be preferable when given more balanced evaluation.

More broadly speaking, incorrectly categorizing problems or solutions can then activate an inappropriate mental model associated with the category.

Decision support software is under development to relieve the human decision maker of the stress of trying to keep too many factors in mind at once. Decision making aids can assist the decision maker in breaking the decision task down into chunks of a humanly manageable size. (Confidence rating = 9)

M9) Satisficing Another heuristic used for dealing with many options is to mentally set a failure threshold for several decision criteria and accept the first option that is not eliminated by one or more criteria not meeting the desired threshold value. This approach leads to choice of an option that is "good enough" rather than optimal.

As discussed under rule M9, decision support software or other methodology may be required to decompose a complex decision making task into subproblems simple enough to be handled within the memory and attention limits of the human mind. (Confidence rating = 8)

M10) Overconfidence People tend to be unduly confident in the correctness of their judgments, in the accuracy of their memories and, in fact, in their abilities in general.

Training to raise awareness of this tendency has been shown to have a corrective effect. (Confidence rating = 11)

1) Formulas for Stress and Strain

R.J. Roark and W.C. Young

Invaluable "cookbook" guide to structural analysis, generally accurate.

Available from: McGraw Hill (800) 262-4729

Approximate cost: $50.00

IBSN 0-07-053031-9

2) Formulas for Natural Frequency and Mode Shape

R.D. Blevins

A guide to structural dynamics in the tradition of Roark, generally accurate.

Available from: Krieger (407) 724-9542

Approximate cost: $43.50

IBSN 0-89874-791-0

3) Structural Analysis of Shells

E.H. Baker, L. Kovalevsky, and F.L. Rish

A guide to the analysis of shells, with summary charts for stress, deflection, and critical buckling load computations for various geometries, generally accurate.

Available from: Krieger (407) 724-9542

Approximate cost: $47.50

IBSN 0-89874-118-1

4) Aircraft Structures

D.J. Peery

Clear presentation of practical hand analysis techniques for flight structures.

Available from: McGraw Hill (800) 262-4729

Approximate cost: $52.95

IBSN 0-07-049196-8

5) Analysis and Design of Flight Vehicle Structures

E.F. Bruhn

Extensive and wide ranging discussion of analysis techniques for flight structures with instructive derivations, poor accuracy.

Available from: Jacobs Publishing Inc.

[via Opamp Books (213) 464-4322]

Approximate cost: $64.44

6) Missile Structures - Analysis and Design

Orlando, Meyers, and Bruhn

Not commercially available - MDAC library has a copy.

1) Engineering Mathematics Handbook

J.J. Tuma

Excellent collection of tabulated mathematical formulas, generally accurate.

Available from: McGraw Hill (800) 262-4729

Approximate cost: $44.50

IBSN 0-07-065443-3

1) Mechanisms, Linkages, and Mechanical Controls

N.P. Chironis

A collection of mechanisms, grouped by function, for those times when you need to pinch hit as a mechanical engineer.

Available from: McGraw Hill (800) 262-4729

Approximate cost: $23.50

IBSN 07-010775-0

1) Standard Handbook for Mechanical Engineers

T. Baumeister and L.S. Marks

Introductory level material on a wide range of engineering disciplines, useful in coping with subjects outside one's immediate area of specialization.

Available from: McGraw Hill (800) 262-4729

IBSN 0-07-004122-9

2) Design of Geosynchronous Spacecraft

B. N. Agrawal

Introductory level material on common spacecraft subsystems,

useful for rough preliminary design of spacecraft during proposal activity.

Available from: Prentice-Hall

IBSN 0-13-200114-4

3) Space Vehicle Design

M.D. Griffin and J.R. French

Introductory level material on space vehicle design. Offers some insight into the spacecraft design process from the customer's viewpoint.

Available from: American Institute of Aeronautics and Astronautics

Approximate cost: $56.00

IBSN 0-930403-90-8

4) Space Mission Analysis and Design

W.J. Larson and J.R. Wertz

Overview of space vehicle design. A good introduction to the basics of the

various subsystems making up a spacecraft. Contains useful tables of data on

existing spacecraft and spacecraft components.

Available from: Microcosm, Inc.

Approximate cost: $44.75 (pb)

IBSN 1-881883-01-9 (pb)

IBSN 0-7923-1998-2 (hb)

1) Introduction to Composite Materials

S.W. Tsai and H.T. Hahn

Standard text for introductory composite courses. Intended more for college students than working engineers, but contains useful basic material.

Available from: Technomic Publishing Company

IBSN 0-87762-288-4

2) Mechanics of Composite Materials

R.M. Jones

Standard introductory level text. Academic rather than design oriented, but a useful reference on micromechanics.

Available from: Hemisphere Publishing Corporation

IBSN 0-89116-490-1

3) SDS Spacecraft Structural Composite Materials Selection Guide

Extensive materials data, test methods, and design guidelines to assist the designer working on Strategic Defense System (SDS) satellites.

Prepared by Ketema Inc. for the Air Force Materials Laboratory

M.L. Hand and J.J Tracy have copies of release 2.0

4) Composite Airframe Structures

M.C.Y. Niu

Although written in broken English and generously sprinkled with typos, this

is by far the most useful guide I've seen for composite structure design

practice. This guide contains much practical information based on the

author's 25 years of experience working with composites, primarily at

Lockheed.

Available from: Technical Book Company, Los Angeles, CA (310) 475-5711

IBSN 962-7128-06-6

1) International Reference Guide to Space Launch Systems

S.J. Isakowitz

Basic information on most of the launch vehicles in the world: capability, fairing sizes, imposed loads, payload stiffness requirements, etc.

Available from: American Institute of Aeronautics and Astronautics

Approximate cost: $30.00 (paperback)

IBSN 1-56347-002-0

1) Why Buildings Fall Down

M. Levy and M. Salvadori

Although dealing with civil structures, this book is highly recommended

reading as it probes the psychology behind design errors that lead to

structural failures. Failures in aerospace structures are caused by precisely

the same sorts of flawed thinking that lead to civil structure failures. This

book presents some design practices to guard against common human errors.

Available from: W. W. Norton & Company

Approximate cost: $24.95

IBSN 0-393-03356-2

2) To Engineer is Human

Henry Petroski

Subtitled "The Role of Failure in Successful Design". A collection of structural disasters and study of how the lessons drawn from them allowed development of superior designs.

Available from: St. Martin's Press

Approximate cost: $16.95

IBSN 0-312-80680-9

3) How Designers Think

Brian Lawson

This book is a general overview of the design process providing an introduction to a variety of cognitive strategies for attacking design problems. The author's architectural background noticeably flavors the book, but some material of interest to any designer will be found. The chapter on "Design Tactics and Traps" is of particular interest.

Available from: Eastview Editions, Inc.

Approximate cost: $30.00

IBSN 0-89860-047-2

1) Engineering Psychology and Human Performance

Christopher D Wickens

Most complete reference I've found on inherent weaknesses in human

reasoning abilities. Contains some material on decision making aids.

Available from: HarperCollins Publishers Inc.

Approximate cost: $67.00

IBSN 0-673-46161-0

2) Fluid Concepts and Creative Analogies

Douglas Hofstadter and the Fluid Analogies Research Group

Report on recent research using computer modeling to probe for the wellsprings of human creativity (and tendencies to make mistakes - the two are intimately related).

Available from: BasicBooks, a division of HarperCollins Publishers Inc.

Approximate cost: $30.00

IBSN 0-465-05154-5

3) The Society of Mind

Marvin Minsky

I'll admit that this book can in no way, shape or form be considered an engineering handbook. It is a piece of the foundation for the way engineering will be done in the near future with individual engineers forming part of a web of combined man/machine intelligence. Part of the promise of artificial intelligence research is a better understanding of how "natural" minds function. Minsky describes how that promise is being realized.

Available from: Simon and Schuster Inc.

Approximate cost: $9.95 (paperback)

IBSN 0-671-60740-5