Applying machine learning techniques to predict detonation initiation fromhot spots  

作　　者：Je Ir Ryu 

机构地区：[1]Department of Mechanical Engineering,University of California-Berkeley,Berkeley,CA 94720,USA

出　　处：《Energy and AI》2022年第3期121-127,共7页能源与人工智能（英文）

摘　　要：As hot spots can be a source of detonation initiation by autoignition in a reactive mixture, understanding hotspot initiated detonation is significantly important for energy-related applications of detonation. Although theZel’dovich reactivity gradient theory is extremely useful to predict reaction propagation modes from hot spots,the prediction is limited due to the changes in the unburnt mixture during the induction time. In the currentstudy, machine learning based estimation methods by training the numerical simulation result data set aresuggested to avoid the extensive and empirical effort to find initial conditions of hot spot initiated detonationin inhomogeneous mixtures. The data set was obtained from the detailed numerical simulations with variousconditions of hot spots and divided into training and test data sets. Some variables were normalized withothers to avoid multicollinearity. Three different machine learning techniques, logistic regression, classificationand regression trees, and artificial neural network, were utilized to develop prediction models. Using thedeveloped models by machine learning approaches, the modes of hot spot initiated reaction front propagationcan be predicted without computationally expensive numerical simulations. The accuracies of the machinelearning based prediction models were significantly improved compared to the baseline model simply usingthe Zel’dovich reactivity gradient theory. These models can be utilized as preliminary prediction methods ofhot spot induced detonation initiation conditions for further detailed numerical simulations and dominantinput variables.

关 键 词：DETONATION DEFLAGRATION Hydrogen SYNGAS Logistic regression CART Artificial neural network 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]