基于频谱分析的磨机负荷检测方法研究

发布时间：2018-08-19 16:07

【摘要】：磨机是工业生产中物料粉碎的核心设备。球磨机通过钢球和研磨物之间的频繁碰撞实现物料的粉碎作业。为保证球磨机高效、安全地运作,必须对球磨机内部工作状态进行检测。球磨机负荷(主要包括物料量和料位)是磨料过程的重要检测指标,直接影响磨机的工作效率。然而球磨机内部的工作环境复杂多变,难以保证稳定的运作状态,对球磨机负荷检测带来极大阻碍。现阶段工厂主要通过人工监听的方式对负荷进行估计,存在较大误差。采用传统的检测方法也存在不能准确检测出磨机负荷状态的问题。本文首先对传统主元分析(PCA)结合极限学习机(ELM)的检测方法进行深入研究,分别从频谱分析、主元提取和模型建立三个方面对该方法在磨机负荷检测中存在的缺陷进行分析,并提出相应的解决方案。然后,结合磨机负荷检测的特点,本文采用基于频谱分析的方式,提出将核主元分析(KPCA)和误差最小化极限学习机(EM_ELM)相结合的建模检测方法,对磨机工作时产生的磨音信号进行建模分析。本文提出的方法结合小波包去噪对信号进行前期预处理,利用最大熵法的现代功率谱估计方法将信号转换到频域分析,采用间接检测的方式建立磨机内部负荷状态和外部检测信号之间的模型关系。最后,结合工业现场采集球磨煤机工作时的数据进行实验测试,检测结果与传统的PCA-ELM检测方法进行测试对比。结果表明,采用本文提出的基于频谱分析的KPCA-EM_ELM检测方法,在测量准确度上得到了提高,保证算法的运行时间,提高了检测准确度和效率。为将该检测方法运用到实际的球磨机负荷检测系统中提供了理论依据,对于提高球磨机研磨效率、稳定生产具有重要的意义和广阔的应用前景。
[Abstract]:Mill is the core equipment for material crushing in industrial production. The ball mill comminutes materials through frequent collisions between steel balls and abrasives. In order to ensure the efficient and safe operation of ball mill, it is necessary to check the internal working state of ball mill. Ball mill load (including material quantity and material level) is an important testing index of abrasive process, which directly affects the working efficiency of mill. However, the internal working environment of ball mill is complex and changeable, it is difficult to ensure stable operation state, which greatly hinders the load detection of ball mill. At present, the factory mainly uses manual monitoring to estimate the load, there is a big error. There is also the problem that the load state of the mill can not be detected accurately by using the traditional detection method. In this paper, the detection methods of traditional principal component analysis (PCA) combined with extreme learning machine (ELM) are studied in detail. The defects of this method in mill load detection are analyzed from three aspects: spectrum analysis, principal component extraction and modeling. And put forward the corresponding solution. Then, considering the characteristics of mill load detection, this paper proposes a modeling and detection method which combines kernel principal component analysis (KPCA) with error minimization learning machine (EM_ELM) based on spectrum analysis. Modeling and analysis of grinding sound signal produced by grinding machine. The method proposed in this paper combines wavelet packet denoising to pre-process the signal and converts the signal to frequency domain analysis by using the modern power spectrum estimation method of maximum entropy method. The model relationship between the internal load state of mill and the external detection signal is established by indirect detection. Finally, the test results are compared with the traditional PCA-ELM detection method, combined with the data collected from the industrial field during the operation of the ball mill. The results show that the proposed KPCA-EM_ELM detection method based on spectrum analysis can improve the measurement accuracy, ensure the running time of the algorithm, and improve the detection accuracy and efficiency. It provides a theoretical basis for the application of this method to the actual load detection system of ball mill. It is of great significance and broad application prospect for improving the grinding efficiency of ball mill and stabilizing production.
【学位授予单位】：重庆邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH69

【参考文献】