Reliability optimization design of hydraulic system considering oil contamination

Zhou, Xinda; Yang, Zhaojun; Tian, Hailong; Chen, Chuanhai; Wang, Liding; Zhu, Yongfu; Liu, Jia

doi:10.1007/s12206-020-1108-1

Cited by 3 publications

(1 citation statement)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Fault diagnoses for aircrafts' hydraulic systems involve four stepped stages: subjective diagnoses by technicians, measurement-based diagnoses using simple instruments, diagnoses using model analysis, and intelligent fault diagnoses [9][10][11]. Currently, measurements and diagnoses using simple instruments are mainly used for aircrafts' hydraulic systems.…”

Section: Introductionmentioning

confidence: 99%

Fault Diagnosis for Aircraft Hydraulic Systems via One-Dimensional Multichannel Convolution Neural Network

Shen

Zhao

2022

Actuators

View full text Add to dashboard Cite

Detecting the faults in hydraulic systems in advance is difficult owing to the complexity associated with such systems. Hence, it is necessary to investigate the different fault modes and analyze the system reliability in order to establish a method for improving the reliability and security of hydraulic systems. To this end, this paper proposes a novel one-dimensional multichannel convolution neural network (1DMCCNN) for diagnosing fault modes. In this work, a landing gear hydraulic system was constructed with a normal model and a fault model; five types of faults were considered. Pressure signals were extracted from this hydraulic system, and the extracted signals were subsequently input into the convolution neural network (CNN) as multichannel data. Thereafter, the data were subjected to a one-dimensional convolution filter. The differences between channels were used to enhance features. The features obtained in this manner were compared for fault diagnoses. Furthermore, this proposed method was verified via simulations; the simulation results indicated that the precision of the 1DMCCNN was considerably higher than that of conventional machine learning algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%