Skip to content

Suggestions as to how to proceed would be appreciated...............

7 messages · Bert Gunter, Bernard McGarvey, David Winsemius +3 more

#
I work in aspects of Cold Chain transportation in the pharmaceutical industry. These shippers are used to transport temperature sensitive products by surrounding the product load box with insulating materials of various sorts. The product temperature has lower and upper allowed limits so that when the product temperature hits one of these limits, the shipper fails and this failure time is teh shipper duration. If the shipper is exposed to very low or very high ambient temperatures during a shipment then we expect the duration of the shipper to be low.

The particular problem I am currently undertaking is to create a fast way to predict the duration of a shipping container when it is exposed to a given ambient temperature.

Currently we have the ability to predict such durations using a calibrated 3D model (typically a finite element or finite volume transient representation of the heat transfer equations). These models can predict the temperature of the pharmaceutical product within the shipper over time as it is exposed to an external ambient temperature profile. .

The problem with the 3D model is that it takes significant CPU time and the software is specialized. What I would like to do is to be able to enter the ambient profile into a spreadsheet and then be able to predict the expected duration of the shipper using a simple calculation that can be implemented in the spreadsheet environment. The idea I had was as follows:

1. Create a selection of ambient temperature profiles covering a wide range of ambient behavior. Ensure the profiles are long enough so that the shipper is sure to fail at some time during the ambient profile.

2. Use the 3D model to predict the shipper duration for the selection of ambient temperature profiles in (1). Each ambient temperature will have its own duration.

3. Since only the ambient temperatures up to the duration time are relevant, truncate each ambient profile for times greater than the duration.

4. Step (3) means that the ambient temperature profiles will have different lengths corresponding to the different durations.

5. Use the truncated ambient profiles and their corresponding durations to build some type of empirical model relating the duration to the corresponding ambient profile.

Some other notes:

a. We know from our understanding of how the shippers are constructed and the laws of heat transfer that some sections of the ambient profile will have more of an impact on determining the duration that other sections.
b. Just correlating the duration with the average temperature of the profile can predict the duration for that profile to within 10-15%. We are looking for the ability to get within 2% of the shipper duration predicted by the 3D model.

What I am looking for is suggestions as to how to approach step (5) with tools/packages available in R.

Thanks in advance

Bernard McGarvey, Ph.D.

Technical Advisor
Parenteral Supply Chain LLC

Bernard.First.Principles at gmail.com mailto:Bernard.First.Principles at gmail.com

(317) 627-4025
#
Please read the posting guide(PG) inked below. Your query sounds more like
a project that requires a paid consultant; if so, this is way beyond the
scope of this list as described in the PG. So don't be too surprised if you
don't get a useful response, which this isn't either of course.


Bert Gunter

"The trouble with having an open mind is that people keep coming along and
sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Sun, May 22, 2022 at 10:40 AM Bernard McGarvey <
mcgarvey.bernard at comcast.net> wrote:

            

  
  
#
Its simply a query to know what tools/packages R has for correlating single values with multivalued vectors. If that is outside the scope of the PG then so be it.

Bernard

Sent from my iPhone so please excuse the spelling!"

  
  
#
There are several CRAN Task Views. Some of them should intersect with your question. I don?t think your description of the problem suggest that multivariate correlation is the best approach.  Some sort of optimization or numerical simulation would seem to be more fruitful.

? 
David 
Sent from my iPhone
#
Would lm, nls, or nlme work for what you need? 
Tim

-----Original Message-----
From: R-help <r-help-bounces at r-project.org> On Behalf Of Bernard Comcast
Sent: Sunday, May 22, 2022 3:01 PM
To: Bert Gunter <bgunter.4567 at gmail.com>
Cc: R-help at r-project.org
Subject: Re: [R] Suggestions as to how to proceed would be appreciated...............

[External Email]

Its simply a query to know what tools/packages R has for correlating single values with multivalued vectors. If that is outside the scope of the PG then so be it.

Bernard

Sent from my iPhone so please excuse the spelling!"
______________________________________________
R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see https://urldefense.proofpoint.com/v2/url?u=https-3A__stat.ethz.ch_mailman_listinfo_r-2Dhelp&d=DwIFaQ&c=sJ6xIWYx-zLMB3EPkvcnVg&r=9PEhQh2kVeAsRzsn7AkP-g&m=Pp0V5t70tdzhilCe5gEo6fR1inb2-RkIPZG4jtPvyUSKaWIjhiPEuI3ROOS_GDYh&s=O1gCEtqvVraPdsMGEWP_LcLa8IFCzatSw16BRhUsEF8&e=
PLEASE do read the posting guide https://urldefense.proofpoint.com/v2/url?u=http-3A__www.R-2Dproject.org_posting-2Dguide.html&d=DwIFaQ&c=sJ6xIWYx-zLMB3EPkvcnVg&r=9PEhQh2kVeAsRzsn7AkP-g&m=Pp0V5t70tdzhilCe5gEo6fR1inb2-RkIPZG4jtPvyUSKaWIjhiPEuI3ROOS_GDYh&s=XggaTfrxoB9UFmUs6lC2e9SOz_QVk29KVRAMmgUT6ng&e=
and provide commented, minimal, self-contained, reproducible code.
#
Hi Bernard,
My first guess would be to simulate changes in the internal
temperature of the shipment in manageable time increments, say one
hour. Then estimate the ambient temperature for each time increment
from historical records. This would require a few thousand
calculations for the usual shipment time if I'm correct. In other
words, just successively calculate the internal temperatures until
failure or successful delivery.

Jim

On Mon, May 23, 2022 at 3:40 AM Bernard McGarvey
<mcgarvey.bernard at comcast.net> wrote:
#
Some thoughts:

1) Actual profiles are only known after the shipment is complete. Thus, the largest source of uncertainty in prediction of shipping duration is likely to be in any prediction made regarding the shape/magnitude of the profile. Your goal of 2% may simply be infeasible.

2) Thermal conduction is fairly accurately modeled as a linear differential equation. As such, exposing your 3d model to a step change profile and fitting an exponential curve is likely to capture a lot of the characteristic behaviour of the shipment. However, time lags between the center and surface of the shipment create an initial transient that can obscure the effective coefficients. 

3) Convection is not so linear as conduction, so if your shipment is in the middle of a larger packing form then the coefficients will be different than if they were at the packing boundary. Also, when at the packing boundary, air velocity can affect the apparent coefficients, whereas if not then wind will not affect results.
On May 22, 2022 3:44:03 PM PDT, Jim Lemon <drjimlemon at gmail.com> wrote: