Choose Language Hide Translation Bar
bakoppisepp-950
New Contributor

Utilisation des équations de surfaces de réponse

Bonjour tout le monde,

 

J'ai obtenu des courbes de surface de réponse à partir d'un plan d'expérience. 

J'aimerais utilisé ces surfaces de réponse pour réaliser des calculs ; j'aimerais donc obtenir l'équation de ces surfaces de réponse. Pour obtenir ces équations, j'ai pris les résultats obtenus dans le tableau "estimations des paramètres". En revanche, en prenant ces résultats et en traçant les courbes sur un autre logiciel, je n'obtiens pas les mêmes courbes.

 

Est ce qu'il y a une chose particulière à savoir pour écrire ces équations de surface de réponse ?

 

Merci d'avance.

 

Laura

0 Kudos
1 ACCEPTED SOLUTION

Accepted Solutions

Re: Utilisation des équations de surfaces de réponse

I would guess that this design was not created in JMP since the coding property is not being used. However, in order to protect you and provide more meaningful analyses, JMP will automatically center the interaction and quadratic terms. That is why you see (t-416.629) and (T-191.357) and is likely what led to your issue.

 

In order to determine values of t and T that will lead to lower concentrations, you have a few options to explore this. Keep in mind that the proper region with low concentrations is likely not rectangular, so you can't just say t < X value and T < Y value.

 

One approach is to use the contour profiler. 

From the red triangle choose Factor Profiling > Contour Profiler. You will see a contour plot of time versus temperature and a single contour drawn for your concentration. Put your upper limit for concentration in the "Hi Limit" field. The areas above that upper limit will be shaded out. Anything in the white space is a predicted good combination of time and temperature.

 

There could be other approaches, but I think this is probably the quickest and easiest way to see the acceptable region for t and T.

Dan Obermiller

View solution in original post

4 REPLIES 4

Re: Utilisation des équations de surfaces de réponse

You are not providing many details here, so I will be making a guess on what is going on.

 

The coefficients that JMP reports on the parameter estimates table are often in terms of the coded units for the factors, not the natural units. There are many reasons for this, but that is beyond the scope of this post. A quick way to tell if this is the case, look at the label for the factors. Does it list something like A(x, y) where x and y are specific numbers? If so, then the parameter estimates are in terms of coded values. From the red popup menu you can ask for the prediction equation which lists everything explicitly. You could also save the prediction formula to ensure that you are getting the proper formula. If this is NOT what is going on, please post some details (what your formula is, ranges for the factors, what you see from the other package, etc.)

 

And finally, one more question: why not just get the graph you need from JMP? I would be curious to know what graph you desire that JMP does not offer.

 

Dan Obermiller
bakoppisepp-950
New Contributor

Re: Utilisation des équations de surfaces de réponse

Merci pour votre réponse.
A partir du menu contextuel rouge (estimations --> afficher l'expression de la prévision), j'obtiens bien une équation.
Dans mon cas, j'obtiens
-16,25164+ 0,1232075*T -0,002904*t -0,063439 * ((T-191,357) * (T(-191,357)) -0,001744 * ((T-191,357) *(t-416,629) -7,258e-5 * ((t-416,629)*(t-416,629))
Il s'agit de la concentration d'un composé en fonction du temps (t) et de la température (T), qui varie de mon expérience d'environ 300 à 500 secondes, et de 185 à 200 °C.
J'aimerais trouver les valeurs de température et de temps que je peux appliquer pour ne pas dépasser un seuil de concentration.
0 Kudos

Re: Utilisation des équations de surfaces de réponse

I would guess that this design was not created in JMP since the coding property is not being used. However, in order to protect you and provide more meaningful analyses, JMP will automatically center the interaction and quadratic terms. That is why you see (t-416.629) and (T-191.357) and is likely what led to your issue.

 

In order to determine values of t and T that will lead to lower concentrations, you have a few options to explore this. Keep in mind that the proper region with low concentrations is likely not rectangular, so you can't just say t < X value and T < Y value.

 

One approach is to use the contour profiler. 

From the red triangle choose Factor Profiling > Contour Profiler. You will see a contour plot of time versus temperature and a single contour drawn for your concentration. Put your upper limit for concentration in the "Hi Limit" field. The areas above that upper limit will be shaded out. Anything in the white space is a predicted good combination of time and temperature.

 

There could be other approaches, but I think this is probably the quickest and easiest way to see the acceptable region for t and T.

Dan Obermiller

View solution in original post

bakoppisepp-950
New Contributor

Re: Utilisation des équations de surfaces de réponse

Merci !
0 Kudos