基于CEEMDAN-多尺度模糊熵和ISRNN的球磨机负荷识别

Load Identification of Ball Mill Based on CEEMDAN-MFE and ISRNN

针对球磨机振动信号具有非线性、非平稳性特点导致的负荷状态难以识别问题,提出一种基于自适应噪声的完备经验模态分解(CEEMDAN),多尺度模糊熵(MFE)和改进堆叠式循环神经网络(ISRNN)的磨机负荷预测方法。首先,采用CEEMDAN算法分解球磨机振动信号以获得本征模态分量;其次,利用MFE提取负荷状态特征;最后,将特征向量作为ISRNN的输入,球磨机负荷状态作为输出,建立球磨机负荷识别模型。试验结果表明,该方法在负荷识别时有较高的精准性,整体识别率高达98.67%,证明了CEEMDAN-MFE特征提取结合ISRNN的方法可实现对球磨机负荷状态的准确识别。

The vibration signal of ball mill has the characteristics of nonlinear and nonstationary,leading to the load state was difficult to identify.In order to solve the problem,a load forecasting method for ball mills was proposed based on CEEMDAN of self-adaptive noise,MFE and ISRNN.Firstly,CEEMDAN algorithm was used to decompose the vibration signal of ball mill to obtain the eigenmode component.Secondly,MFE was used to extract the load state characteristics.Finally,the eigenvector was used as the input of ISRNN,and the load state of ball mill was used as the output to establish the load identification model of ball mill.The experimental results showed that the method had a high accuracy in load identification,and the overall identification rate was up to 98.67%,which proved that the method of CEEMDAN-MFE feature extraction combining with ISRNN can realize the accurate load identification of the ball mill.