Article
Automation & Control Systems
Le Yao, Bingbing Shen, Linlin Cui, Junhua Zheng, Zhiqiang Ge
Summary: In this article, a dynamic mixture variational autoencoder regression model is proposed to handle the multimode industrial process modeling with dynamic features. Furthermore, a semi-supervised mixture variational autoencoder regression model is introduced to deal with the partially labeled process data with rare quality values and large-scale unlabeled samples, where a corresponding semi-supervised data sequence division scheme is introduced. The proposed methods are applied to a numerical case and a methanation furnace case, and the results demonstrate their superior soft sensing performance compared to the state-of-the-art methods.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Chemistry, Analytical
Yu-Hsuan Tseng, Chih-Yu Wen
Summary: This paper proposes a novel approach for real-time human activity recognition using a wearable IMU sensor. By utilizing pre-trained XGBoost and CVAE models as the classifier and generator, respectively, the system achieves an accuracy of 96.03% in activity classification.
Article
Chemistry, Analytical
Yun-Chieh Fan, Yu-Hsuan Tseng, Chih-Yu Wen
Summary: This paper proposes a novel approach for indistinguishable activity recognition based on human wearable sensors. A multistage deep neural network framework is built to interpret accelerometer, gyroscope, and magnetometer data and generate realistic human activities. Transfer learning is also applied to enhance the performance of the target domain.
Article
Computer Science, Artificial Intelligence
Cong Geng, Jia Wang, Li Chen, Zhiyong Gao
Summary: Variational Autoencoder (VAE) and Generative adversarial network (GAN) are two classic generative models that generate realistic data from a predefined prior distribution. VAE has the advantage of generating high-dimensional data and learning useful latent representations. However, there is a tradeoff between reconstruction and generation in VAE, as matching the prior distribution for latent representations may destroy the geometric structure of the data manifold. To address this issue, we propose an autoencoder-based generative model that allows the prior to learn the embedding distribution. We provide theoretical and experimental support for the effectiveness of our method.
Article
Neurosciences
Jung-Hoon Kim, Yizhen Zhang, Kuan Han, Zheyu Wen, Minkyu Choi, Zhongming Liu
Summary: By training a variational autoencoder with rsfMRI data, researchers have been able to untangle the underlying sources of brain cortical activity and connectivity, representing spatiotemporal characteristics and driving changes in cortical networks. The resultant latent variables can be used as a reliable feature for accurate subject identification, even with limited data available. This demonstrates the value of VAE for unsupervised representation learning in resting state fMRI activity.
Review
Computer Science, Interdisciplinary Applications
Shiran Levy, Eric Laloy, Niklas Linde
Summary: In this study, we combine inverse autoregressive flows (IAF) with variational Bayesian inference (variational Bayes) for geophysical inversion using deep generative models with complex priors. Variational Bayes parameterizes the unnormalized posterior distribution within a given family of distributions through optimization. To enhance the expressiveness of the variational distribution, we explore its combination with IAFs, which transform samples from a base distribution onto an approximate posterior. The trained IAF provides a good reconstruction of channelized subsurface models for GAN and VAE-based inversions using synthetic crosshole ground-penetrating-radar data.
COMPUTERS & GEOSCIENCES
(2023)
Article
Engineering, Electrical & Electronic
Juan Zhang, Xiaoqi Lang, Bo Huang, Xiaoyan Jiang
Summary: This paper proposes an unpaired image-to-image translation method based on coupled generative adversarial networks (CoGAN) to solve low-level vision problems. The method, called VAE-CoGAN, introduces a shared-latent space and variational autoencoder (VAE) in its framework. The method has been evaluated using synthetic datasets and real-world images, and has shown favorable performance compared to state-of-the-art methods.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Jiyoung Song, Young Chul Lee, Jeongsu Lee
Summary: The study addresses the challenge of implementing deep learning technology for fault detection in manufacturing processes by proposing a novel approach that combines deep generative models and encoding of time series into images. The results show that the proposed method outperforms existing methods in distinguishing between good and defective products.
JOURNAL OF INTELLIGENT MANUFACTURING
(2023)
Article
Computer Science, Information Systems
Xuying Meng, Suhang Wang, Zhimin Liang, Di Yao, Jihua Zhou, Yujun Zhang
Summary: To address security concerns in communication networks, a semi-supervised anomaly detection framework called SemiADC is proposed, which improves the accuracy of anomaly detection through self-learning processes.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Changde Du, Changying Du, Huiguang He
Summary: The paper proposed a novel doubly semi-supervised multimodal learning (DSML) framework for the comprehensive utilization of incomplete multi-modality data. By using a modality-shared latent space and multiple modality-specific generators, DSML effectively associates multiple modalities together. Experimental results demonstrate that DSML outperforms baselines on tasks such as semi-supervised classification, missing modality imputation, and cross-modality retrieval.
INFORMATION FUSION
(2021)
Article
Computer Science, Artificial Intelligence
Hui Zhou, Xin Liu, Xiangju Li, Zhongying Zhao, Chao Li
Summary: Attributed graph embedding is an important research direction that aims to learn low-dimensional representations by leveraging both topological structure and attribute information. Existing algorithms often ignore high-frequency noise in features and do not normalize the embeddings, limiting their performance in downstream tasks. To address this problem, we propose an Adversarial Attributed Graph Embedding algorithm (AAGE) with a Laplacian smoothing filter. Experimental results demonstrate that AAGE outperforms state-of-the-art methods.
APPLIED INTELLIGENCE
(2023)
Article
Engineering, Chemical
Yifan Liu, Runze Liu, Jinyu Duan, Li Wang, Xiangwen Zhang, Guozhu Li
Summary: Commercial fuel discovery is facing decreasing return of investment due to stricter environmental criteria and reducing potential uses for each new fuel. This study proposes a deep generative model called LIGANDS, which screens desired fuel molecules in a large chemical space without manually setting design rules. LIGANDS integrates a variational autoencoder, a generative adversarial network, and a stacking model to generate new fuel molecules with similar properties and improved energy performance. The model imitates key properties of target fuel to expand and enrich the fuel-relevant chemical space with innovative molecular entities.
CHEMICAL ENGINEERING SCIENCE
(2023)
Article
Chemistry, Multidisciplinary
Ali Reza Sajun, Imran Zualkernan
Summary: This paper presents a survey method using Generative Adversarial Networks (GANs) for semi-supervised learning (SSL), analyzing and identifying the state-of-the-art GAN architecture for SSL. It also identifies future research opportunities involving the adaptation of SSL elements into GAN-based implementations.
APPLIED SCIENCES-BASEL
(2022)
Article
Engineering, Civil
Ruiqi Ren, Peixin Shi, Pengjiao Jia, Xiangyang Xu
Summary: In this study, we propose a semi-supervised learning approach based on generative adversarial networks for identifying pixel-level anomalous image segments in pavement distress detection. By using multiple style discriminators and an end-to-end mask channel, our approach is capable of maintaining background pixels, modifying anomalous foreground regions, and detecting pixel-level abnormal areas. Experiment results show that our approach achieves a high accuracy rate of 80.75% on the dataset without pixel-level or patch-level annotations, demonstrating its superiority over several prior semi-supervised methods in quantitative comparisons.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Mathematics
Claudia Pinheiro, Francisco Silva, Tania Pereira, Helder P. Oliveira
Summary: The use of deep learning methods in medical imaging has shown promising results but requires large annotated datasets. This study proposes a semi-supervised learning approach that combines labeled and unlabeled data to improve predictive model efficiency for predicting mutation status in lung cancer.
Article
Computer Science, Artificial Intelligence
Runze Lin, Junghui Chen, Lei Xie, Hongye Su
Summary: This paper proposes a novel RL control algorithm, CBR-MA-DDPG, which combines case-based reasoning, model-assisted experience augmentation, and deep deterministic policy gradient to improve adaptability and control performance in industrial process control applications.
Article
Engineering, Chemical
Pei Sun, Junghui Chen, Lei Xie, Hongye Su
Summary: Modelling is a basic requirement for model-based controlling, monitoring, or other process strategies. This study focuses on non-linear model predictive control (NMPC) and proposes an actively improved Gaussian process (GP) model building strategy for incomplete models. The proposed method utilizes Bayesian optimization and expected improvement strategy to efficiently build models with insufficient initial training data for NMPC. It also considers multi-step ahead prediction and a novel disturbance rejection strategy based on GP outputs. Simulation results demonstrate the effectiveness of the proposed method compared to traditional algorithms.
CANADIAN JOURNAL OF CHEMICAL ENGINEERING
(2023)
Article
Automation & Control Systems
Wangwang Zhu, Zhengjiang Zhang, Junghui Chen, Yi Liu, Tao Xia, Antonios Armaou, Sheng Zhao
Summary: For a stochastic PID feedback control system, the uncertainty of the working environment often leads to unsatisfactory performance. This paper focuses on the performance degradation caused by Gaussian/non-Gaussian disturbances and measurement noise. A method called dynamic data reconciliation (DDR) is proposed to filter out noise and disturbances and improve the system's performance.
Article
Engineering, Environmental
Jingxiang Liu, Guan-Yu Hou, Weiming Shao, Junghui Chen
Summary: A supervised transfer-learning based functional Bayesian inference method is developed to improve the monitoring performance of batch processes with nonlinearity, uneven-length, and limited-data issues. The raw uneven-length batch data is transformed into functional data using orthogonal wavelet basis functions. Gaussian process models and Bayesian inference methods are applied to the latent features represented by the approximation coefficients, and the established models are transferred to enhance modeling performance for the target process with limited batches. The proposed functional method avoids distortion of the raw data structure and enables effective within-batch detection.
PROCESS SAFETY AND ENVIRONMENTAL PROTECTION
(2023)
Article
Computer Science, Artificial Intelligence
Kai Wang, Junghui Chen, Zhihuan Song, Yalin Wang, Chunhua Yang
Summary: The proposed process monitoring model uses deep neural networks to effectively handle the complexities of nonlinearity, dynamics, and uncertainties, outperforming other comparative methods by at least 10% in industrial experimental data.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Chemical
Shaowu Gu, Junghui Chen, Lei Xie
Summary: This study proposes an Auto-Segmentation Subspace Identification (AS-SID) method for modeling a nonlinear batch process. AS-SID automatically reinforces data that pertains to the corresponding local models and weakens data that does not pertain to other local models. Multiple state-space models are constructed using the partitioned data to model the entire operation range. AS-SID is also used for online detection by monitoring the running batch using only the online collected data.
CHEMICAL ENGINEERING SCIENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Jingxiang Liu, Guoqing Mu, Junghui Chen
Summary: Establishing efficient monitoring models for complex batch processes is challenging due to the 3-D data array and serious dynamics. Existing methods often destroy the batch data structure and increase data dimensions. This research proposes a novel high-order tensor slow feature analysis model to simultaneously handle the 3-D and dynamical issues. The model tackles two sub-optimal problems iteratively and performs within-batch detection based on monitoring statistics, considering the multi-phase property for improved accuracy. Simulated and industrial cases demonstrate the merits of the proposed method.
COMPUTERS & CHEMICAL ENGINEERING
(2023)
Article
Automation & Control Systems
Mifeng Ren, Yan Liang, Junghui Chen, Xinying Xu, Lan Cheng
Summary: With the era of big data, data-driven models are increasingly vital to just-in-time decision support in pollution emission management and planning. This article evaluates the usability of a proposed data-driven model for monitoring NOx emissions from a coal-fired boiler process using easily measured process variables. The proposed SIP-PCA model extracts more information from non-Gaussian distributed process variables and enables real-time detection of possible failures to prevent excessive NOx emissions.
Article
Automation & Control Systems
Weiming Shao, Xu Li, Yating Yao, Junghui Chen, Dongya Zhao
Summary: Data-driven soft sensors play a crucial role in predicting key quality variables in the process industry. This paper proposes a semi-supervised manifold regularization model based on dual representation (SsMRM-DR) and a semi-supervised local manifold regularization model based on dual representation (SsLMRM-DR) to address the challenges of limited labeled samples and strong nonlinear relationships between variables. The proposed models improve prediction accuracy and computational efficiency by reducing the influence of measurement noise and utilizing unlabeled samples effectively.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2023)
Article
Thermodynamics
Li Zhu, Meng-Qin Yu, Junghui Chen
Summary: In this paper, a novel cyber-physical energy monitoring and diagnosis scheme is proposed for large-scale plant-wide chemical processes. The scheme constructs a cyber-physical model based on process knowledge and data, and uses a distributed monitoring approach for energy state estimation and fault diagnosis. The effectiveness and practicality of the proposed scheme are demonstrated through numerical simulation and practical production examples.
ENERGY CONVERSION AND MANAGEMENT
(2023)
Article
Automation & Control Systems
Zhengjiang Zhang, Junghui Chen, Xiaofei Wu, Lei Xie, Chun-I. Chen
Summary: Data reconciliation and gross error detection (DRGED) is important for improving the precision and reliability of process data in the semiconductor industry. This study develops a novel correntropy estimator based iterative neural network (C-INN) for DRGED, which effectively solves the problem in complex semiconductor processes.
JOURNAL OF PROCESS CONTROL
(2023)
Article
Engineering, Electrical & Electronic
Ji Fan, Tao Liu, Yan Huo, Yonghong Tan, Junghui Chen
Summary: In this article, a novel in situ measurement method with stereoscopic image analysis is proposed to monitor crystal length and width distributions during a cooling crystallization process. The method includes establishing a stereoscopic imaging calibration model, improving crystal image matching using an enhanced algorithm, determining key corners related to crystal dimensions, and computing crystal length and width based on reconstructed key corners in a 3-D coordinate space. Experimental validation on a microscale checker-board plate and betta-form L-glutamic acid (LGA) crystals is conducted to verify the effectiveness and advantage of the proposed measurement method.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Automation & Control Systems
Pei Sun, Junghui Chen, Haoyi Que
Summary: This paper introduces a new deep learning method for modeling spatial-temporal industrial processes. The proposed model incorporates both domain knowledge and physical rules, and utilizes multiple RNNs to model spatial and temporal relationships. It achieves more accurate predictions with fewer parameters.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Proceedings Paper
Automation & Control Systems
Lee Yi Shan, Junghui Chen
Summary: This paper proposes a semi-supervised latent dynamic variational autoencoder (S-2-LDVAE) to learn features between the process and quality data. It also addresses the issue of the uneven length of the process and quality data. In the case of missing quality data, artificial data generated by the trained prediction network are used for online quality estimation.
2022 IEEE INTERNATIONAL SYMPOSIUM ON ADVANCED CONTROL OF INDUSTRIAL PROCESSES (ADCONIP 2022)
(2022)
Article
Engineering, Electrical & Electronic
Guoqing Mu, Tao Liu, Junghui Chen, Chao Shang, Chongquan Zhong
Summary: A semi-supervised calibration model is proposed to measure the moisture content of granules during the industrial fluidized bed drying process using near-infrared spectroscopy. The model utilizes both labeled and unlabeled spectra to overcome the lack of labeled samples in batch FBD processes. An adaptive Gamma distribution-based sparsing algorithm is used to select spectral variables for modeling and overcome high-dimensional input collinearity.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Automation & Control Systems
Haifei Peng, Jian Long, Cheng Huang, Shibo Wei, Zhencheng Ye
Summary: This paper proposes a novel multi-modal hybrid modeling strategy (GMVAE-STA) that can effectively extract deep multi-modal representations and complex spatial and temporal relationships, and applies it to industrial process prediction.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2024)