Development of a reliability tool

Industrialising a prototype code for reliability computation

Renault Group (RG) has developed code to perform predictive reliability computation. As this code is a prototype, RG wishes to industrialise the developments in the form of a Python module, properly programmed, tested and documented. RG also wants to implement new functionalities.

The code includes several sections:

• Inferences from marginal distributions and correlation of raw data;
• Modelling of different business laws;
• Several evaluations: failure probabilities, validation plan, quantile calculation, using the excesses method;
• Export of results obtained in graphical and tabular format.

Technical specifications

In this context, RG approached Phimeca to :

• Develop the StaRe module, for STAtistical REliability, in Python;
• Improve the overall performance of the methods;
• Develop the code unit tests and the module documentation;
• Add new functionalities:
  • Implement a specific method for inferring the tails of distributions;
  • Implement a new method for calculating constraints.

Results

• Development of the StaRe module and its unit tests in compliance with the PEP8 standard.
• Development of the Python API documentation with examples of how to use the module, using sphinx and sphinx-gallery technology.
• Improved code performance by adding parallelization.
• Implementation of inference of distribution tails using the Generalized Pareto-Distribution (GPD).