Computers are both overused and underused. First we will discuss analysis programs because that area probably gets the most computer use. Highly sophisticated programs such as frame analysis or the finite element method, require a huge volume of input and output, and consequently a large amount of time and consequent expense. For that reason, they should be used only to confirm a design. Preliminary and intermediate methods should be used to set the member sizes and to discover other problems. Then it will be necessary to make only a few runs to get the information to complete the final design. In this way the preliminary analysis can be used to check the sophisticated computer analysis.

Otherwise there may be huge piles of computer output for each of the many, successive runs. only a few values in these listings of results will have any meaning. If many runs are required, then you probably do not understand the structural system.

Another fault that I have observed, is that many load cases are run when only one or two are really significant. To evaluate the data produced, takes time that should otherwise be used to study the structural integrity of the design, and develop strong and economical connections, usually the most vulnerable part of a design.

Computer calculations are sometimes required to confirm a decision in the early stages of design. A case in point is the design of tall buildings where the lateral deflection of the structure must be studied so the structural system can be selected. In this case, it is often not necessary to use the full model, but rather a simple model can be devised having only a fraction of the number of members, so consequently it will be easier to code and to run. often symmetry or antisymmetry can be used to reduce the size of building* frames. These are all techniques that must be studied and devised by the computer user.

How accurate must an analysis be? What inconsistencies between the analysis and the member sizes can be tolerated for the final design? Actually, in many cases, considerable deviation is acceptable if the total capacity of an indeterminate frame is satisfied. Witness the acceptance of plastic design of steel frames. Young engineers need constant supervision of the use of analysis programs. it is possible to get a closed loop where the difference between member sizes supplied for the analysis never converges to those required by the forces found in the analysis.

For analysis, the preparation of computer input, and the timing of computer runs requires a considerable expertise, and novice engineers should be trained carefully in order to get the maximum efficiency from the use of computers in a structural office.

The use of design programs also has problems, though not as aggravated because it may not be necessary to make repeated runs when the right guesses are not made. How automated should design programs be? Should the computer furnish a the completed design or should the program be a set of tools that can be orchestrated by the designer as required? There are many design problems that can not be automated but must be solved step by step. In the past these have required tedious manual calculations.

My opinion is that you need both types of design programs. However, you should not have short routines on large computers. Rather they should be on small computers that are on the desk of each engineer. I believe that this area of computing has not been fully exploited in structural engineering offices, probably because small computers are relatively new, it takes time to develop programs, and many firms are essentially conservative. It will take effort to provide a large enough number of suitable programs. Each engineer has a different way of solving problems, and it is difficult to write programs that will satisfy everyone.

In the future, the basic design and analysis routines could be developed by each engineering student in school, and they could carry these programs to the job. Then programs can be a personal part of the engineers equipment just as are his school books.

Computers are only as useful as their accessibility. my concept of computer usage is that there should be a hierarchy of computers:

1. Programmable calculators and pocket computers.

The new, very small computers with dot matrix LCD readout are now much cheaper than programmable calculators and are very easy to program in BASIC. Also they may have built?in tape recorders and printers. There should be one of these on each desk. These computers may have limited memory but small analysis and design programs usually do not take much space. Is hard copy is necessary with small computers? Hard copy takes extra equipment, longer programs and more time to run a program. My own point of view is that it is better merely to record the results directly in the calculations. (Where they should be anyway). It is very easy to justify the cost of these small computers, because the equipment is only a small fraction of the salary of the engineer.

2. Microcomputers of the Apple, Radio Shack, IBM PC variety.

These computers will actually run very large and sophisticated programs but they will also run small task oriented programs that allow you to do bits and pieces of a design. For large analysis programs they can be very slow, especially for the print out. There could be, say, one of these computers to every three engineers. With the cost of printers so low, there should be a printer of each machine.

3. Main frame computers.

They are necessary to run the sophisticated analysis programs and large scale design programs that take much output. Not every structural office can afford this type of equipment, but microcomputers can be used as terminals for commercial time sharing systems.

In summary, more emphasis should be give by consulting firms on the use of very small computers. For example, it is senseless to do complicated shear and moment diagrams requiring 30 minutes to solve and to check, when a small hand held computer will give an answer in several minutes. There is much to be learned about computer usage.