Accurate and timely prediction of remaining useful life (RUL) a machine enables the to have an appropriate operation maintenance decision. Data-driven RUL methods are more attractive researchers because they can be deployed quicker cheaper compared other approaches. The existing deep neural network (DNN) models proposed for applications mostly single-path top-down propagation. In order improve prognostic accuracy network, this paper proposes directed acyclic graph (DAG) that combines long...

This paper deals with gear pitting fault diagnosis problem and presents a method by integrating convolutional neural network (CNN) gated recurrent unit (GRU) networks vibration acoustic emission signals to solve the problem. The presented first trains one-dimensional CNN GRU signals. Then features obtained two are concatenated form deep learning structure for diagnosis. Seven different conditions used test feasibility of method. result shows that accuracy reaches above 98% only relatively...

Research on data-driven fault diagnosis methods has received much attention in recent years. The deep belief network (DBN) is a commonly used learning method for diagnosis. In the past, when people DBN to diagnose gear pitting faults, it was found that result not good with continuous time domain vibration signals as direct inputs into DBN. Therefore, most researchers extracted features from However, desirable use raw achieve results. this paper proposes novel by stacking spare autoencoder...

The prediction of remaining useful life (RUL) mechanical equipment provides a timely understanding the degradation and is critical for predictive maintenance equipment. In recent years, applications deep learning (DL) methods to predict RUL have attracted much attention. There are two major challenges when applying DL prediction: (1) It difficult select model structure hyperparameters such as network depth, rate, batch size, etc. (2) developed domain dependent, i.e., it can only give good...

Abstract Wafer defect classification is a key component in the wafer manufacturing process. Under stable operating conditions and sufficient test data, an effective model can help engineers quickly accurately judge solve problems production However, complexity of process leads to serious imbalance between various types defects, which greatly reduces performance traditional method. This paper proposes Jacobi regularized generative adversarial network (JRGAN) for sample imbalanced image...

In recent years, research on gear pitting fault diagnosis has been conducted. Most of the focused feature extraction and selection process, diagnostic models are only suitable for one working condition. To diagnose early faults under multiple conditions, this article proposes to develop a domain adaptation model–based improved deep neural network transfer learning with raw vibration signals. A particle swarm optimization algorithm L2 regularization used optimize improve stability accuracy...

Abstract Remaining useful life (RUL) prediction methods based on deep neural networks (DNNs) have received much attention in recent years. The collected time-series signals are usually processed by the sliding time window method into several segments with same sequence length as input. However, signal processing is not only time-consuming, but also relies too personal experience. Moreover, of affects results and range. Obviously, it more desirable to remove data use an entire series input...

Gear pitting fault is one of the most common faults in mechanical transmission. Acoustic emission (AE) signals have been effective for gear detection because they are less affected by ambient noise than traditional vibration signals. To overcome problem low recognition rate using AE and convolutional neural networks, this paper proposes a new method named augmented convolution sparse autoencoder (ACSAE) diagnosis raw First, proposed combines one-dimensional networks unsupervised learning...

Prognostics and health management (PHM) has received much attention as an emerging discipline. And the prediction of remaining useful life (RUL) is core PHM. The data-driven RUL methods are more favored because they can be developed faster cheaper. However, existing models usually only used under same data domain (DD), require a lot labeled to retrain new model. So adaptation model desirable. In this paper, proposed by integrating adaptive batch normalization (AdaBN) into deep convolutional...

Abstract Using deep learning to classify the time-frequency images of bearing vibration signals has become a mainstream method in field fault diagnosis. Most studies, however, assume constant rotational speed, and accuracy reliability diagnosis model diminishes once speed changes. Moreover, due large size high computational complexity, convolutional neural networks are not suitable for industrial applications. This paper proposes novel rotating machinery with variable based on multi-feature...

The early gear pitting fault diagnosis has received much attention in the industry. In recent decades, with popularity growth of artificial neural network, researchers have applied deep learning methods to figure out faults. However, classical usually use networks according time sequence collected signals. this case, feature extraction direction inverse time-domain signals is ignored. Aimed at overcoming shortage, ground on a traditional Long Short Term Memory (LSTM) paper proposes...

In recent years, the application of deep neural networks containing directed acyclic graph (DAG) architectures in mechanical fault diagnosis has achieved remarkable results. order to improve ability networks, researchers have been working on developing new network and optimizing training process. However, this approach requires sufficient time empirical knowledge try potential optimal framework. Furthermore, it is time-consuming laborious retune architecture hyperparameter values when faced...

Deep learning models, particularly those based on transformers, often employ numerous stacked structures, which possess identical architectures and perform similar functions. While effective, this stacking paradigm leads to a substantial increase in the number of parameters, posing challenges for practical applications. In today's landscape increasingly large depth can even reach dozens, further exacerbating issue. To mitigate problem, we introduce LORS (LOw-rank Residual Structure). allows...

In the field of emotion recognition and Human-Machine Interaction (HMI), personalised approaches have exhibited their efficacy in capturing individual-specific characteristics enhancing affective prediction accuracy. However, personalisation techniques often face challenge limited data for target individuals. This paper presents our work on an enhanced strategy, that leverages augmentation to develop tailored models continuous valence arousal prediction. Our proposed approach, Distance...

Tool wear prediction is of great significance in industrial production. Current tool methods mainly rely on the indirect estimation machine learning, which focuses more estimating current state and lacks effective quantification random uncertainty factors. To overcome these shortcomings, this paper proposes a novel method for predicting cutting wear. In offline phase, multiple degradation features were modeled using Brownian motion stochastic process SVR model was trained mapping values....