Physics-based and data-driven prediction method for features and type boundaries of oblique detonation wave systems in hydrogen-air mixtures

Predicting the wave systems in the initiation region of oblique detonation wave (ODW) is critical to the development of oblique detonation engine. Despite the extensive studies on ODW initiation characteristics, this remains a challenging task. In this study, a physics-based and data-driven prediction method is proposed to predict the features and type boundaries of oblique detonation wave systems based on Gaussian process regression (GPR) surrogate model and support vector classifier (SVC). The ODWs of hydrogen-air mixtures under wide realistic flight conditions are numerical simulated by solving the two-dimensional reactive Euler equations. Four different wave systems are analyzed and quantified based on actual ODW physics to extract the main features as the training data, reducing the size of required data and improving data interpretability. A GPR surrogate model is constructed with 36 initial samples and further refined with 36 infill samples using mean square error (MSE) infilling criterion. The refined surrogate model is validated based on 10 test samples, showing good prediction performance on the wave system features in the feasible region. Moreover, SVC is utilized for finding the boundaries of different wave system types based on the criterion of ODW characteristic height ratio and predicted data from GPR surrogate model. The prediction type boundaries are obtained as specific formulas with a total accuracy of 94.55% and an average F1-score of 0.9410. This work addresses the challenge for predicting features and type boundaries of ODW systems, showing great potential in engineering applications.