What is the differences between RUL models and ML models?

I would like to work about Remining Useful Life estimation. Matlab has several models for it such as Survival, Degradation, and Similarity. What are them based on? If I work by ML models(SVM, DT, kNN) is there any differences woking with ML models instead of Matlab's model?

Answer 1

[the cyclist]

This is not an easy question to answer, and in some ways it is not really a MATLAB question. It is more of a generic modeling question. "What is the best model for my problem?" does not have a simple answer.

I do not have expertise in estimating remaining useful life, but I do have some experience in survival analysis. I have two opinions to share here.

First, I would say that the main difference between the ML models and the RUL models (whether you use MATLAB or not) is that the RUL models are typically going to be parameterized models, where you fit a specific function to your data, and get parameter estimates out. ML models are typically non-parametric. Broadly (but not always), ML models can be better at predicting your outcomes, but might be more difficult to interpret why you get those outcomes. The parameterized models will typically be easier to interpret (e.g. "variable A has the largest impact on remaining life, so I should try to improve it").

Second, I would point you to this documentation for deciding from among the different RUL models, based on the data you have, if you choose to go with an RUL model.