Convolutional neural network based combustion mode classification for condition monitoring in the supersonic combustor

Abstract

Abstract Supersonic combustor is one of the core components in the scramjet, so it is of great significance to monitor the combustion modes in the combustor to ensure the safe and stable operation of the scramjet engine. Traditionally, several key parameters or manually-engineered features are selected as the indicators to evaluate the operation conditions, which usually heavily depends on the professional experience and carries considerable limitations. Convolutional neural networks have been proved to be effective in automatic feature extraction and have shown better generalization performance. Hence, it is attractive and promising to apply Convolutional neural networks to condition monitoring in mechanical systems due to their excellent ability of pattern recognition. To accomplish the classification of combustion modes, a convolutional neural network based method is proposed, which can learn features directly from the raw pressure data collected during supersonic combustion experiments. Meanwhile, the proposed method is compared with the traditional machine learning methods, such as multilayer perceptron, k-nearest neighbor, single-hidden layer feedforward neural network, and support vector machine. Furthermore, feature data is constructed by manual statistical features from time domain and frequency domain. The raw data and feature data are both considered to study the influence of feature extraction methods on the performance of different models. The results show that the proposed convolutional neural network based method is able to reveal intrinsic features from raw data and effectively complete the classification of four main combustion modes occurring in the combustor. The novel approach achieves a higher classification accuracy and better generalization performance than other comparative methods.