Article
Deep Auto-Encoder and Deep Forest-Assisted Failure Prognosis
for Dynamic Predictive Maintenance Scheduling
Hui Yu
1
, Chuang Chen
2
, Ningyun Lu
2,
* and Cunsong Wang
3
Citation: Yu, H.; Chen, C.; Lu, N.;
Wang, C. Deep Auto-Encoder and
Deep Forest-Assisted Failure
Prognosis for Dynamic Predictive
Maintenance Scheduling. Sensors
2021, 21, 8373. https://doi.org/
10.3390/s21248373
Academic Editor: Steven Chatterton
Received: 27 November 2021
Accepted: 14 December 2021
Published: 15 December 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Integrated System Integration Department, No. 38 Research Institute of CETC, Hefei 230088, China;
yuhuihustac@163.com
2
College of Automation Engineering, Nanjing University of Aeronautics and Astronautics,
Nanjing 211106, China; chenchuang@nuaa.edu.cn
3
Institute of Intelligent Manufacturing, Nanjing Tech University, Nanjing 210009, China;
wangcunsong@njtech.edu.cn
* Correspondence: luningyun@nuaa.edu.cn
Abstract:
Prognostics and health management (PHM) with failure prognosis and maintenance
decision-making as the core is an advanced technology to improve the safety, reliability, and oper-
ational economy of engineering systems. However, studies of failure prognosis and maintenance
decision-making have been conducted separately over the past years. Key challenges remain open
when the joint problem is considered. The aim of this paper is to develop an integrated strategy
for dynamic predictive maintenance scheduling (DPMS) based on a deep auto-encoder and deep
forest-assisted failure prognosis method. The proposed DPMS method involves a complete process
from performing failure prognosis to making maintenance decisions. The first step is to extract repre-
sentative features reflecting system degradation from raw sensor data by using a deep auto-encoder.
Then, the features are fed into the deep forest to compute the failure probabilities in moving time
horizons. Finally, an optimal maintenance-related decision is made through quickly evaluating the
costs of different decisions with the failure probabilities. Verification was accomplished using NASA’s
open datasets of aircraft engines, and the experimental results show that the proposed DPMS method
outperforms several state-of-the-art methods, which can benefit precise maintenance decisions and
reduce maintenance costs.
Keywords:
failure prognosis; maintenance decision-making; deep auto-encoder; deep forest;
maintenance cost
1. Introduction
The rapid development of industrial Internet of Things technology greatly promotes
the complexity, integration, and intelligence of modern engineering systems [
1
–
5
]. It also
raises significant challenges in the safety and reliability of systems’ operation. Due to
the unavoidable degradation of various components caused by wearing, aging, fatigue,
functional design defects, and complicated environmental factors, the failure probability
of the whole system is high and the consequences may be intolerable [
6
–
10
]. Precise
predictive maintenance is an urgent need in those systems, especially for applications such
as nuclear power plants, missile weapons, and aerospace vehicles, which have extremely
high reliability and safety requirements. Therefore, prolonging the effective service life and
ensuring the reliability with less maintenance cost is of great significance in practice. The
key is to realize failure prognosis-based maintenance decision-making, that is, to predict
the failure probabilities and carry out just-in-time maintenance activities [11–15].
In recent decades, failure prognosis has received extensive attention, and many re-
search results have been achieved. For instance, an integrated feature-based failure prog-
nosis method was developed in [
16
], where the dynamics of various failures were detected
using signal processing technology, and the adaptive Bayesian algorithm was used to
Sensors 2021, 21, 8373. https://doi.org/10.3390/s21248373 https://www.mdpi.com/journal/sensors
