船用中速双燃料发动机放热规律神经网络预测模型的开发

Development of Neural Network Prediction Model for Heat Release Law of Marine Medium Speed Dual Fuel Engine

以船用中速双燃料发动机为研究对象,提出其放热规律神经网络预测模型的开发方法。首先建立船用中速双燃料发动机的多维性能仿真模型,对增压空气压力、燃气喷射量和引燃油喷射提前角等不同控制参数进行数值组合,计算多组不同工况条件下的放热率曲线;通过对多条放热率曲线进行参数化分析,明确描述放热率曲线的4个曲线特征参数和特征方程;建立双燃料发动机放热规律神经网络预测模型,以控制参数作为输入量,以放热率曲线特征参数作为输出量,利用多组放热率数据对神经网络模型进行训练和测试。该模型揭示了控制参数与放热率之间的规律,可由控制参数对放热率曲线进行预测。仿真计算结果表明:相比一般的发动机实时仿真模型,神经网络预测模型结果更加贴近发动机实际工作状态。

The marine medium speed dual fuel engine taking as the research object, a neural network prediction model of the engine is developed. A multi-dimensional performance simulation model of dual-fuel engine is established and the ROHR curves are calculated under various numerical control parameters such as charge air pressure, gas injection amount and pilot injection angle. The multiple ROHR curves are parametrically analyzed to determine the four characteristic parameters of the curve, y0, A, xc and w, as well as the curve characteristic equation, Gauss Amp equation. A neural network prediction model of the heat release law of dual-fuel engine was established. The neural network model was trained and tested by using the control parameters as inputs and four curve characteristic parameters as the output variables. The model correlates the control parameters with the heat release law, and predicts the rate of heat release（ROHR） curve according to the control parameters. The simulation results show that compared with the general real-time simulation model of the engine, the neural network prediction model results are closer to the actual working state of the engine.