50 Common Errors in Powering Predictions

Discussion in 'Hydrodynamics and Aerodynamics' started by Leo Lazauskas, May 13, 2011.

  1. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 151, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    I have attached an interesting paper by John Almeter of the US Navy's Carderock Division. I'm sure others here can push the number of potential errors well above 50.

    "Local Angle": papers by two NAs (Don Macpherson and Dr. Patrick Couser) who sometimes bob up on boatdesign.net are cited.

    Hiper08
    6th Int. Conference on High-Performance Marine Vehicles,
    Naples, Italy, 18-19 Sept. 2008.

    Leo.
     
    Last edited: Aug 12, 2015
    1 person likes this.
  2. Alik
    Joined: Jul 2003
    Posts: 3,007
    Likes: 301, Points: 93, Legacy Rep: 1306
    Location: Thailand

    Alik Senior Member

    That's an excellent paper, have read it before. Every 'software user' should read it prior to pressing keys in performance prediction software.
     
  3. BYDE
    Joined: Mar 2011
    Posts: 54
    Likes: 3, Points: 0, Legacy Rep: 66
    Location: Italy-China

    BYDE Junior Member

    Yes, interesting paper. Thanks for sharing.

    By the way, Alik I'm reading your paper from the HPMV conference in Shanghai last month, the one about catamarans. I'd like to thank you, there's a lot of valuable data and experience about catamarans. A very nice gift from you to other yacht designers! :)
    May I ask if I could have also your referenced paper [7]? namely the one you did for the Chesapeake Symp. 2010. Thanks
     
  4. Alik
    Joined: Jul 2003
    Posts: 3,007
    Likes: 301, Points: 93, Legacy Rep: 1306
    Location: Thailand

    Alik Senior Member

    Papers from CPBS-2010 are supposed to be available from SNAME... Just send me Your email by PM.
     
  5. BMcF
    Joined: Mar 2007
    Posts: 1,061
    Likes: 79, Points: 48, Legacy Rep: 361
    Location: Maryland

    BMcF Senior Member

    Thanks for that Leo.

    I noted with a chuckle that I've worked in the past (recent past even) with every single one of the fine gentlemen who were acknowledged at that end of the paper.;) I can also smile privately knowing exactly the details of a couple ill-conceived HPMV projects to which John is almost referring to by name. To his credit though, he kept the criticism generic and oblique. I've often failed to be so charitable myself.
     
  6. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 151, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    He was less than generic and oblique in his opinion of one piece of software.
    Thankfully, for me, it was a good review :)
     
  7. BMcF
    Joined: Mar 2007
    Posts: 1,061
    Likes: 79, Points: 48, Legacy Rep: 361
    Location: Maryland

    BMcF Senior Member

    That is always a nice thing, eh? ;) I was recently put in the position of once again criticizing a certain piece of Scandinavian motion prediction software, now so commonly used. I believe my comment on the disparities noted between the predictions and actual results were something charitable though: "I see that after almost 20 year now, V{program} continues to reliably produce consistently unreliable results. At least that means the code is stable."
     
  8. DMacPherson
    Joined: Mar 2005
    Posts: 139
    Likes: 28, Points: 28, Legacy Rep: 209
    Location: Durham, NH USA

    DMacPherson Senior Member

    Leo:

    I'm writing this with a bit of a grin here, as I was going to avoid commenting on John's mention of NavCad. (It would have been OK with me if you had mentioned it by name, but I appreciate your consideration, none-the-less.)

    I think John was talking more about the failure of the methods that happen to be in NavCad, rather than the program itself. (Which I believe he explicitly states.) John's comments may be fair, but it it is hard to tell as there is no data to support it - only a results plot. One would have no way of knowing if he followed the advice of the "Method Expert" ranking system, or went through a typical NavCad user vetting process for the methods.

    Using a tool like NavCad requires the engineer to be involved. It is a forecasting tool that includes a collection of resistance prediction methods from planing hull to tankers. If appropriate methods are not chosen, results will be bogus. Our philosophy (right or wrong) is to allow the user to conduct whatever calculation they want to do - but we will give you as many tools as possible to determine if you are getting reliable results. These include the aforementioned "Method Expert", as well as parameter range tables, pointers to the original references, and comments that are gleaned from our own and user experiences with the methods.

    (It is also fair to point out that published methods vary in their sensitivity and behavior to hulls that are outside their original data set. One "added value" to NavCad is that we leverage a broad body of knowledge from our users. Some methods are given additional constraints based on this knowledge. The Holtrop method is an example of this, where very heavy hulls do not do well. Likewise with some of the semi-displacement methods, where transom immersion and waterplane area were important, these parameters were not identified in the original publication. In both cases, we conducted a re-evaluation of the original data, to the extent that we could find it, and developed a set of new criteria that greatly improved the ranking of the methods for this class of vessels.)

    Some years ago, I was at a trade show and a student walked up and engaged me in discussion. He was using one of the utility programs that were included in a larger hull design suite, and even though their school had NavCad, he was using the utility tool. In reply to my query about this, he stated emphatically that "A******** make my boats go faster than NavCad". He was intentionally seeking out the method that made his boat "look" the fastest. As much as I tried to explain that the boat would physically go only as fast as it was going to go, and that these tools were to forecast this speed, and that methods are like weather forecasting models, he was having none of it. He was using a model for Arizona, when his boat was living in Alaska.

    So, in the same way that it is easy to intentionally make a prediction look good (underpredict), it is also easy to make a prediction look bad (overpredict). You just have to ignore the advice and warnings in the software. I'm not saying that this was John's intention, but without the supporting data, the criticism is really not worth much.

    Don MacPherson
    HydroComp
     
  9. DMacPherson
    Joined: Mar 2005
    Posts: 139
    Likes: 28, Points: 28, Legacy Rep: 209
    Location: Durham, NH USA

    DMacPherson Senior Member

    Leo:

    It just occurred to me that you probably were talking about the SWPE comments. Sorry that I presumed you were referring to the comments about NavCad.

    Don
     
  10. Leo Lazauskas
    Joined: Jan 2002
    Posts: 2,696
    Likes: 151, Points: 63, Legacy Rep: 2229
    Location: Adelaide, South Australia

    Leo Lazauskas Senior Member

    No worries, Don.

    I agree that some users expect a lot from computer codes, and they often use them without thinking.

    My pet peeve is when users claim that Michlet or other thin-ship codes I produce have been "validated" for ships with L/B as low as 4! They never mention the thousands of cases for stubby hulls where predictions are terrible.

    Leo.
     
  11. BMcF
    Joined: Mar 2007
    Posts: 1,061
    Likes: 79, Points: 48, Legacy Rep: 361
    Location: Maryland

    BMcF Senior Member

    My pet peeve is related to my line of work which is the prediction (and control) of ship motions. I'm happy to leave the prediction of speed and powering to others. :D

    Until very recent times..very recent..the most accurate prediction tools for advanced hullforms were physics-based frequency-domain programs. A couple of those were so routinely accurate for the specific range of hull types modeled that model testing for motions almost became only an exercise in validation of the predicted motions.

    Yet year after year, and still to this day, I see motion predictions produced from 'other' commercially applied or available frequency-domain packages - one in particular - that are routinely off the mark by significant amounts. When/if the discrepancies fall out during comparisons with actual model or full scale test data, the standard explanation for the errors is always the same: "It's probably due to non-linearities that cannot be handled in the frequency domain".:rolleyes:

    Unfortunately..and as is so often the case..the highly validated frequency-domain tools 'belonged' to military organizations and are simply lost or forgotten when those organizations have large gaps between projects that require the software tools be dusted off and used.

    Evey new program with its attendant batch of new engineers and scientists then states that "Simulation and modeling tools for this vessel type do not exist or are inadequate and we need lots o money to..."

    pffft
     
  12. DMacPherson
    Joined: Mar 2005
    Posts: 139
    Likes: 28, Points: 28, Legacy Rep: 209
    Location: Durham, NH USA

    DMacPherson Senior Member

    Couldn't agree more. Tech transfer from the naval community can be like that from the academic community. There is a lack of continuity, yes, but perhaps the biggest issue is simply that of expectation.

    A tool that is written for yourself (which could mean an individual or a larger group) has no need for "tech support" or "new user" considerations in the development. It is just a deliverable for a particular calculation objective. The authors and users are regularly in the same body. (The circumstances of academic software probably is worse, in that the accuracy of results just is not all that important. The objective is the fulfillment of some project requirement, whose value is measured in the learning involved rather than the veracity of results. This is appropriate, of course, as learning is the purpose - but it makes for a risky commercial tool.)

    From our side, we put "tech support" and "new user" considerations right at the top of the list. Everyone is a new user once, and if you don't have need to apply the tool frequently, you can be a "new user" many times. So, we place calculation behavior and stability at a higher priority than precision. The sharpest point is easiest to break. I would rather have a prediction method that was consistently off by some small amount and never greatly wrong, than one that was perfect most of the time but could go off on a flier.

    Don
     
  13. BMcF
    Joined: Mar 2007
    Posts: 1,061
    Likes: 79, Points: 48, Legacy Rep: 361
    Location: Maryland

    BMcF Senior Member

    To put the fairest possible point on it; although some of the motion prediction tools I was referring to were refiend to the point or remarkable fidelity and accuracy, they were for the oddball craft types. SES and SWATH vessels, high-speed catamarans, even planing monohull motion prediction was a tricky bit.

    So while I have often criticised the ability of 'wanna be' programs that purported to usefully predict the motions of various HPMV types, I wouldn't want to be stood up in front of an audience and pressed too hard to answer how many numbers of those vessel types have been built where that great prediction software was usefull..compared to all other types.:D

    To try and sneak this back on topic..I've noted that its the very same vessel types I play with, particularly the fastest ones, that are so routinely the ones that suffer the consequences of erroneous speed/powering predictions. Yet some were predicted entirely correct. SEMO's 40m SES..the USN SES-200...both classes of RNoN SES vessels.
     
  14. DMacPherson
    Joined: Mar 2005
    Posts: 139
    Likes: 28, Points: 28, Legacy Rep: 209
    Location: Durham, NH USA

    DMacPherson Senior Member

    Luck? Shoot at enough targets and you are bound to hit a few.
     

  15. BMcF
    Joined: Mar 2007
    Posts: 1,061
    Likes: 79, Points: 48, Legacy Rep: 361
    Location: Maryland

    BMcF Senior Member

    Not hardly. Performance targets were hit due to very careful adherence to proven prediction methods..methods developed and refined over many years by, coincidentally enough, by the author and nearly all those listed in the author's acknowledgements in the paper that Leo put out to start this thread.


    Go figure.:D
     
Forum posts represent the experience, opinion, and view of individual users. Boat Design Net does not necessarily endorse nor share the view of each individual post.
When making potentially dangerous or financial decisions, always employ and consult appropriate professionals. Your circumstances or experience may be different.