PSAM 16 - Paper Overview

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Lead Author: Maria Valentina Clavijo Mesa Co-author(s): - Sergio Cofre-Martel, email: scofre@umd.edu - Carmen Elena Patino-Rodriguez, email: elena.patino@udea.edu.co - Fernando Jesus Guevara Carazas, email: fjguevarac@unal.edu.co - Indra Gunawan, email: indra.gunawan@adelaide.edu.au - Enrique López Droguett, email: eald@g.ucla.edu

Asset Health Management using Deep Learning Techniques for Gas Engine
With the objective to reduce greenhouse emissions, efforts have gone to increase the integration of alternative fuels in the transportation sector such as natural gas. Using natural gas as automobile fuel has several advantages over petrol and diesel: lower costs, better combustion efficiency, and the possibility to produce it through a biomass conversion process. Natural gas-based engines have become a crucial asset in the South American transportation sector. However, the share of natural-gas vehicles in the current vehicle market is still estimated to be below 5%. Therefore, to incentivize investments and development in gas engines, it is imperative to ensure their reliability, availability, and sustainability by developing reliability analysis and identifying critical components, probability of failure, and better operating conditions. These assessments can later be used to design tailored maintenance policies, thus reducing maintenance and operational costs. This paper presents a deep learning-based prognostics analysis for gas engines installed in a fleet of heavy-duty trucks (HDTs) from a Colombian company. These HDTs operate under varying demand profiles, including continuous stops and runs, long trajectories, steep hills, and frequent load-unload cycles. Data is provided for 18 months of operation consisting of 14 operational variables. The dataset presents two challenges during the preprocessing stage, namely: the raw dataset does not include any kind of labels, and the sensors present an irregular sampling frequency. Thus, the analysis on this paper focuses on addressing these preprocessing challenges to later train prognostics models for the remaining useful life (RUL) estimation of the gas engine fleet. Results show that by implementing an adequate preprocessing methodology, promising results can be obtained for the engine’s RUL.

Asset Health Management using Deep Learning Techniques for Gas Engine

With the objective to reduce greenhouse emissions, efforts have gone to increase the integration of alternative fuels in the transportation sector such as natural gas. Using natural gas as automobile fuel has several advantages over petrol and diesel: lower costs, better combustion efficiency, and the possibility to produce it through a biomass conversion process. Natural gas-based engines have become a crucial asset in the South American transportation sector. However, the share of natural-gas vehicles in the current vehicle market is still estimated to be below 5%. Therefore, to incentivize investments and development in gas engines, it is imperative to ensure their reliability, availability, and sustainability by developing reliability analysis and identifying critical components, probability of failure, and better operating conditions. These assessments can later be used to design tailored maintenance policies, thus reducing maintenance and operational costs. This paper presents a deep learning-based prognostics analysis for gas engines installed in a fleet of heavy-duty trucks (HDTs) from a Colombian company. These HDTs operate under varying demand profiles, including continuous stops and runs, long trajectories, steep hills, and frequent load-unload cycles. Data is provided for 18 months of operation consisting of 14 operational variables. The dataset presents two challenges during the preprocessing stage, namely: the raw dataset does not include any kind of labels, and the sensors present an irregular sampling frequency. Thus, the analysis on this paper focuses on addressing these preprocessing challenges to later train prognostics models for the remaining useful life (RUL) estimation of the gas engine fleet. Results show that by implementing an adequate preprocessing methodology, promising results can be obtained for the engine’s RUL.

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Paper VA38 Preview

Author and Presentation Info

Download paper VA38.

Download the presentation pdf file.