Data-Driven Two-Dimensional Deep Correlated Representation Learning for Nonlinear Batch Process Monitoring

Jiang, Qingchao; Yan, Shifu; Yan, Xuefeng; Yi, Hui; Gao, Furong

doi:10.1109/tii.2019.2952931

Cited by 54 publications

(21 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These data are used together to train a VAE model. Once the parameters of VAE are trained with the optimization objective in Equation (10), the latent variable z can be randomly sampled from N(0, I) . Then synthetic data are generated by the decoder p θ (xjz) through decoding the synthetic feature vectors into the original space.…”

Section: Vae Based Generative Modelmentioning

confidence: 99%

“…[1][2][3] ] The most widely used MSPM techniques include principal component analysis (PCA), partial least square (PLS), canonical correlation analysis (CCA), and their extensions. [4][5][6][7][8][9] Besides these methods, in recent years, deep learning methods, such as stacked autoencoders (SAE) [10] and deep belief networks (DBN), [11] have also been widely applied to the field of process monitoring and they have achieved considerable monitoring effects because of excellent feature extraction ability.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Synthesizing data by transferring information in data‐intensive regions to enhance process monitoring performance in data‐scarce region

et al. 2021

View full text Add to dashboard Cite

Sufficient data are necessary for valid process monitoring results. However, modern industrial processes sometimes switch to new modes to meet the changes in market demand. The available data in such a new mode are initially quite scarce and it brings huge obstacles to data-based model construction. In this paper, a novel data synthesis method based on variational autoencoders is proposed to generate synthetic data for the data-scarce region. The proposed method utilizes not only the original data in the data-scarce region but also the data in other data-intensive regions, which share some common information with the scarce data. To avoid model biases caused by the data imbalance between these regions, a model correction mechanism is also developed. Once the ultimate synthetic data of the data-scarce region are acquired, they are combined with the original data to establish a local monitoring model. Finally, the effectiveness of the proposed method is demonstrated through a real ammonia synthesis process.

show abstract

Section: Vae Based Generative Modelmentioning

confidence: 99%

mentioning

confidence: 99%

Synthesizing data by transferring information in data‐intensive regions to enhance process monitoring performance in data‐scarce region

et al. 2021

View full text Add to dashboard Cite

show abstract

“…During fault detection of the pitch system of the wind turbine generator system, an important step is to acquire the status parameters that effectively reflect the features of the pitch system from a lot of SCADA data. Due to the particularity of the SCADA system that involves the complex and diverse parameters of the pitch system, including strong coupling parameters, it is necessary for feature selection to optimize the model complexity to reduce the calculation time and the amount and select the effective status parameters and also to take redundant items into account to delete excess parameters and avoid model overfitting [27,28].…”

Section: Pitch System Of Wind Turbinementioning

confidence: 99%

Fault Detection of Wind Turbine Pitch System Based on Multiclass Optimal Margin Distribution Machine

Tang

Kuang

Zhao

et al. 2020

Mathematical Problems in Engineering

View full text Add to dashboard Cite

In response to the unbalanced sample categories and complex sample distribution of the operating data of the pitch system of the wind turbine generator system, this paper proposes a method for fault detection of the pitch system of the wind turbine generator system based on the multiclass optimal margin distribution machine. In this method, the power output of the wind turbine generator system is used as the main status parameter, and the operating data history of the wind turbine generator system in the wind power supervisory control and data acquisition (SCADA) system is subject to correlation analysis with the Pearson correlation coefficient, to eliminate the features that have low correlation with the power output status parameter. Secondary analysis is performed to the remaining features, thus reducing the number and complexity of samples. Datasets are divided into the training set for training of the multiclass optimal margin distribution machine fault detection model and test set for testing. Experimental verification was carried out with the operating data of one wind farm in China. Experimental results show that, compared with other support vector machines, the proposed method has higher fault detection accuracy and precision and lower false-negative rate and false-positive rate.

show abstract

“…The requirements for the reliability and safety of modern industrial process are increasing. Therefore, anomaly detection technology for complex industrial processes has become a key research topic in recent years [2][3][4][5]. It can not only ensure the safety and property loss of the production process, but also reduce the fluctuation of product quality and realize the efficient operation of industrial production [6][7].…”

Section: Introductionmentioning

confidence: 99%

Anomaly Detection Method for Multimode Complex Industrial Process Based on Multiple Subspaces Slow Feature Analysis

2021

IEEE Access

View full text Add to dashboard Cite

The multimode operation feature is widely presented in the modern industrial process, which is a typical cyber physical system. In order to achieve the accurate anomaly detection for multimode process, a novel method named multiple subspace slow feature analysis is proposed in this paper. Firstly, the neighborhood subtractive clustering algorithm is used to divide the mode. Then, to consider the local information and conduct fine-scale anomaly detection, Gaussian and non-Gaussian subspaces are built in each mode. Secondly, the static and dynamic features in each Gaussian and non-Gaussian subspaces are extracted through the slow feature analysis, and then the monitoring statistic and control limit are constructed. The control limit in each subspace is estimated according to different methods. During the online phase, the local outlier probability is used to determine the current mode for the online data, and the anomaly detection result can be achieved based on the built anomaly detection model. Finally, the effectiveness of the proposed MSSFA method is demonstrated in a numerical example and a typical industrial case.

show abstract

Data-Driven Two-Dimensional Deep Correlated Representation Learning for Nonlinear Batch Process Monitoring

Cited by 54 publications

References 34 publications

Synthesizing data by transferring information in data‐intensive regions to enhance process monitoring performance in data‐scarce region

Synthesizing data by transferring information in data‐intensive regions to enhance process monitoring performance in data‐scarce region

Fault Detection of Wind Turbine Pitch System Based on Multiclass Optimal Margin Distribution Machine

Anomaly Detection Method for Multimode Complex Industrial Process Based on Multiple Subspaces Slow Feature Analysis

Contact Info

Product

Resources

About