【摘要】 本文通过对某船用柴油机的匹配涡轮增压器进行研究，得到了压气机的特性曲线。分析了压气机转速、质量流量、压比和效率之间的关系。为了拟合回归和预测压气机特性曲线，使用了一些广泛使用的代理模型，包括Kriging模型、响应面法（RSM）、人工神经网络（ANN）、径向基函数（RBF）和支持向量机（SVM）。代理模型用于精确预测涡轮增压器在特定时间的运行参数。采用代理模型对不同速度组的曲线进行拟合，研究了压力比与效率的整体预测建模方法，并通过数据分析了不同训练样本量对拟合精度的影响;通过遗传算法优化后的全转速下的代理模型的拟合精度提高，在样本数量较少时，其代理模型拟合结果的R~2超过95%。更多还原

【Abstract】 This study gets the matched turbocharger of a marine diesel engine to acquire the compressor characteristic curve. It was done to analyze the relationships between compressor speed, mass flow rate, pressure ratio, and efficiency. For fitting regression and predicting the compressor characteristic curve, a number of widely used surrogate models, including the Kriging model, Response Surface Methodology(RSM), Artificial Neural Networks(ANNs), Radial Basis Function(RBF) and Support vector machines(SVM), were used. The surrogate model is used to precisely anticipate the turbocharger’s operational parameters at a specific time. The surrogate models are used to fit the curves of different speed groups, the result shows that the whole prediction modeling method of pressure ratio and efficiency is studied, and the influence of different training sample size on fitting accuracy is analyzed by data. After optimization by genetic algorithm, the fitting accuracy of the surrogate model at full speed is improved. When the samples size is small, the fitting result R~2 of the surrogate model is more than 95%.更多还原