A convolutional neural network (CNN) is proposed to learn multiple useful feature representations for a classification from low level (raw pixels) high (object).Convolutional kernels are initialized by the learned filter that come sparse auto-encoders.Unlike some traditional methods, which divide abstracting and classifier training into two separated processes, discriminative vector single multi-class of softmax regression simultaneously during process.Based on high-quality representation,...

Abstract Aiming at the characteristics of orthotropic anisotropy, multiangle and multilayer carbon fiber composites, based on relevant theories mechanics materials a method calculating strain in main direction each ply inside composite laminate by using measured reverse order surface layer is proposed. In this study, laminates with different lay-up angles thicknesses are selected as test objects, accuracy gauge measurement rotational axis calculation verified comparing values extensometer...

Iterative learning control (ILC) can synthesize the feedforward signal for trajectory tracking of a repetitive task, even when system has strong nonlinear dynamics. This makes ILC be one most popular methods control. Restriction on however, limits its application to multiple trajectories. article proposes neural-network-based (NN-ILC) deal with nonrepetitive tasks very effectively. A position-based is designed compensate error, based which outputs (ILC outputs) are expressed as function...

Sintered silver pastes have emerged as a kind of packaging material for wide-bandgap (WBG) power devices because their superior electrical and thermal conductivity. Herein, containing monodisperse triangular nanoflakes with size distribution 40–260 nm were prepared. The sintered joints exhibited significant high shear strength 49.8 MPa, larger than that traditional paste, relatively low porosity 5.1%. These remarkable properties attributed to the special anisotropic sintering process...

Linear motion commands of multi-axis computer numerical control (CNC) machine tools need to be smoothed at the transition corners, because velocity discontinuities corners can result in fluctuations on tool motions and lead poor surface quality. However, no research has been reported local corner smoothing algorithm for four-axis CNC with two rotary axes by considering their special kinematic characteristics. To this end, paper proposes an analytical C3 continuous machines axes. After...

During tunneling processes, disc cutters of a tunnel boring machine (TBM) usually need to be frequently and unexpectedly replaced. Regular inspections are needed check cutters’ status, which significantly reduces the work efficiency increases cost. This paper proposes new prediction model based on TBM operational parameters geological conditions that determines whether cutter replacement is needed. Firstly, an evaluation criterion for replaced constructed. Secondly, specific related analyzed...

Rolling bearing fault diagnosis is one of crucial tasks in mechanical equipment diagnosis. Currently, artificial intelligence and machine learning-driven methods are extensively utilized for rolling bearing. When compared to traditional techniques, the diagnostic accuracy has significantly improved. These methods, however, need a substantial amount labelled training data, which difficult obtain actual failures. In order resolve this problem, Transfer Learning (TL) was created learn target...

Objective To investigate the characteristics of response inhibition overweight/obese people, using behavior experiments combine with neural electrophysiological technology and discussing difference in impulse level between obesity/overweight normal-weight people through EEG data, questionnaire, experiment. Method (1) All participants completed Go/Nogo task; meanwhile, data 64 channel were recorded. (2) Participants Stop-Signal task date was Results During task, no significant differences...

The position angle is very important in the current control of permanent-magnet linear motors (PMLMs), and its error affects precision currents. However, often not accurately detected because faults motor making, pulse loss encoders, so on. In order to precisely currents PMLMs, a compensation scheme errors proposed based on mathematical relation between scheme, compensator designed compensate angles by encoders. Furthermore, filter class iteration learning algorithm are avoid produced phase...

Photonic crystal (PC)-based devices have been widely used since 1990s, while PC has just stepped into the research area of nanofluidic. In this paper, photonic had as a complementary metal oxide semiconductors (CMOS) compatible part to create nanofluidic structure. A structure prototype fabricated with CMOS-compatible techniques. The channels were sealed by direct bonding polydimethylsiloxane (PDMS) and periodic gratings on was 4-in. Si wafer Si3N4 guided mode layer SiO2 film substrate...

In a tunneling boring machine (TBM), to obtain the attitude in real time is very important for driver. However, current laser targeting system has large delay before obtaining attitude. So, an adaptive-neuro-fuzzy-based information fusion method proposed predict of time. method, dual-rate used fuse and two-axis inclinometer, then roll pitch angles with higher rate provide smoother prediction. Considering that measurement error exists, adaptive neuro-fuzzy inference (ANFIS) model error,...

A nanofluidic biosensor based on nanoreplica molding photonic crystal (PC) was proposed. UV epoxy PC fabricated by a master wafer. The nanochannels were sealed between the gratings surface and taped layer. resonance wavelength of PC-based used for testing sealing effect. According to peak value sensor, an initial label-free experiment realized with R6g as analyte. When illuminated monochromatic light source specific angle, sensor will match amplified electromagnetic field. field enhance...

As one of the key components for active compliance control and human–robot collaboration, a six-axis force sensor is often used robot to obtain contact forces. However, significant problem distortion between forces data conveyed by because its zero drift, system error, gravity end-effector. To eliminate above disturbances, an integrated compensation method proposed, which uses deep learning network least squares realize zero-point prediction tool load identification, respectively. After...

An approach to a class-specific and shared dictionary learning (CDSDL) for sparse representation is proposed classify surface defects of steel sheet. The CDSDL algorithm modelled as unified objective function, covering reconstructive error, discriminative promotion constraints. With the high-quality dictionary, compact, feature an image can be extracted. Then classification efficiently performed by information obtained from error or vector. Based on dataset images captured practical...

The fiber optic gyroscope (FOG) is a high precision inertial navigation device, and it necessary to ensure its reliability for effective use. However, the extracted fault features are easily distorted due interference of vibrations when FOG in operation. In order minimize influence greatest extent, fusion diagnosis method was proposed this paper. It from data with Fast Fourier Transform (FFT) wavelet packet decomposition (WPD), built strong diagnostic classifier sparse auto encoder (SAE)...