::000_A::
//* {Parametric Bronze Floor Castings}::
*//{2010.12.08 ~ Professional Work : Loisos & Ubbelohde}

Shireen is a 51' Frederick Shepherd designed ketch wooden boat, built by Camper-Nicholson Yachts in the south of England 1934. Shireen's teak planking on oak frames have twice been around the world, and across the Atlantic 9 times. I was actively engaged in her restoration from June of 2009 to January 2011.

One of the larger components of Shireen's restoration was the bronze casting of 36 different structural members along the keel, known as floor members. Traditionally these members were individually wrought of iron to fit the boat, and then hot-dip galvanized to prevent corrosion. Over time, the salty air and pooling bilge water wasted away the iron, which is traditionally replaced with bronze during restoration. The task at hand was to cast 36 new bronze members to replace the iron components, no two of which are the same.

The traditional process for sand casting requires making a positive pattern for each component, against which foundry sand is packed creating a negative, compressed-sand mold to pour the molten metal into. The owners had previously completed four of these components, and each pattern took them almost a month to complete, and then cost over 2000$ to have a foundry make the molds, and cast the bronze.

This traditional patternmaking process consists of measuring the boat precisely for the final desired dimensions, scaling these dimensions up to accommodate the contraction allowance (or shrinkage bronze incurs when it solidifies), and then building/carving/shaping/sanding/painting the pattern to make a 3d positive model of the desired component. This pattern would then be handed to the foundry, and they would begin a process to determine the best way to cast the part. This process includes determining how the part can be drafted (which way the sand can release from the pattern), the best strategies to make the mold, and determining the proper gating/risering arrangement - tubes and reservoirs that allow the liquid metal to both quickly fill the mold cavity, and compensate for uneven shrinkage distribution during solidification. It can take many iterations to determine the correct gating arrangement for each component, depending largely on the complexity of its geometry and size of the part.

It quickly became evident that the project was of a scale that could payback investments spent in process engineering, so figuring out a faster and more economical way to produce the components was of the essence. The first step was to internalize production by building our own foundry, and the second, employing digital technologies in the process.

A parametric model was created for the basic genetic coding of the members, which defined relationships for the following: a basic structural diagram distributing the mass, scaling up the iron to a softer silicon bronze, creating draft angles on all sides for molding in sand (release), creating flat planes for the bolt-head locations, and other geometric/aesthetic criteria. The definition was then instantiated into the measured geometry for each location throughout the length of Shireen. Using the parametric definition not only saved from having to model each component individually, it ensured the components would all follow the rules necessary for casting and structural integrity, while sharing a kindred aesthetic. Having everything parametrically controlled allowed for changing the rules, geometry, aesthetic criteria, or thicknesses downstream - for all the components simultaneously. The models also easily fed back volume distributions, component weight, and drawings for the owner’s review.

Significant savings are made in the process of mold-making simply by virtue of having the floor geometry defined digitally. Rather than producing a free component in-the-round, one can produce two half-patterns mounted on a board (in relief) called a match-plate, which speeds up the molding processes significantly. The gating & riser design can then also be drawn digitally in the computer based upon volumetric distributions etc., and incorporated directly into the half-patterns, resulting in a complete model of the casting cavities and precise data on how much weight/volume was necessary to cast the part. The product of this is two perfectly registered and drafted half-patterns that can be easily used to ram the sand molds against.

The final and most critical part of this project was the switch from making traditional wood patterns to milling the sand molds directly on the CNC router. Please see "milling sand molds" for a more complete description of this process. This step alone saved the project thousands of dollars in labor and material costs by eliminating the need to produce positive patterns altogether, as well all of the associated waste. The digital form is simply inverted into negative geometry and milled from solid billets of resin-bonded foundry sand, into which the metal is cast.

What has been outlined by this project is a process that could be scaled for the mass-customization of metalcastings, which I believe is without precedent. In this particular project there was a lot of long-handed digital design dealing with the gating strategies and the complex parting lines of the various components, however I believe these things could be parametrically defined and automated to production given greater resources & time quite readily.

Using this process, myself and one colleague were able to cast all 36 members in two months flat, at a rate of nearly one component per day. Prior to implementing this process, producing one per week would have been fast. In total, we cast 1,800 lbs. of bronze into 16,500 lbs of sand. I expect these efforts to have saved the project tens of thousands of dollars in materials and labor costs, in addition to funding the building of a foundry, saving truckloads of material waste, and making possible the production of higher quality, more beautiful components in a quarter of the time. ~