气动加载系统具有严重非线性和参数不确定性，很难用精确的数学模型来描述，而在系统运行过程中，存在大量的输入输出数据，因此，提出一种基于数据驱动的气动加载系统在线建模方法.离线辨识阶段，基于Takagi-Sugeno模糊模型，采用减法聚类和模糊c -均值相结合的模糊聚类算法从输入输出样本数据中提取模糊规则，然后利用递推最小二乘法对后件参数进行辨识，得到离线模型；在线调参阶段，采用最小均方差算法在线修正模型线性参数，得到在线模型.基于VC++6.0软件开发平台，设计了在线建模程序，在气动变载荷摩擦磨损试验机上进行了实验研究.实验结果表明：该模型具有较高的辨识精度、较理想的泛化能力和跟踪能力，为建立气动加载系统的数学模型提供了一条新途径，同时也为气动加载系统的智能控制奠定了基础.%Pneumatic loading system has the characteristics of serious nonlinearity and parameter uncertainty,it is difficult to de-scribe the system by accurate mathematical model. There are a large number of input/ output data while the system is running. So an online modeling method based on data-driven was proposed for the pneumatic loading system. On the offline stage,subtractive cluste-ring and fuzzy c - means clustering (FCM)were combined to extract fuzzy rules from the input and output data,and then recursive least squares method was used to identify the consequent parameters. On the online stage,least mean square algorithm was adopted to adjust the parameters of the model. Based on VC++ 6. 0 software development platform,online fuzzy modeling program was designed and tested on a pneumatic variable-loading friction and wear tester. The results show that the model has high identification accuracy, satisfactory generalization ability and tracking ability. It provides a new way to establish the mathematical model for pneumatic loading system. Furthermore,it lays foundation for the design of intelligent control algorithms for pneumatic loading system.
展开▼
