François Gallard

Questions and Answers (13) View all

• Answer added in Aerospace
  13 What will happen if we make a temperature difference between the upper and lower sides of the wing of an aircraft?
  By Geen Paul · Anna University, Chennai
  François Gallard · Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique

  Another question would be what heat flux is required to maintain this temperature gradient (wall constant temperature heat flux), then how much energy... [more]

  Another question would be what heat flux is required to maintain this temperature gradient (wall constant temperature heat flux), then how much energy do you need to put in the system, and finally does it worth the benefits ? Also I don't see the point of increasing lift of an airfoil (if we are not speaking about maximum lift at stall). Increasing the angle of attack will increase the lift. What is important is the drag at a given lift.
  Following
• Answer added in Multidimensional analysis
  17 How to minimize a set of fitted multidimensional data?
  By Anna Obertacke · Bergische Universität Wuppertal
  François Gallard · Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique

  Note : there are many other multidimensional interpolation/regression methods like krigging, slplines are usually taken for dimension 3 at most. 

  Note : there are many other multidimensional interpolation/regression methods like krigging, slplines are usually taken for dimension 3 at most.
  Following
• Answer added in Multidimensional analysis
  17 How to minimize a set of fitted multidimensional data?
  By Anna Obertacke · Bergische Universität Wuppertal
  François Gallard · Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique

  Hi, yes, first you can create a continuous model of the data by interpolation for example. Using radial basis function or Proper orthogonal deompositi... [more]

  Hi, yes, first you can create a continuous model of the data by interpolation for example. Using radial basis function or Proper orthogonal deomposition. In dimension 15 it will be very costly and you will need a lot of samples (tipically millions or billions). The best or maybe the only way in this dimension would be to try to first remove the useless variables or split them by their influences using analysis of variance and at least by exploiting the properties that you know of your problem. Then you would need and efficient optimization strategy, and you have to know about your problem and take advantage of it, see "no free lunch theorem for optimization". Basically you can try gradient based algorithms or heuristic methods like geneic algorithms depending on if you need the global minimum or a local one. Or even couple the strategies.
  Following
• Answer added in Wind Power and Power Systems
  11 How can I predict the behavior of Aerofoils?
  By Karthikeyan N · Anna University, Chennai
  François Gallard · Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique

  Hello, you should use a CFD software, there are many. It depends if you can pay or not... But there are open source codes such as OpenFOAM and recentl... [more]

  Hello, you should use a CFD software, there are many. It depends if you can pay or not... But there are open source codes such as OpenFOAM and recently SU2, the Stanford code with airfoils test cases. It is parallel, unstructured and really easy to compile and run, with a great python interface. OpenFOAM is a bit harder to learn, but easy to modificate, also you don't need that. If you need to create an optimization loop then it is a much harder problem if you don't have any experience in it. You can start with hand design.
  Following
• Answer added in Turbulence
  11 Do numerical solutions explain the origin of turbulence?
  By Amador Muriel · Harvard University
  François Gallard · Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique

  I think that you are right Ricardo. It has not been proven, to my knowledge, that the NS equations do not contain all the information required to unde... [more]

  I think that you are right Ricardo. It has not been proven, to my knowledge, that the NS equations do not contain all the information required to understand and model turbulence. The capability of numerical methods to properly resolve these equations in the aim of understanding turbulence, assuming that analytical NS contains the required information, is another problem. Again, up to now, without considering the numerical cost of simulations, no fundamentamental impossibility of solving numerically a partial derivatives equation has been shown. Also doing DNS in high Reynolds regimes is still unafordable due to computational cost.
  Following