Sequence Adaptation Adversarial Network for Remaining Useful Life Prediction Using Small Data Set

Remaining useful life prediction is the core of condition-based maintenance under the technology framework of prognostic and health management. But the remaining useful life of airborne fuel pump after maintenance is difficult to predict because of the multi-stage noise and small data size. A new method is proposed to solve the remaining useful life prediction of repaired fuel pump. Firstly, an alternative smooth transition auto-regression model logistic smooth transition auto-regression or exponential smooth transition auto-regression is proposed to reduce the multi-stage noise. Secondly, random effect Wiener process is utilized to model the de-noised degradation data, and the posterior parameters of remaining useful life prediction after maintenance are derived by the Bayesian method based on the parameters before maintenance. Finally, the method proposed above is compared with the methods which neglect the multi-stage noise and information before maintenance, comparative results show that the proposed method can improve the remaining useful life prediction accuracy significantly.

Download Full-text

A Deep Adversarial Approach Based on Multi-Sensor Fusion for Semi-Supervised Remaining Useful Life Prognostics

Sensors ◽

10.3390/s20010176 ◽

2019 ◽

Vol 20 (1) ◽

pp. 176 ◽

Cited By ~ 4

Author(s):

David Verstraete ◽

Enrique Droguett ◽

Mohammad Modarres

Keyword(s):

Sensor Fusion ◽

Asset Management ◽

Life Prediction ◽

Health Management ◽

Remaining Useful Life ◽

Rolling Element Bearing ◽

Sensor Data ◽

Data Sets ◽

Data Set ◽

Useful Life

Multi-sensor systems are proliferating in the asset management industry. Industry 4.0, combined with the Internet of Things (IoT), has ushered in the requirements of prognostics and health management systems to predict the system’s reliability and assess maintenance decisions. State of the art systems now generate big machinery data and require multi-sensor fusion for integrated remaining useful life prognostic capabilities. When dealing with these data sets, traditional prediction methods are not equipped to handle the multiple sensor signals in unison. To address this challenge, this paper proposes a new, deep, adversarial approach to any remaining useful life prediction in which a novel, non-Markovian, variational, inference-based model, incorporating an adversarial methodology, is derived. To evaluate the proposed approach, two public multi-sensor data sets are used for the remaining useful life prediction applications: (1) CMAPSS turbofan engine dataset, and (2) FEMTO Pronostia rolling element bearing data set. The proposed approach obtains favorable results when against similar deep learning models.

Download Full-text

Remaining useful life prediction for the electro-hydraulic actuator based on improved relevance vector machine

Proceedings of the Institution of Mechanical Engineers Part I Journal of Systems and Control Engineering ◽

10.1177/0959651819862948 ◽

2019 ◽

Vol 234 (4) ◽

pp. 501-511

Author(s):

Runxia Guo ◽

Zhenghua Liu ◽

Jiaqi Wang

Keyword(s):

Computational Complexity ◽

Flight Control ◽

Computational Efficiency ◽

Life Prediction ◽

Relevance Vector Machine ◽

Remaining Useful Life ◽

Hydraulic Actuator ◽

Data Set ◽

On Line ◽

Useful Life

The electro-hydraulic actuator plays a significant role in the automatic flight control system, which is critical to the flight security, so it is vital to monitor the state of health and predict the remaining useful life for the electro-hydraulic actuator. Relevance vector machine is flourishing in the field of remaining useful life prediction and gradually applied to the prediction of complex systems or components. However, the general relevance vector machine cannot achieve on-line prediction efficiently due to its high computational complexity, besides, the sparse relevance vector machine model which is only based on historical data set could cause a great prediction error in the long term. To deal with these plights, an optimized incremental learning and on-line training algorithm based on the relevance vector machine is presented taking full advantage of the on-line updating samples to improve the precision of prediction, and in order to remedy the problem of computational complexity and improve the computational efficiency, the “sample entropy” is introduced as an effective signature of the electro-hydraulic actuator’s health to effectively reduce the size of training samples. The effectiveness of the proposed on-line training approach is evaluated through an experiment of an electro-hydraulic actuator, and the results show a satisfactory prediction accuracy as well as an improvement in the computational efficiency for remaining useful life prediction.

Download Full-text

Remaining useful life prediction with insufficient degradation data based on deep learning approach

Eksploatacja i Niezawodnosc - Maintenance and Reliability ◽

10.17531/ein.2021.4.17 ◽

2021 ◽

Vol 23 (4) ◽

pp. 745-756

Author(s):

Yi Lyu ◽

Yijie Jiang ◽

Qichen Zhang ◽

Ci Chen

Keyword(s):

Deep Learning ◽

Short Term Memory ◽

Time Window ◽

Synthetic Data ◽

Remaining Useful Life ◽

Generative Adversarial Network ◽

Data Set ◽

Adversarial Network ◽

Degradation Data ◽

Useful Life

Remaining useful life (RUL) prediction plays a crucial role in decision-making in conditionbased maintenance for preventing catastrophic field failure. For degradation-failed products, the data of performance deterioration process are the key for lifetime estimation. Deep learning has been proved to have excellent performance in RUL prediction given that the degradation data are sufficiently large. However, in some applications, the degradation data are insufficient, under which how to improve the prediction accuracy is yet a challenging problem. To tackle such a challenge, we propose a novel deep learning-based RUL prediction framework by amplifying the degradation dataset. Specifically, we leverage the cycle-consistent generative adversarial network to generate the synthetic data, based on which the original degradation dataset is amplified so that the data characteristics hidden in the sample space could be captured. Moreover, the sliding time window strategy and deep bidirectional long short-term memory network are employed to complete the RUL prediction framework. We show the effectiveness of the proposed method by running it on the turbine engine data set from the National Aeronautics and Space Administration. The comparative experiments show that our method outperforms a case without the use of the synthetically generated data.

Download Full-text

Dynamic Modeling and Wear-Based Remaining Useful Life Prediction of High Power Clutch Systems

Tribology Transactions ◽

10.1080/05698190590927451 ◽

2005 ◽

Vol 48 (2) ◽

pp. 208-217 ◽

Cited By ~ 32

Author(s):

Matthew Watson ◽

Carl Byington ◽

Douglas Edwards ◽

Sanket Amin

Keyword(s):

Dynamic Modeling ◽

High Power ◽

Life Prediction ◽

Remaining Useful Life ◽

Useful Life

Download Full-text

Remaining Useful Life Prediction Based on Stacked Sparse Autoencoder and Echo State Network

2020 39th Chinese Control Conference (CCC) ◽

10.23919/ccc50068.2020.9188550 ◽

2020 ◽

Author(s):

Yinghua Yang ◽

Dandan Yao ◽

Xiaozhi Liu

Keyword(s):

Life Prediction ◽

Remaining Useful Life ◽

Echo State Network ◽

Useful Life ◽

Sparse Autoencoder ◽

Stacked Sparse Autoencoder

Download Full-text

The Lyapunov Exponent Variance of an Electronic Manufacturing Enterprise’s Daily Qualified Rate Time Series by Improved Small Data Sets Approach

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.197.271 ◽

2012 ◽

Vol 197 ◽

pp. 271-277

Author(s):

Zhu Ping Gong

Keyword(s):

Time Series ◽

Lyapunov Exponent ◽

Chaotic Time Series ◽

Quality System ◽

Quality Level ◽

Largest Lyapunov Exponent ◽

Small Data ◽

Data Set ◽

Electronic Manufacturing ◽

Small Data Set

Small data set approach is used for the estimation of Largest Lyapunov Exponent (LLE). Primarily, the mean period drawback of Small data set was corrected. On this base, the LLEs of daily qualified rate time series of HZ, an electronic manufacturing enterprise, were estimated and all positive LLEs were taken which indicate that this time series is a chaotic time series and the corresponding produce process is a chaotic process. The variance of the LLEs revealed the struggle between the divergence nature of quality system and quality control effort. LLEs showed sharp increase in getting worse quality level coincide with the company shutdown. HZ’s daily qualified rate, a chaotic time series, shows us the predictable nature of quality system in a short-run.

Download Full-text