Case Studies in using Consumer Analytics to drive PHM Strategy

Sameer Vittal and Mark Sporer
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_17_040.pdf978.85 KBSeptember 15, 2017 - 9:35am

As part of the “Digital-Industrial Revolution”, the world is seeing the rapid transformation and digitization of the world’s energy value network – from generation, through transmission & distribution, to end user consumption. This new paradigm comprises of new business products and services built on data flows that accompany energy flows; where the insight gained from sensors and analytics drives better decision making and customer outcomes. This is what drives the digital strategies of Original Equipment Manufacturers of large industrial assets like power plants, oil & gas equipment, aviation fleets, etc.

In this paper, we look at how analytical methods originally developed in the consumer industry can be applied to industrial data. This helps guide the development of Prognostics & Health Management strategies that are tuned to customer preferences and value models, in addition to engineering inputs. These methods complement, rather than replace, FMEA-driven strategies that are traditionally used in PHM systems design.

Traditional PHM systems are designed “bottom up”, starting with Failure Modes and Effects Analysis, and then progressing through a series of trade studies where sensors, anomaly detection and remaining life algorithms are selected and integrated to reduce unplanned failures from specific failure modes. The methods so not typically consider marketing or survey data, qualitative customer information, or other exogenous economic variables that are needed to “sell” the PHM system and realize it’s true value. The availability of retail and e-commerce generated data on the other hand, has led to the maturity of many consumer analytics techniques like Latent Class Analysis, Text Mining, Multiple Correspondence Analysis, Choice and Conjoint models, etc., that work with traditional data mining classification and clustering methods to parse out preferences and differentiate products from a customer’s perspective. These methods and workflows can also be applied to industrial data, and can help drive PHM systems architectures that can be customized to consumer segments, increasing their adoption, usage and ultimately, business value.

In this paper, we provide an overview of consumer analytics techniques that are relevant to industrial data, and show how they can be applied via two cases studies. The first deals with usage-based segmentation analysis of coal-fired power plants, and the second is based on risk-based segmentation of pitch system failures observed in wind turbine fleets. In both cases, we offer insights that would not be available using traditional PHM design methods. Finally, it is hoped that theses case studies would motivate the use of consumer analytics within the broader toolkit of PHM system design methods.

Publication Year: 
2017
Publication Volume: 
8
Publication Control Number: 
040
Page Count: 
8
Submission Keywords: 
PHM
Consumer Analytics
Latent Class Analysis
Fleet Management
internet of things
Submission Topic Areas: 
Economics and cost-benefit analysis
Health management system design and engineering
Industrial applications
Submitted by: 
  
 
 
 

follow us

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