基于粒子群核极限学习机的涡扇发动机加速过程模型辨识

Turbofan Engine Model Identification of Acceleration Process Based on Particle Swarm Optimization Kernel Extreme Learning Machine

关键词:  涡扇发动机  加速过程  核极限学习机  数据驱动  模型辨识

Key words:  Turbofan engine  Acceleration process  Kernel extreme learning machine  Data-driven  Model identification

摘要:

针对解析法建立涡扇发动机加速过程模型精度和实时性不高的问题，提出了一种基于粒子群核极值学习机（PSO-KELM）的涡扇发动机加速过程模型数据驱动辨识方法，构建涡扇发动机加速过程模型，结合加速过程试车数据，利用PSO-KELM方法对该加速模型进行辨识。试验结果表明：低压转子转速、高压转子转速和低压涡轮出口燃气总温都较好地逼近了试车数据，最大相对误差均值分别为1.013%，0.355%和1.055%，平均计算时间为0.04ms。精度和实时性均优于反向传播神经网络和粒子群支持向量回归方法，可用于发动机状态监控和性能优化控制。

Abstract:

In order to solve the problem of low accuracy and real-time in establishing the model of turbofan engine acceleration process by analytic method， a data-driven method on identifying the turbofan engine acceleration process model based on kernel extreme learning machine （KELM） optimized by particle swarm optimization （PSO） was proposed. Firstly， a turbofan engine acceleration process model was constructed. Then， PSO-KELM was adopted to identify the model using engine acceleration process test data. The results show that identification results of the low-pressure rotor speed， the high-pressure rotor speed and the low-pressure turbine outlet gas temperature are all close to measured data， the mean maximum relative errors are 1.013%， 0.355% and 1.055%， respectively， and the mean computing time is 0.04ms. The precision and real-time are better than back propagation （BP） neural network method and PSO-support vector regression （SVR）. This method can be used for engine condition monitoring and performance optimization control.