Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Inverse calculating of a convective heat flux coefficient
Posted Dec 10, 2015, 10:06 p.m. EST Heat Transfer & Phase Change, Optimization 1 Reply
Please login with a confirmed email address before reporting spam
Hi guys,
Geometry: an approximate cube (see Fig. 1), heat-exchange only happens between air and the cube through the curved side surface, and other surfaces are adiathermal. The convective heat flux coefficient is just what we want to obtain.
Monitored temperature data: the temperatures of the two blue points (coordinates: (0.5,0,0.2) and (0.5,0,0.8)) at 0.5 and 1 days are 10 and 5 oC respectively.
Physics: heat transfer & Optimization
Study: Optimization (method: Levenberg-Marquardt; Optimality tolerance: 0.01) & time dependent (times: range (0,0.02,1) time unit: d)
Questions:
1. In the tips of help document about least-suqares objective, it says ” There is no need to add points to the geometry at the measurement locations specified in the file.” However, an error will occur (see Fig. 2) if the two points are not selected under the Settings-Point, and the corrdinates are only assigned under the Experimental Data-Filename (see Fig. 3 .csv file). So, why? The first column is time; it seems that default unit is second (see Fig. 4). The second to forth columns are the corrdinates. The fifth column indicate monitored temperatures.
2. The monitored temperatures at one of the two points (see Fig. 4). Why no unit of y-coordinate? since the unit is defined as degC under Value Coumn-Value Column 1-least Squares Objective 1.
3. Why showing time=86400d when setting time as 1 day under the Settings-Temperature-Results? Why there is a big difference between the monitored and simulated temperatures (see Fig. 5)? Why no changes happen to the simulated temperatures by changing the monitored temperatures (that are 10 and 5 oC), decreasing the optimality tolerance, increasing the mesh density, etc.?
4. How to set different contribution weights for the two points, like 0.1 and 0.9? Please help check the settings of Least-Squares Objects 1 including time, corrdinates and value columns as well as the experimental data (.csv file).
5. Why the least squares objective values are so large (see Fig. 6)?
Please help me solve the problems. Thank you very much.
Kind regards,
Honglie Zhang
Geometry: an approximate cube (see Fig. 1), heat-exchange only happens between air and the cube through the curved side surface, and other surfaces are adiathermal. The convective heat flux coefficient is just what we want to obtain.
Monitored temperature data: the temperatures of the two blue points (coordinates: (0.5,0,0.2) and (0.5,0,0.8)) at 0.5 and 1 days are 10 and 5 oC respectively.
Physics: heat transfer & Optimization
Study: Optimization (method: Levenberg-Marquardt; Optimality tolerance: 0.01) & time dependent (times: range (0,0.02,1) time unit: d)
Questions:
1. In the tips of help document about least-suqares objective, it says ” There is no need to add points to the geometry at the measurement locations specified in the file.” However, an error will occur (see Fig. 2) if the two points are not selected under the Settings-Point, and the corrdinates are only assigned under the Experimental Data-Filename (see Fig. 3 .csv file). So, why? The first column is time; it seems that default unit is second (see Fig. 4). The second to forth columns are the corrdinates. The fifth column indicate monitored temperatures.
2. The monitored temperatures at one of the two points (see Fig. 4). Why no unit of y-coordinate? since the unit is defined as degC under Value Coumn-Value Column 1-least Squares Objective 1.
3. Why showing time=86400d when setting time as 1 day under the Settings-Temperature-Results? Why there is a big difference between the monitored and simulated temperatures (see Fig. 5)? Why no changes happen to the simulated temperatures by changing the monitored temperatures (that are 10 and 5 oC), decreasing the optimality tolerance, increasing the mesh density, etc.?
4. How to set different contribution weights for the two points, like 0.1 and 0.9? Please help check the settings of Least-Squares Objects 1 including time, corrdinates and value columns as well as the experimental data (.csv file).
5. Why the least squares objective values are so large (see Fig. 6)?
Please help me solve the problems. Thank you very much.
Kind regards,
Honglie Zhang
Attachments:
1 Reply Last Post Dec 10, 2015, 10:40 p.m. EST
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Supplementary information (see the attached picture)
Attachments: