Interpretable Deep Learning for Monitoring Combustion Instability

摘要: Abstract Transitions from stable to unstable states occurring in dynamical systems can be sudden leading catastrophic failure and huge revenue loss. For detecting these transitions during operation, it is of utmost importance develop an accurate data-driven framework that robust enough classify scenarios. In this paper, we propose deep learning frameworks show remarkable accuracy the classification task combustion instability on carefully designed diverse training test sets. We train our model with data a laboratory-scale system showing states. The dataset multimodal correlated hi-speed video acoustic signals. labeling mechanism for sequences by implementing Kullback-Leibler Divergence time-series data. using 3D Convolutional Neural Network Long Short Term Memory network task. To go beyond gain insights into predictions, incorporate attention across time-steps. This aids understanding time-periods which contribute significantly prediction outcome. validate domain knowledge perspective. By exploring inside black-box models, used development better detection different systems.