I am new here so hope this is the way to ask.
I have got a machine I want to test. I must stop the test when the machine is about 100 degrees celsius. I have got a logging device and logged the cooling down. I logged for about 10 minutes and I want to predict the temperature in 1 hour cooling.
Here is what I did measure.
$$ \begin{array}{c|lcr} T(min) & \text{T(C Measured)} & \text{T(C Calculated)} \\ \hline 0& 95.28& 95.28&\\ 1& 85.39& 85.44&\\ 2& 76.23& 76.74&\\ 3& 69.03& 69.04&\\ 4& 63.78& 62.24&\\ 5& 58.84& 56.22&\\ 6& 54.99& 50.90&\\ \end{array} $$
Ages ago I had Math lessons, and I though well lets see what I can remember. I did a search on the internet and found this site and it looked familiar.
http://www.ugrad.math.ubc.ca/coursedoc/math100/notes/diffeqs/cool.html
It is about a pan of soup cooling down, I should be able to use this method to calculate the cooling down of the machine right? But when I try and use the first few minutes of the data the calculated graph is different from the measured graph. I changed some numbers in the end formula with the method of “gamble and see what happens” and then it gets better, but then it will be a gamble and not a real calculation. Here is my calculation. The last 3 numbers are not close engough to the real measured data and I want to go to T = 60 min.
Calculation:
So I tried to use the first 3 minutes and see if the formula could predict te following minutes, this is in the column “T(C Calculated)”.
T(t) = To + (To-Ta)e(-kt) 69.03 = 10.25 + (95.28-10.25)e(-3k) 69.03 - 10.25 = 85.03e(-3k) 58.78 = 85.03e(-3k) 58.78/85.03 = e(-3k) 85.03/58.78 = e(3k) 1.447 = e(3k) ln(1.447) = ln (e(3k)) 0.3695 = 3k 0.3695/3 = k k = 0.123
T(t) = 10.25 + 85.03e-0.123*t
With the same method I used Calc (excel of open office) to calculate the k values for all the rows of time and I got this table:
Automatic calculated k values:
$$ \begin{array}{c|lcr} T(min) & \text{k} & \text{Delta k} & \text{AVERAGE} \\ \hline 1 & 0.1236510953 & . & . & . \\ 2 & 0.1268262348 & 0.0031751395 & . &. \\ 3 & 0.1230674915 & -0.0037587433 & . & . \\ 4 & 0.1156904727 & -0.0073770187 & . & . \\ 5 & 0.1119172774 & -0.0037731954 & . & \\ 6 & 0.1070226966 & -0.0048945808 & -0.0033256797 & \\ \end{array} $$
And changed the formula to:
T(t) = 10.25 + 85.03e-1*(0.123-(0.003326*t))*t
And got these values:
$$ \begin{array}{c|lcr} T(min)& \text{T(C)} & \text{Difference from measured values} \\ 0 & 95.2799999989 & 1.07968389784219E-009& \\ 1 & 85.3900071572 & -7.15722647726125E-006& \\ 2 & 77.0935046247 & -0.8635046247& \\ 3 & 70.1099181698 & -1.0799181698& \\ 4 & 64.2137289838 & -0.4337289838& \\ 5 & 59.2230004683 & -0.3830004683& \\ 6 & 54.9904554258 & -0.0004554258& \\ \end{array} $$
It is closer to the real measured data but still I do not trust my method enough to predict the coming hour. Can anyone please tell me what I might be doing wrong and what to change in my method? It would be helpful if someone did something similar, maybe the theory is different from the real thing?