Tutorials

(Tutorial Slides)

Presenter: Chris Teubert; NASA Ames Research Center

Description: This year, NASA Released the Prognostics Python Packages, a collection of research tools for developing prognostic models, simulating degradation of systems, performing prognostics, analyzing results, and developing new prognostic algorithms. The Prognostics Python Packages encapsulate design improvements based on experience gained with the previously open-sourced Generalized Software Architecture for Prognostics (GSAP) and Prognostics Matlab Libraries. The tools have been extended to include new features such as approaches for resource-constrained prognostics (e.g., sample shedding), hybrid models, model validation tools, online system identification, and surrogate model generation. This tutorial is a hands-on overview of these tools, including how to use and extend them. This tutorial will also include examples of how the tools are used at NASA.

There are two ways to run the tutorial, described below:

Option 1: Online

Use the following links to run the tutorial:

1. Part 1 prog_models: https://mybinder.org/v2/gh/nasa/prog_models/master/?labpath=tutorial.ipynb
2. Part 2 prog_algs: https://mybinder.org/v2/gh/nasa/prog_algs/master/?labpath=tutorial.ipynb

Option 2: Running on Your Machine

• To run this on your own machine, we suggest you download the following tools and software:
  • Python (3.6-3.9 supported) – https://www.python.org/downloads/
  • Pip – https://pypi.org/project/pip/
  • Jupiter Notebooks – https://jupyter.org. I personally use VS Code with the Jupiter
  Notebooks plugin.

Once these are installed, download the tutorials from here:

1. Part 1 prog_models: https://nasa.github.io/prog_models/_downloads/200bc74cf412da163fa7fd0ac3a7bb56/tutorial.ipynb
2. Part 2 prog_algs: https://nasa.github.io/prog_algs/_downloads/200bc74cf412da163fa7fd0ac3a7bb56/tutorial.ipynb

Finally, install prog_algs using the following command:
pip install prog_algs

(Tutorial Slides)

Presenter: David Siegel; Predictronics

Description: With recent advancements in sensors, IoT, and machine learning-based analytics, it’s an exciting time for leveraging the promising new developments and research in prognostics and health management for fielding solutions in various industrial applications. This tutorial will discuss some of the challenges that arise in deploying solutions, including data quality, training data requirements, maintaining algorithm accuracy/re-training, and how to make better decisions from the fielded system. A methodology on how to approach the development and deployment of a PHM solution for each application will be presented, along with best practices and lessons learned. Case studies in manufacturing, heavy industry, and aerospace, among others, will be presented. Lastly, new developments and ideas will be shared, exploring thoughts on the improvement of existing PHM fielded solutions and addressing the current unmet challenge.

Presenter: Felipe Viana; University of Central Florida

Description: We present a modeling framework that allows for the simultaneous use of physics-informed and machine learning by implementing recurrent neural networks for cumulative damage modeling. The main advantage of this approach is the compensation of limitations in physics-informed kernels as well as labeled data. This framework handles highly unbalanced datasets formed by a few output observations and data-lakes containing time series used as inputs. This tutorial will focus on both formulation and computational implementation. There will be two hands-on examples on model identification of dynamic systems as well as fatigue crack growth modeling written in TensorFlow using the Python programming language API. The participants will have direct access to the Python scripts and will be able to run them on their personal laptops.