Abstract
The problem we were trying to solve in 2013 PHM Society Conference Data Challenge competition 1 is closely related to remote monitoring and diagnostics in industrial applications. This data was generated from an industrial piece of equipment with a sensor network to measure several parameters and an onboard condition monitoring system. The measured data goes through a control logic in order to monitor the equipment’s operating regime. At any time instant when some of these parameters meet a specific condition, the control system generates an unique event id/code. Each case is described by a set of event codes which characterize the atypical operating condition of the equipment. Some of these cases with specific event code combinations may be operationally significant and could be indicative of “Problem Types”, some of which are assumed to be known to the subject matter experts. As a response to these problems, domain experts recommend appropriate diagnostic measures (or maintenance actions) depending on the problem types. The goal of this data competition is to build an automated system that can recommend particular maintenance action(s) to mitigate these problem(s).
Highlights
Though active research has been conducted in recommender systems for the last two decades, more recently this field has gained tremendous attention in service industry
It should be noted that the training data does not contain any examples for P 0932 and P 6880 types present in the test set and the model is not expected to make any kinds to predictions on these types as they have been anyway eliminated from Fcters while computing Zetsct during the preprocessing stage
It requires some experimentation with different variations of the above mentioned cost function along with an appropriate set of features and correct parameter settings before we obtain the model that best fits this data set
Summary
Though active research has been conducted in recommender systems for the last two decades, more recently this field has gained tremendous attention in service industry. Filtering is synonymous to estimating Fu(c, u) using either item’s content based features e.g., keyword weights or user’s explicit feedback on items e.g., ratings. Content filtering assumes access to item’s content (Dc) In this approach the first step is content based information retrieval where a set of attributes/features (Ac) that best describes an item c are extracted. Once we obtain the key features (Ac) the remaining task is to adopt a heuristics or a model based approach to compute the score of the utility function Fu(Ac,Dcj ) over all new items’ content Dcj. The item with the highest score or the n top scoring items are recommended to the user. As mentioned earlier collaborative filtering based approach does not assume access to item’s content Dc, instead it expects user’s explicit feedback (ratings rc) on items i.e. useritem interactions (Yuc). The readers are advised to consult the following literatures (Koren & Bell, 2011; Koren et al, 2009; Adomavicius & Tuzhilin, 2005) for detailed description of variations on aggregation functions
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: International Journal of Prognostics and Health Management
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
