William M. Bulleit

Michigan Tech

Uncertainty in the Design of Non-prototypical Engineered Systems

ConceptDesignPrototype – with feedback to designProductionQA & Testing

(Element 14, Journal 1)

Product Development Cycle Electronic Products

ConceptDesignConstruction – feedback to design mostly

changes, not necessarily improvements

Non-prototypical Systems

Aleatory Of or related to chanceUncertainty generally not reduced by

increased knowledgeFlipping a coin - frequentist or subjective

EpistemicOf or related to lack of knowledgeUncertainty generally reduced by increased

knowledgeFlipping a coin - physics

Types of Uncertainty

Time – past and futureStatistical limits – never enough dataRandomness – nothing is one numberHuman error – screw ups happen

Sources of Uncertainty - Basic

Use changesPredict future loads based on past loadsDeteriorationIncreased time causes increased probability

of extreme load

Time

Only can take so many samples of anythingWe only have about a 100 years of load dataNever sure if the sample represents the

population

Statistical Limits

Seismic ground motions are random processes

Wind pressure is a random processCross sectional dimensions varyLive load varies spatially

Randomness

“To err is human, to anticipate is design.”Anonymous

“Good judgment comes from experience, and experience comes from bad judgment.”

Attributed to Mark Twain

Design

Modeling – simplifications or misconceptionsContingency – it does not existInconsistent crudeness – one refined, one

not…Code complexity – what to choose?

Sources of Uncertainty - Design

Occupancy live load is assumed to be uniformly distributed

Wind load is assumed to be staticLoad variability is assumed to be

representative of load effect variabilityStrain distribution assumed to be linear

Modeling

“I am persuaded that many more failures of foundations or earth structures occur because a potential problem has been overlooked than because the problem has been recognized but incorrectly or imprecisely solved.”

Ralph B. Peck

Human Error/Modeling Error

Tacoma Narrows

Contingent: dependent on something not yet certain.

In engineering design contingency refers to the need to visualize a system and perform analysis and design on the envisioned system before it can be built. (Scientists typically analyze existing systems.)

[H. Simon, The Sciences of the Artificial]

Contingency increases uncertainty

Contingency

Engineers’ designs are not consistently crude.

Some portions of a code are well researched and based on engineering science, and some have been in the code for decades (EFW for concrete T-beams).

Inconsistent Crudeness

“A heuristic is anything that provides a plausible aid or direction in the solution of a problem but is in the final analysis unjustified, incapable of justification, and potentially fallible.”

B. V. Koen, Discussion of the Method

Heuristic

We use them to help solve problems and perform designs that would otherwise be intractable or too expensive to perform.

Ex. 1: 0.2% offset method gives the yield stress of the steel.

Ex. 2: The dynamics of the wind load can be ignored in the design of buildings.

Ex. 3: Occupancy live load is uniformly distributed.

Heuristics

Use characteristic values (e.g., 5th percentile)

Use design provisions that have stood the test of time, but update as necessary (possibly due to failures)

Check designs and inspect construction (Quality control)

Make appropriately conservative assumptions in analysis (What is appropriate?)

Dealing with Uncertainty

Check complex analyses with simpler methods where possible.

Use your own experience.Recognize that heuristics are used in all

engineering design and think about their limits

Dealing with Uncertainty (Cont.)

“The person who insists on seeing with perfect clearness before deciding, never decides.”

Henri F. Amiel

“Choosing not to decide is a decision.”Anonymous

Decisions

Reflection by the engineer on a design may be a way to enhance future similar designs

Reflection may also work as a type of feedback (e.g., Citicorp Building, 1978, William Le Messurier)

Reflection

Prototypical versus non-prototypical systems are distinguished by the amount and timing of feedback

Design of prototypical systems involves relatively rapid feedback during design and more feedback during operation (e.g., automobiles, computers, light bulbs)

Non-prototypical systems receive essentially no feedback during design, and only slow feedback during their life (e.g., Tacoma Narrows, Deepwater Horizon)

Time and Again

Low probability – high consequence eventsBlack swan eventsHuman/societal limitations

Conclusion

Questions?