Multivariate time series (MTS) forecasting plays an important role in industrial process monitoring, control and optimizations. Usually, hierarchical interactive behaviors among industrial MTS have formed complex nonlinear causal characteristics, which greatly hinders the applications of existing predictive models. It is found that graph attention networks (GAT) provide technical ideas to meet this challenge. However, the unknown directed graph and linear conversions of node information make conventional GAT less popular for the industrial fields. In this paper, we propose a novel prediction model termed as temporal causal graph attention networks with nonlinear paradigms (TC-GATN) to adequately capture inherent dependencies on industrial MTS. Specifically, the graph learning algorithm concerning the granger causality (GC) is exploited to extract potential relationships among multiple variables for guiding directional edge connections of the hierarchy. Then, parallel GRU encoders located in the graph neighborhood space are introduced to perform the nonlinear interaction of node features, which accomplishes the adaptive transformation and transmission. The self-attention mechanism is further employed to aggregate encoder hidden states across all stages. Finally, a temporal module is supplemented to process information from the graph layer, achieving satisfactory predictions. The feasibility and effectiveness of the TC-GATN are validated by two actual datasets from the methanol production and the chlorosilane distillation.
1 August 2018 in Chemometrics and Intelligent Laboratory Systems
The Internet has given learners the ability to learn anytime and anywhere at their own pace facilitated by interactive and adaptive software. Digital learning technology is more than just providing students with a laptop. Digital learning requires a combination of technology, digital content, and instruction. This study aims to identify and observe the impact and mediation of top management support in relation to the successful adoption of digital learning technologies in developing countries. A questionnaire was designed and distributed to rate the successful adoption and utilization of digital learning technologies in developing countries and data was analyzed using structural equation modeling. This study provides empirical evidence and explains many complex factors, such as user awareness, perceived usefulness, perceived ease of use, and information communication technology infrastructure, in the context of top management support to facilitate the effective utilization of a digital learning technology. Mediating the top management support between adoption of a digital learning technology and user awareness, perceived usefulness, and perceived ease of use provides clear and crucial evidence to support the effective adoption of a digital learning technology.
Imbalanced data samples can adversely affect the performance of industrial process fault diagnosis models. Recently, it has become a valued challenge to expand data samples and reasonably assess their quality. To address this issue, this paper presents an imbalanced data augmentation and assessment method that integrates Wasserstein Time Generative Adversarial Network with Gradient Penalty (WTGAN-GP) and Maximum Information Coefficient with Improved Dynamic Time Warping Distance (MIC-IDTW) indicator. Firstly, the WTGAN-GP effectively tackles the scarcity of fault data by incorporating the Wasserstein distance with gradient penalty into TimeGAN, significantly enhancing the data generation capability and stability of the network. Additionally, the MIC-IDTW is established as a quantitative and interpretable indicator for assessing the quality of generated samples. Finally, this paper validates the performance of WTGAN-GP and MIC-IDTW in addressing the issue of imbalanced data in vibration fault diagnosis for an actual factory centrifugal air compressor. It is demonstrated that the proposed methods can effectively enhance various fault data in the presence of imbalanced fault samples and significantly improve the fault classification performance.