A collaborative data library for testing prognostic models

Joanna Z. Sikorska, Melinda R. Hodkiewicz, Ashwin D'Cruz, Lachlan Astfalck, and Adrian Keating
Submission Type: 
Full Paper
Supporting Agencies (optional): 
See acknowledgments in paper.
AttachmentSizeTimestamp
phmec_16_005.pdf379.21 KBJune 21, 2016 - 10:47pm

A web-based data management system for use by researchers and industry around the world to access suitable datasets for testing prognostic models is developed. The value of the project is in the provision of, and access to, real-world data for asset failure prediction work. In practice, it is difficult for researchers to obtain data from industrial equipment. Industry datasets are rarely shared and hardly ever published. When such data is made available, very little meta-data about the underlying asset is provided. This restricts the number and type of models that can be applied.

The solution is a data management system for three groups: researchers needing datasets, industry and academics with datasets. This paper identifies the data being sought, the system requirements and architecture, and discusses how the design is being implemented using an Agile development approach. Crucially, meta-data is stored in the database and accessed using a secure web-based front-end so as to maximize the available information, whilst obfuscating any corporate-sensitive material. The success of this prognostics data library depends on the support of the prognostic community to contribute and use the data; similar projects have been successful in the Machine Learning and Big Data communities.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
005
Page Count: 
11
Submission Keywords: 
relational database
benchmarking datasets
engineering prognostics
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Verification and validation
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed