Voices of Science

By this time one had the little Hewlett Packard calculators which had little memory cards you could stick in and you could actually write your own little programmes for doing calculations on a tiny little handheld calculator. So we’d gone from having no calculators to having a calculator that would just add and multiple and divide numbers that was about a metre square, which was the first calculator we ever had in the office which I guess must have been about 1972. To then having these little Hewlett Packard things that you could programme yourself that were quite small and amazing, to having mainframe computers to do calculations for you. All that happened in this period between 1970 and 1980. And I suppose by the time I was on site I was probably using a little Hewlett Packard calculator to do the calculations on these rams, but if I needed to check something out I could – I could use the company’s mainframe computer in London and send down some FORTRAN programme sheets. So most of the work we did then was not standardised calculation programmes, you actually had to write the FORTRAN yourself to actually do every single analysis of a bridge model or something. So you had to, you know, you actually had to know the programme language and how to punch cards and all that sort of stuff, so that was yet another thing you had to learn. And it was trial and error ‘cause the first time you – I – probably up to about the fifth time you did it there was always something wrong, you know, a card was punched wrong or you’d missed out a dash or a slash or a star or something in the programming. We had programme manuals to explain how to do it which were quite good but nevertheless you seemed to have to have two or three goes at it before you got the thing to run properly. I think the big problem with when we got to the stage where a bridge could be analysed with a – we used to call a grid model or a plane stress plate model was that people started to use those models to predict the performance of bridges, people started to use them as black boxes. And I can still remember to this day a particular example where a young engineer had analysed a bridge and gone all through the detailed design of a bridge and had come up with solutions which weren’t quite right, they didn’t look quite right, and so I questioned it and went back and found that there was a fundamental error in the original coding of some of the members of the bridge, which had never ever been questioned and so this whole design had got to quite an advanced stage with a fundamental error in the original model that could have been picked up if somebody had – it was a factory of ten in a stiffness, if somebody had really thought they would have picked this up but they didn’t, and it always worried me that people tended to use computer models as black boxes. I worried that engineers would lose the ability to think at a fundamental level about structural behaviour.