A physicists perspective on a material science/engineering problem I am looking into some research that involves engineering and material science. As a physicists I wondered what other physicists would think of this problem and how they would approach it. Much of the experimental work is already in progress so this can be ignored.
We would like to know the impact of humidity and temperature on the stress/strain in very large fabrics.
To start let's say we have a very large piece of some fabric (e.g. wool). It may be pristine or may have degraded over time (e.g. a historical tapestry). We wish to measure the stress/strain relationship using sensors (that do not impact the stress/strain within the large fabric) as well as the temperature and humidity in or around the material and send this data to a computer (I would assume).
How we would begin to model this relationship and what would the model look like? Can anyone recommend any papers upon which I can expand my research or has a similar model I can work with?
The main problem I am having is where to begin, I can see where it would begin from a technical stand point but (setting the test up to collect data). But how would this then be used to create or expand upon an existing model? What would such a model look like? Finally if a model were to be developed what accuracy could one place upon it?
Any help with this would be so amazing! If I figure out how to add a bounty for a good question then I surely will!
Regards,
 A: A great question to ask. You can find some ideas in the paper
Multidimensional analysis of excitonic spectra of monolayers of tungsten disulphide: toward computer-aided identification of structural and environmental perturbations of 2D materials.
Although it is about sheets of nanomaterials, the general idea is very much transferable to other fields.
Shortly, you would need to come up with a mean/tool/method that allows you to attach certain information to each pixel of your fabric. Ideally this information should contain more than one value so you could gather some multidimensional dataset.
Assuming your fabric is $N\times M$ pixels and each pixel is attached $D$ numbers (features) representing attached information. The dataset then will contain  $N\times M$ data-samples with each data-sample consisting of $D$ numbers. Once you collect your data, you can apply machine learning tools such as principal component analysis and K-means clustering.
When trying to find which parameters to attach to each of the pixels, it is important to think of those that are somewhat sensitive to strain/stress and temperature/humidity if those cannot be retrieved directly. Ideal case if those parameters are measured directly, then you will need only 3 features (strain/stress, temperature and humidity) and your phase space will be three-dimensional.
Then the model would be a multi-dimensional data-cloud embedded in a multi-dimensional parametric phase-space to be analyzed by machine learning algorithms and personnel with relevant expertise.
In the exemplary paper above, authors could tell apart areas on a material sheet affected by strain, abundance of electric charge, water, and locations of thickened regions.
