基于决策树算法的AMT挂挡过程冗余控制研究

为了对AMT系统挂挡过程进行良好的冗余控制，针对挂挡位移传感器失效状况，提出了一种基于CART决策树算法的挂挡时间预测模型.模型建立过程中，选取挂挡阶段的气压p、变速器输入轴转速n₁、变速器输出轴转速n₂和同步器的同步速差Δ；n作为特征值变量，选取挂挡时间作为预测值.通过交叉验证，对原始决策树进行修剪，得到最优决策树模型.在使用测试集进行模型验证时，在预测误差不超过50 ms的情况下，对挂挡时间的预测具有90%以上的准确率.台架及实车试验表明，在挂挡位移传感器失效的情况下，基于该算法的电磁阀控制策略可以完成正常的换挡操作，通过实际传动比和控制策略预测得到的理论传动比的对比发现，预测的换挡点时刻的判断误差在50 ms以内. 

Research on Redundant Control of AMT System Gear Shifting Process Based on Decision Tree Algorithm

In order to improve redundant control of the AMT system gear shifting process, considering the failure condition of the gear shifting displacement sensor, a redundant control strategy was proposed based on decision tree algorithm. The air pressure p, the transmission input shaft speed n₁, the transmission output shaft speed n₂ and the synchronous speed difference of the synchronizer Δn were selected as the characteristic variables, and the gear shifting time was selected as the predicted value to establish a gear shifting time decision tree prediction model. To obtain an optimal decision tree model, a cross-validation and the original decision tree pruning were arranged. The results show that the prediction accuracy rate of the gear shifting time can reach up to 90% with the error less than 50 ms. The bench test and actual vehicle verification results show that, when the gear shifting sensor fails, the solenoid valve control strategy based on this algorithm can ensure the normal shifting operation. Comparing the actual transmission ratio with the theoretical transmission ratio predicted by the control strategy, the operation results show that the judgment error of the predicted shifting time is all within 50 ms.