(150a) Visual Analytics: A New Paradigm for Process Monitoring

Inspired by the recent breakthroughs in computer vision, we propose a novel paradigm for process monitoring. Visual analytics is the process of combining visual representation of the data, computer vision tools, and analytical reasoning to support decision-making and extract knowledge from the data. Using the process data as visual clues, we convert the process monitoring problem into an image recognition problem. In this talk, a novel end-to-end visual analytics pipeline for industrial process monitoring, using 1D and 2D convolution operations, is proposed. We learn a visual representation of the data that captures the temporal and local features from historical 1D time-series signals that would otherwise be spread over time. Next, the learned features in a 2D format are visually recognized and classified using 2D convolutional neural networks. Our experimental results demonstrate that this approach achieves better performance on an industrial multi-variate dataset compared to other state-of-art signals imaging tools (e.g., Gramian Angular Field and Recurrence Plot).