Image-based process monitoring using deep learning framework

With the advances in optical sensing and image capture systems, process images offer new perspectives to process monitoring. Compared to the process data collected by traditional sensors at local regions, process images enhance data-driven process monitoring a lot by capturing more significant variations in the whole space. Using the easily available industrial process images, a new deep learning framework based on the deep belief network (DBN) is proposed for feature extraction and timely fault detection. Unlike the traditional DBN methods inputting the images into the network directly, in the proposed framework, the sub-networks are used to extract local features from the sub-images. The global network fuses all of them for global feature extraction to remarkably improve the training efficiency without deteriorating the fault detection accuracy. Meanwhile, a new statistic is specially developed for the proposed deep learning framework. Finally a real combustion system is introduced to demonstrate the effectiveness of the proposed method.