The estimation and prediction of state-of-health (SOH) state-of-charge (SOC) Lithium-ion batteries are two main functions the battery management system (BMS). In order to reduce computation cost enable deployment BMS on low-cost hardware, a Lebesgue-sampling-based extended Kalman filter (LS-EKF) is developed estimate SOH SOC. An LS-EKF able eliminate unnecessary computations, especially when states change slowly. this paper, first estimated remaining useful life predicted by LS-EKF. Then,...

Aiming to condition based maintenance for complex equipment, numerous intelligent fault diagnosis and prognostic methods on machine learning have been researched. Compared with the traditional shallow models, which problems of lacking expression capacity existing curse dimensionality, using deep theory can effectively mine characteristics accurately recognize health condition. In consequence, turned into an innovative promising research field. This paper gives a review learning. First all,...

Lithium-ion batteries play critical roles in many electronic devices. It is necessary to develop a reliable and accurate remaining useful life (RUL) prediction approach provide timely maintenance or replacement of battery systems. A fusion RUL based on Deep Belief Network (DBN) Relevance Vector Machine (RVM) proposed this paper. In the approach, DBN responsible for extracting features from capacity degradation lithium-ion batteries, RVM takes extracted as input prediction. The CALCE datasets...

Abstract Due to the heavy reliance on computers and networks, security issues have become a major concern for individuals, companies, nations. Traditional measures such as personal identification numbers, tokens, or passwords only provide limited protection. With development of intelligent keyboard (IKB), this paper proposes deep‐learning‐based keystroke dynamics method increased security. The IKB is kind self‐powered, nonmechanical‐punching keyboard, which converts mechanical stimuli...

In recent years, a notable development for predicting the remaining useful life (RUL) of components is prognostics that use data-driven approaches based on deep learning. particular, long short-term memory networks (LSTMNs) have been successfully applied in RUL prediction. However, to best our knowledge, these learning-based do not take into account uncertainty, and their prediction performance needs improvement. Bayesian model averaging (BMA) very ensemble method because it can quantify...

Bearings play a critical role in maintaining safety and reliability of rotating machinery. health condition prediction aims to prevent unexpected failures minimize overall maintenance costs since it provides decision making information for condition-based maintenance. This paper proposes Deep Belief Network (DBN)-based data-driven method bearings. In this method, DBN is used as the predictor, which includes stacked RBMs regression output. Our main contributions include development deep...

Lithium-ion batteries play critical roles in many electronic devices. It is necessary to develop a reliable and accurate remaining useful life (RUL) prediction approach provide timely maintenance or replacement of battery systems. A novel RUL based on Long Short-term Memory (LSTM) Recurrent Neural Network (RNN) proposed this paper. LSTM able capture long-term dependencies model sequential data among the capacity degradation lithium-ion batteries. The advantages our method include: 1)...

Information security plays critical roles in modern society. However, traditional measures like passwords, tokens and personal ID numbers only provide limited protection. Inspired by the fast training speed of extreme learning machine (ELM) promising feature extraction capability auto-encoder (ELM-AE), this paper proposes a keystroke dynamics identification method for intelligent keyboard (IKB) based on multi-layer (ML-ELM). The IKB, as first demonstrated Wang's group, is self-powered,...

Least Squares Support Vector Machines(LS-SVM), which is an efficient supervised learning tool, has been widely applied to real-time on-line data processing in many fields. However, the training of LS-SVM always suffers from huge computation greatly limits its practicability especially embedded systems. By leveraging flexibility and high degree parallelism offered by reconfigurable fabrics, we propose a Run-Time Reconfiguration(RTR) framework accelerate LS-SVM. To realize maximum...

Traditional analog circuit fault diagnosis usually consists of feature extraction, selection and classifier. However, extraction take much manual effort, which make the design system complicated not universal. This paper proposes an method based on Extreme Learning Machine (ELM). The proposed uses single pulse as stimulus signal to circuit-under-test (CUT), then raw time domain response CUT is directly sampled construct samples, are inputted into ELM network obtain results. experimental...

Fault diagnosis and prognosis (FDP) plays more important role in industries FDP aims to estimate current fault condition then predict the remaining useful life (RUL). Based on estimation of health state RUL, essential decisions maintenance, control, planning can be conducted optimally terms economy, efficiency, availability. With increase system complexity, it becomes difficult model dynamics, especially for multiple interacting modes that are affected by many internal external factors....

Traditional feature extraction methods for analog circuits incipient fault diagnosis rely on signal processing technology and expert experience, the diagnostic accuracy faults is not satisfactory. To deal with these problems, a novel approach based deep neural network presented in this paper. The proposed method includes stacked autoencoders softmax classifier. extract features from raw time-domain responses, then classifies mode of automatically. experiment results Sallen-Key band-pass...

A small simulator for AOS high-speed payload multiplexer is designed and implemented based on the idea of satellite hierarchical test, which will be used in test data transmission subsystem. The uses FPGA as control center, including multi-input multi-output LVDS RS422 interfaces. It caches received using large-capacity SRAM dual caching technology, selects downlink channel rate both can configured, a virtual (VC) scheduling algorithm Consultative Committee Space Data Systems (CCSDS)...

Traditional turbine engine health indicator (HI) construction methods generally require manual feature extraction, selection and even fusion, besides, training labels need to be designed in advance, which make the whole procedure time consuming not universal. Therefore, this paper proposes a novel unsupervised method of based on stacked denoising autoencoders (SDAE). In method, deep structure adaptively extracts features raw monitoring signals an way obtain its indicator. Experimental...

In recent years, machine learning technology is widely used in the field of fault diagnosis for bearings. Although these methods usually work well, following defects still exist when they are dealing with large amount data: (1) feature extraction need to rely on expertise or signal processing technologies. Therefore, there a lack method that common different diagnostic problems; (2) shallow models can't learn more complex mapping relationships well; (3) traditional intelligent...

Analog circuits play important roles in modern electronic systems. Incipient fault diagnosis of analog is a recognized challenging research direction due to the difficulty feature extraction and identification. This paper proposes an early algorithm for based on multilayer extreme learning machine (ML-ELM).Its basic idea comes from Auto-encoder(AE) Extreme Learning Machine (ELM). The proposed method which combines characteristics both, so it has ability fast training speed. In this method,...

The traditional FIR filters based on reconfiguration have disadvantages with difficult to control and low-level automation. In addition, the take long time configure. To solve these problems, a real-time reconfigurable filter is proposed which dynamic partial technology of EAPR multiply-accumulate structure. Finding common distinguish part by analyze transfer function within 1-15 order. Then are divided into static region region. design for FPGA implementation filter, supports up 121.265MHz...

The fusion approaches with multi-model ensemble can present a better performance than the simple single model in Prognostics and Health Management (PHM). Bayesian Model Averaging (BMA) is very useful method these because of its ability uncertainty quantification. A based on BMA relevance vector machine (RVM) presented this paper. Multi RVM models eight different kernel functions are constructed for degradation training data. Then, used to integrate sub-models into one framework reliability...

With the update and improvements of radar technology, testing has gained wider higher demand. Radar facilities are developing towards miniaturization high real time. The emergence ZYNQ, which is a kind all programmable SoC designed by Xilinx, allows us to design portable facility with performance based on it. ZYNQ composed two parts, processing system (PS) ARM Cortex-A9 logic (PL) Xilinx FPGA, convenient for building freely customizable embedded platform. This paper presents developed ZYNQ....