A Novel Cyber-Physical System For Combustion Instability Detection Using Deep Neural Network

For effective monitoring of disparate cyber-physical systems involving sequential data,precise frameworks are required to predict system states.Interpretation provides insights that can improve scientific understanding of the system if the framework is interpretive.It is critical to detect impending instability shifts in order to initiate effective combustion system control.In this case,it’s critical to create solid frameworks.We train our proposed deep convolutional neural network(CNN)model on continuous image frames extracted from high-speed flame videos by inducing instabilities in the system under specific circumstances,as one of the first applications of deep neural networks to characterize instabilities in combustion systems.Changing the length of the protocol We apply inexpensive but noisy labeling techniques to sound pressure data to define a dimensionless instability metric for training.We are attempting to detect the onset of instabilities in transient datasets where instabilities are caused by different protocols.We can successfully detect critical transitions of high combustion instability states by changing control parameters,demonstrating the robustness of our proposed detection framework,which is not dependent on combustion-induced protocols.We propose a new model that takes into account temporal correlation.Without affecting the accuracy,this model can account for the contribution of each image in the input sequence to the generation of individual predicted labels.We use a temporal attention mechanism to capture the global temporal structure after encoding the images of the input sequence with a 2D convolutional neural network and a long short-term memory recurrent neural network.The importance of image frames that are most relevant to each prediction is highlighted by attention weights.We show how well our model performs in a problem where interpretability and robustness have never been investigated before.This can lead to a better grasp of the situation and more effective control.