基于集成ELM模型的航空发动机性能参数预测

Aircraft engine performance parameter prediction based on ensemble ELM model

为预测航空发动机性能参数,提出了一种动态集成极端学习机模型。采用AdaBoost.RT集成算法对极端学习机(ELM)进行集成,并针对AdaBoost.RT集成算法中固定阈值的局限性,采用自适应动态调整阈值的方法来提高预测精度。通过比较前后两次迭代的均方根误差(RMSE)的大小,对阈值进行调整。最后,以排气温度裕度为预测参数,使用动态集成ELM模型对其进行预测,并与单一ELM模型和原始集成ELM模型进行比较。结果表明:改进模型的预测结果好于其他模型,更适合航空发动机性能参数预测。

In order to predict performance parameters of aircraft engine accurately, a dynamic ensemble extreme learning machine （ELM） model is proposed. Ada Boost.RT algorithm is used to integrate ELM to construct the ensemble model. Aiming at the limitation of static threshold in original AdaBoost.RT algorithm, a self-adaptive and dynamic adjusting method is used to improve the forecasting precision. A given method makes adjustments to the threshold by comparing RMSE of two neighboring iterations. Finally, compared with single ELM model and original ensem- ble ELM model, dynamic ensemble ELM model is used to predict EGTM. Results show that the improved model is better than other models for the performance parameters prediction of aircraft engine.