Modeling the Bowl – Part 2 : First Attempts

2010 August 8

Our computer modeling subcontractor, Phoenix Analysis and Design Technologies, has been working diligently on how to model the bowl-shaped roasting vessel for the project. They have tried several different method and finally decided on one that seems to work well. This week, Jim Peters, our contact with PADT, invited us to the office to demonstrate the initial findings before signing a contract for the full modeling work.

PADT has many engineers on staff, of which some are off-site. In this case, Carlos, in their Philadelphia office did the initial work on refining the modeling of the bowl. We contacted Carlos by speakerphone and began discussing the models. In all cases, they used a perfect hemisphere for the bowl shape. Carlos came up with a model for an individual bean composed of several elements. He then began running the model with a few beans to test it. This takes just a couple of minutes for the computer to do the calculations and create a video of the results. Later, he tried 30,000 individual beans (about a half batch) to see how the bowl performed. The model is calculating the position of each bean, so the run time to make a single movie of the bowl in operation is approximately 1 day. When we model 60,000 beans (a full batch), the run time is about 1 week per video.

Reviewing the Results

Carlos lead us through the model results one by one while Jim Peters projected them on a screen for us to observe. Carlos also began to share some important observations  he had noticed while working with the model. First of all, the diverter cone in the bottom is very important to get the beans moving in the first place by getting them out of the central “dead spot”. I had intended for this to happen, as well as to divert the airflow from the heating jet into an appropriate pattern. Secondly, he noted that rotational speed of the bowl probably does not need to exceed 100 rpm but should be variable, which is easy to accomplish. Thirdly, the beans at any fixed rotational speed eventually “stall” or “stick” in a ring around the bowl. This is at the point when the rotational forces, gravity, and friction are all in balance. The faster the rotation, the higher up the “stall” ring occurs.

The Sticking Problem

Several videos were shown to demonstrate this phenomenon. Please click on this link to view a Bowl Model Video. Clearly, this is not a good thing. One of the design goals is excellent and consistent mixing of the beans. Having them stick in a ring means they are essentially fixed in place and not mixing at all (some burning, some raw). Carlos suggested that we consider a “scraper” to break up the ring and return it to the center of the bowl. He ran the model with a scraper to demonstrate the idea.

It seems to me that a better solution may be to make the bowl more shallow, perhaps on the order of a typical kitchen wok. This would speed up the beans and there would be no sticking. The key question is how does one force the beans to curl back over to the center of the bowl with out reaching some sort of equilibrium and stalling. This may include stationary vanes similar to the scrapers. As Carlos has demonstrated, scrapers can work and perhaps that is a good solution. The initial modeling has left quite a few questions to ponder.

The Next Step

So as not to waste money on endless modeling, I will need to narrow down the experiment to the most promising methods and have PADT model those. I would like to look at a number of variables, including bowl shape, bean return methods, speed, and batch size. Since the run time is highly dependent on the number of beans being tracked, I may suggest that we do quarter batch modeling until we find the best design, then ramp up the batch.

One Response
  1. Dean Richardson permalink
    August 10, 2010

    Very cool research…

    I might suggest that since a scraping mechanism is functional, why not use your hot air source, instead of directly down the center but aimed at a specific or actually intentionally inspecific spot to drive the beans off of the static site. This can be done VERY simply by setting your hot air source near your “equator” and then the hot air will scrape your beans. You already have mechanical agitation or “Forced Convection” so hot spotting will not really be a factor. Simple solution for a problem you have already solved by modeling.

Comments are closed.