Testing before building
Before computers, engineers tested designs mostly in two ways: by calculation, using simplified formulas and slide rules, and by building physical models or prototypes and trying them out. Both had limits. Hand calculation worked well for simple shapes such as beams or pipes but became impractical for complicated structures. Prototypes were reliable, but expensive and slow, especially for aircraft, bridges or large machines.
Computer-aided engineering, usually shortened to CAE, grew from the idea that a computer could do more of this testing numerically. The term covers a family of activities, mainly analysis of how a design behaves under load, heat, fluid flow or electrical signals. This article focuses on the tools and the people who used them rather than the underlying mathematics.
There was no single founder
The field developed in several places at once, often tied to the needs of a particular industry. Aerospace was a major driver, because aircraft structures had to be both light and safe, and failures were costly. Nuclear and defense laboratories, which had some of the most powerful computers of the time, developed methods for physical problems. Electronics, civil engineering and automotive firms each followed their own paths.
For that reason, any statement that one person or program created CAE should be treated cautiously. It is more accurate to speak of a number of groups working on related ideas in the 1950s and 1960s.
Structural analysis and the finite element approach
One of the most important developments was the finite element method. The core idea is to divide a complicated structure into many small, simple pieces, write down how each piece behaves, and combine the equations so the whole structure can be solved. Engineers at aerospace firms and universities developed early forms of this approach in the 1950s, with a well-known paper from a team including Boeing engineers appearing in the middle of that decade. The name “finite element” was applied somewhat later, around 1960.
The method suited computers well. It produced large systems of equations, which are tedious by hand but suited to repeated machine arithmetic. As computers grew faster and memory increased, models grew from a few dozen elements to many more, though the numbers involved in particular projects varied widely.
General-purpose programs
At first, engineers often wrote their own programs for specific jobs. Over time, organizations began building general-purpose analysis programs that could be applied to many structures without rewriting the code. One well-known example is NASTRAN, developed under the sponsorship of the U.S. space agency NASA in the late 1960s and made available to outside users in the early 1970s. Several commercial and university programs emerged around the same period.
The value of such programs was reuse. An organization could invest in one validated body of code and employ it across projects, and results from different groups could be compared more easily. Government sponsorship mattered, since large programs were costly to build, and sharing them was one way to spread the cost.
Circuit and system analysis
CAE was not limited to mechanical structures. As electronic circuits grew more complex, hand analysis became hard, especially once integrated circuits arrived and individual transistors could no longer be probed on a breadboard. Circuit simulation programs developed at universities in the 1960s and 1970s allowed engineers to predict how a circuit would behave before manufacture. This line of work grew into the electronic design automation tools that are essential to chip design today, though that story is a large one of its own.
How the work was done
In the 1960s and 1970s, a typical analysis job followed a pattern that now seems slow.
- An engineer or analyst prepared a description of the structure, often as punched cards, listing points, connections, materials and loads.
- The deck was submitted to one of the mainframe computers of the day, which might be shared with many other users.
- Results returned as printed tables, sometimes hours or days later.
- The analyst studied the numbers, looked for mistakes in the input, and ran again.
Because the process took time and expertise, analysis was commonly carried out by specialists, separate from the designers. Preparing the model could take far longer than running it. Computer graphics tools, which later displayed meshes and colored stress plots, were gradually added so that people could check input and understand output.
What it changed and what it did not
CAE enabled engineers to examine more design variations, find problems earlier and reduce, though not eliminate, the number of physical tests. In industries where tests are costly, such as aerospace, this was a significant advantage.
It also had limits that early users learned the hard way. A computer solves the model it is given, not necessarily the real object. If the model leaves out an important effect, uses a poor assumption or is given wrong input, the answer may look precise and still be wrong. For this reason, physical testing continued alongside analysis, and engineering judgment remained essential.
Closer to design
By the 1980s, as CAD and analysis programs improved, the two began to share geometry, so that an analyst did not have to rebuild the model from scratch. That integration is a recurring theme: the more closely design and analysis are linked, the more quickly a design can be refined, and the more important it becomes to understand what the analysis does and does not show.