A BiGRU Autoencoder Remaining Useful Life Prediction Scheme With Attention Mechanism and Skip Connection

Duan, Yuhang; Li, Honghui; He, Mengqi; Zhao, Dongdong

doi:10.1109/jsen.2021.3060395

“…Related scholars have conducted a series of analyses on the practice of smart cities in Europe [ 13 ]. Through technical methods such as data collation and model establishment, the relevant technical strategies of smart city construction are summarized and applied to the planning and construction project of Chongming Smart Ecological Island [ 14 ]. The big data and smart city research team has made systematic research in related fields from big data application in urban planning, urban and rural planning, and design based on big data to urban and rural special planning and urban and rural evaluation management based on big data [ 15 ].…”

Section: Related Workmentioning

confidence: 99%

Prediction of Urban Street Public Space Art Design Indicators Based on Deep Convolutional Neural Network

Yu

¹

,

Wang

²

2022

Computational Intelligence and Neuroscience

1

0

View full text Add to dashboard Cite

This paper analyzes and studies the structure and parameters of the VGGNet network model and selects the most commonly used and efficient VGG-16 as the prototype of the improved model. A multiscale sampling layer is added at the end of the VGG-16 convolution part so that the model can input images of any size for training and testing while reducing the number of neurons in the fully connected layer. This improves the training speed of the model under the premise of ensuring the accuracy. This paper uses multisource street spatial data combined with geographic information spatial analysis technology to measure and evaluate the spatial quality of streets in the main urban area. From the three dimensions of vitality, safety, and greenness of urban street space quality, a systematic structure for evaluation and analysis of street space quality is constructed. Street vitality includes eight index factors: entrance and exit density, street furniture density, street sketch density, street characteristic landscape density, POI density, POI diversity, commercial POI ratio, and street population density. There are five index factors: degree, roadside parking occupancy ratio, traffic signal system density, sidewalk width proportion, and street facility density. We use ArcGIS to build an index factor information database for statistical analysis and visualization. According to the natural discontinuous point classification method, the safety level of urban street public space is divided into five grades. The sample size of the first four grades has a small fluctuation range. The sample sizes are 153, 172, 153, and 158, respectively, accounting for 21%, 23%, 21%, and 21% of the total street samples, of which the first two grades occupy a total of 44%, so 44% of the streets in the main urban area have a low-quality level of street space. Level 5 has a sample of 102 streets, accounting for 14%, with an average street space quality value of 0.43.

show abstract

“…In [25] and [26], authors took a same approach as in [16] but replaced LSTM architecture in the encoder and decoder with bi-directional recurrent networks, BiLSTM and BiGRU respectively. Both works reported enhancements in prediction accuracy than [16].…”

Section: B 2-step Hi-based Approachesmentioning

confidence: 99%

Improving Tool Wear Prediction with Synthetic Features from Conditional Generative Adversarial Networks

Feng¹

2022

Preprint

2

0

View full text Add to dashboard Cite

<p>Amid the current digital transformation wave, predictive maintenance (PdM) using machine learning has become prevalent due to its potential to reduction for total cost and turnaround time in the manufacturing and MRO (maintenance, repair, and operations) sectors. Predicting machinery tool wear or remaining useful life is one of the essential applications in modern PdM. As a result, many deep learning based prediction methods emerged in recent years’ literature and continuously advancing the state-of-the-art. In this paper, we propose a tool wear prediction framework that orchestrates a range of machine learning techniques to enhance prediction accuracy. These techniques include sequence modeling with convolutional neural networks, synthetic data generation with conditional generative adversarial networks, and customized regularization strategy. Experiments were performed with two public milling datasets to evaluate the effectiveness of the proposed framework. Results show that our prediction model, sCNN-Ex, outperforms state-of-the-art tool wear prediction methods.</p>

show abstract

“…In stride step size 1, 573 sequences were obtained from each cutting run. When the flank wear value is greater than 0.45 mm for the first time, we set the true RUL value to 0, the same as the BiGRU-AS [26] A bold face indicates a better performance compared to the existing methods. The values here are averaged over all test cases.…”

Section: A Nasa Milling Datasetmentioning

confidence: 99%

Improving Tool Wear Prediction with Synthetic Features from Conditional Generative Adversarial Networks

Feng¹

2022

Preprint

0

View full text Add to dashboard Cite

<p>Amid the current digital transformation wave, predictive maintenance (PdM) using machine learning has become prevalent due to its potential to reduction for total cost and turnaround time in the manufacturing and MRO (maintenance, repair, and operations) sectors. Predicting machinery tool wear or remaining useful life is one of the essential applications in modern PdM. As a result, many deep learning based prediction methods emerged in recent years’ literature and continuously advancing the state-of-the-art. In this paper, we propose a tool wear prediction framework that orchestrates a range of machine learning techniques to enhance prediction accuracy. These techniques include sequence modeling with convolutional neural networks, synthetic data generation with conditional generative adversarial networks, and customized regularization strategy. Experiments were performed with two public milling datasets to evaluate the effectiveness of the proposed framework. Results show that our prediction model, sCNN-Ex, outperforms state-of-the-art tool wear prediction methods.</p>

show abstract

A BiGRU Autoencoder Remaining Useful Life Prediction Scheme With Attention Mechanism and Skip Connection

Cited by 87 publications

References 28 publications

Prediction of Urban Street Public Space Art Design Indicators Based on Deep Convolutional Neural Network

Prediction of Urban Street Public Space Art Design Indicators Based on Deep Convolutional Neural Network

Improving Tool Wear Prediction with Synthetic Features from Conditional Generative Adversarial Networks

Improving Tool Wear Prediction with Synthetic Features from Conditional Generative Adversarial Networks

Contact Info

Product

Resources

About