Parametric modelling 3
Last updated
Last updated
Up to now, we have been working on a design scheme i.e. a family of designs, making it more and more general and flexible. We have not however, considered at all how one design option may differ from another, other than perhaps visually.
The goal here is to incorporate in our model some quantitative design metrics. We wish to express some design performance characteristics that may help us not only differentiate between options but potentially inform our decision making.
Let’s say we did market research and determined that stainless steel round hollow sections come in linear stock sizes of 15 meters at $300 / meter. Certainly we can build longer spans by welding but it is costly. Having too many different beam sizes is also undesirable because it implies lots of cutting and resource waste.
To understand the various beam sizes we use here a “bar graph” component which renders a histogram of the edge lengths. The way this is done is by first computing the smallest (say 5m) and largest elements (say 20m). We then divide the range (20 – 5=15m) into a number of equally spaced bins (say 10 bins of 1.5 units width each). Finally, we assign every beam into its nearest bin and accumulate the number of elements in it. The histogram shows us the number of beams per size-type, and the percentage of each type in terms of the total. It is a good visual indication of how much variability we have in our design.
The “gradient” component converts a sequence of numbers into colours using the linear interpolation scheme we discussed earlier. Using the “custom preview” component we can visually inspect the color-coded version of our centreline geometry. We can adjust now our design parameters while the ensuring we do not have beams above the maximum threshold.
Finally, we can sum the total length of the structure and compute its material cost. Of course this does not include volume discounts, material transport or fabrication and processing. But in similar manner, if we determine the mass per meter we can also compute the total tonnage of the bridge as well as other key performance indicators. Structural performance is a very important topic especially for bridges, but you will have to wait till pillar years for this.
Creating a parametric model results often into multiple design input parameters scattered all over the graph. A good practice is to collect all those together on one side and create a control panel group. It is also advisable to group the various logical processing tasks into semantic regions e.g. control, spine, points, centrelines, dress-up geometry, performance analytics etc.
After this short maintenance tasks we suspend updating the design logic and focus exclusively on investigating the effects of control parameters to the design output. A parametric model creates a multi-dimensional space of designs, with each parameter representing one dimension. The model thus spans a “design space” with the content observed on screen being one instance, or a single point in this space.
The objective of exploring various combinations of design control parameters is to: (a) understand the interactions between parameters; (b) to observe their severity or sensitivity in producing very similar or wildly different results; (c) to identify unique instances and families of designs with similar characteristics; (d) to train the eye in appreciating subtleties and aesthetic nuances; (e) to catch logical mistakes and verify assumptions; (f) to test extreme cases and boundary conditions; (f) to capture some unforeseen, unexpected and exotic design species.
It is a good practice while performing this exploration, to methodically document the results. There are many ways to approach this: (a) change one variable at the time, (b) randomly choose parameters, (c) fine tune a design then restart avoiding previously explored outcomes etc. To find interesting combinations you may (a) Use extreme values for ranged parameters; (b) Test the limits of the resolution available, (c) Think of geometries and/or values that may break some parts of the logic. The goal is to have fun and toy with serendipity.