We develop a novel deep learning framework to achieve highly accurate machine fault diagnosis using transfer enable and accelerate the training of neural network. Compared with existing methods, proposed method is faster train more accurate. First, original sensor data are converted images by conducting Wavelet transformation obtain time-frequency distributions. Next, pretrained network used extract lower level features. The labeled then fine-tune higher levels architecture. This paper...

Deep learning (DL) architecture, which exploits multiple hidden layers to learn hierarchical representations automatically from massive input data, presents a promising tool for characterizing fault conditions. This paper proposes DL-based multi-signal diagnosis method that leverages the powerful feature ability of convolutional neural network (CNN) in images. The proposed deep model is able types sensor signals simultaneously so it can achieve robust performance and finally realize accurate...

A convolutional discriminative feature learning method is presented for induction motor fault diagnosis. The approach firstly utilizes back-propagation (BP)-based neural network to learn local filters capturing information. Then, a feed-forward pooling architecture built extract final features through these filters. Due the of BP-based network, learned can discover potential patterns. Also, able derive invariant and robust features. Therefore, proposed representation from raw sensory data...

Extracting features from original signals is a key procedure for traditional fault diagnosis of induction motors, as it directly influences the performance recognition. However, high quality need expert knowledge and human intervention. In this paper, deep learning approach based on belief networks (DBN) developed to learn frequency distribution vibration with purpose characterizing working status motors. It combines feature extraction classification task together achieve automated...

As a rising star in the field of deep learning, Transformers have achieved remarkable achievements numerous tasks. Nonetheless, due to safety considerations, complex environment, and limitation deployment cost actual industrial production, algorithms used for fault diagnosis often face three challenges limited samples, noise interference, lightweight, which is an impediment practice transformer with high requirements number samples parameters. For this reason, article proposes lightweight...

As one of the key components in mechanical systems, rotatory machine plays a significant role safe and stable operation. Accurate prediction Remaining Useful Life (RUL) contributes to realization intelligent operation maintenance for manufacturing. In order overcome limitations traditional learning algorithms dealing with complex nonlinear signals, novel framework RUL based on deep is proposed this paper. One-dimensional convolutional neural network utilized extract local features from...

High precision and fast fault diagnosis is an important guarantee for the safe reliable operation of machinery. In recent years, due to strong recognition ability, data-driven technology based on deep learning has attracted enormous attention. The module proposed by many scholars achieved excellent results, but some them are too complex deploy in practice, high costs. this article, efficient feature extraction method convolutional neural networks (CNN) was proposed, high-precision task...

This paper aims at proposing an abnormality detection framework for electrocardiogram (ECG) signals, which owns unbalance distribution among different classes and gaining high accuracy in rhythm/morphology abnormalities classification. The proposed is composed of two models: data augmentation model classification model. In this framework, designed to recast a class-balanced training dataset by generating artificial minor class. outputs are transferred into identify accurately after using...

Aiming at automated and intelligent state monitoring of induction motors, which are an integral component a broad spectrum manufacturing machines, this paper presents Deep Belief Network (DBN)-based approach to automatically extract relevant features from vibration signals that characterize the working condition motor. The DBN model employs structure with stacked restricted Boltzmann machines (RBMs), is trained by efficient learning algorithm called greedy layer-wise training. Vibration used...

Rolling bearings are some of the most crucial components in rotating machinery systems. bearing failure may cause substantial economic losses and even endanger operator lives. Therefore, accurate remaining useful life (RUL) prediction rolling is tremendous research importance. Health indicator (HI) construction critical step data-driven RUL approach. However, existing HI methods often require extraction time-frequency domain features using prior knowledge while artificially determining...

To realize high-precision and high-efficiency machine fault diagnosis, a novel deep learning framework that combines transfer transposed convolution is proposed. Compared with existing methods, this method has faster training speed, fewer samples per time, higher accuracy. First, the raw data collected by multiple sensors are combined into graph normalized to facilitate model training. Next, utilized expand image resolution, then images treated as input of for fine-tuning. The proposed...

Purpose The current studies on remaining useful life (RUL) prediction mainly rely convolutional neural networks (CNNs) and long short-term memories (LSTMs) do not take full advantage of the attention mechanism, resulting in lack accuracy. To further improve performance above models, this study aims to propose a novel end-to-end RUL framework, called recurrent network (CRAN) achieve high Design/methodology/approach proposed CRAN is CNN-LSTM-based model that effectively combines powerful...

Abstract The existing fault diagnosis algorithm based on domain adaptation solves the problem of degradation model performance due to different data distributions under variable working conditions and cross-machine conditions, its excellent relies assumption that category space source target domains is same; however, it difficult meet above in practical application scenarios. For this reason, focusing matter imbalance within category, paper proposes a novel unsupervised partial adaptational...

Deep learning has a strong feature ability, which proved its effectiveness in fault prediction and remaining useful life of rotatory machine. However, training deep network from scratch requires large amount data is time-consuming. In the practical model process, it difficult for to converge when parameter initialization inappropriate, results poor performance. this paper, novel framework proposed predict machine with high accuracy. Firstly, parameters ability pretrained are transferred new...

Postcholecystectomy bile duct injury (BDI) remains a devastating iatrogenic complication that adversely impacts the quality of life with high healthcare costs. Despite decrease in incidence laparoscopic cholecystectomy-related BDI, absolute number as cholecystectomy is commonly performed surgical procedure. Open Roux-en-Y hepaticojejunostomy meticulous technique gold standard procedure excellent long-term results most patients. As many hepatobiliary disorders, minimally invasive approach has...

Nowadays, most deep-learning-based bearing fault diagnosis methods are studied under the condition of steady speed, while performance these models cannot be fully played time-varying conditions. Therefore, in order to facilitate practical application a deep learning model diagnosis, vibration–speed fusion network is proposed, which utilizes transformer with self-attention module extract vibration features and sparse autoencoder (SAE) from speed pulse signal. The enables efficient different...

Abstract In the process of ensemble empirical mode decomposition (EEMD) for motor rolling bearing time series, if classifier is trained directly using eigenvalues extracted from pattern components, there are two shortcomings leading to reduction fault identification accuracy as follows: has serious endpoint effects; correlation between features lead confusion feature vector classification boundary. Aiming at problems, in this paper, a diagnosis model was built. Firstly, LSTM used extend...

Diesel fuel system is a significant part of the diesel engine, whose stable and reliable working state key guarantee for safety efficiency whole system. It essential to conduct health monitoring fault diagnosis based on intelligent technology. Recent years, deep learning has become an effective means perform various mechanical convolution neural networks have achieved remarkable success in applications. In order achieve efficient accurate system, improved network combined with attention...

Nowadays, deep learning has made great achievements in the field of rotating machinery fault diagnosis. But practical engineering scenarios, when facing a large number unlabeled data and variable operating conditions, only using algorithm may reduce performance. In order to solve above problem, this paper uses method combining transfer with learning. First, shrinkage residual network is constructed by adding soft thresholds extract characteristics bearing vibration under noise redundancy....

Abstract In recent years, intelligent fault diagnosis algorithms based on domain adaptation have provided a feasible solution to the problem of diagnosing performance degradation caused by different data distributions and lack target labels. However, most existing are highly dependent label space prior knowledge source domain, which greatly limits their application in practical scenarios. this paper, faced with circumstances that information mechanical device completely unknown, novel...