The Research On Ball-mill Vibration Signal Feature Extraction And Mill Load Modeling

The ball mill is the key equipment for the development and utilization of ore resources.The production safety and grinding efficiency of the entire grinding industry are directly affected by the load changes in ball mill.The closed and continuously rotating complex operating environment of the ball mill makes it difficult to describe the load of the ball mill in the grinding process.Therefore,carrying out research on mill load detection methods so that the mill load status is accurately identified,which provide accurate and reliable basis for the optimization control and efficiency improvement of the grinding.In this thesis,the industrial wet ball mill as the research object,the load status of the ball mill is accurately identified through the study of vibration signal feature extraction and load modeling methods.The main research contents are as follows:The social significance and economic value of the ball mill load detection research are introduced.The domestic and foreign research status of the ball mill load detection method is described.The common vibration signal feature extraction and methods are summarized.The advantages and disadvantages of existing load detection methods based on ball mill vibration fusion grinding process parameters are analyzed.The load state of the ball mill is divided into three states: underload,normal load,and overload by studying the grinding production process and the working principle of the wet ball mill combined with the actual situation of the industry.The relationship between the barrel vibration signal and the mill load is deeply analyzed.The main factors affecting the load change of the mill are studied and 7 grinding process parametersare used as input parameters for the construction of load detection model.Aiming at the problem that the vibration signal load feature of the ball mill barrel is difficult to accurately extract,a feature extraction method based on adaptive VMD and improved power spectrum estimation is proposed in this thesis.The number of modes of VMD is adaptively determined according to the principle of cylinder vibration signal generation and using the adaptive adaptation of EMD combined with the sensitivity of the kurtosis parameter to the impact amount.The autocorrelation function decreases the amplitude of the waveform as the time delay increases,which is effectively avoided by the data extension method.The intrinsic mode function of vibration signal is processed by introducing the Nuttall self-convolution window and the energy centrobaric method.By introducing Nuttall self-convolution window combined with energy centroid method,the frequency corresponding to the maximum power spectrum of the modal component of the vibration signal is extracted as the load feature.The industrial measured results show that the proposed method is more accurate than the EMD method in mill load identification.The complex operating environment of the ball mill makes the load detection of the ball mill based on a single signal unreliable.In this thesis,in order to achieve accurate and reliable determination of the load status of the ball mill,the GA-BP neural network ball mill load status identification model is established and multi-source characteristics as input parameters of load identification model are constructed by combining 7 grinding process parameters and vibration characteristics.The industrial measured results show that compared with BP neural network,GA-BP neural network has faster iteration speed and high load identification rate and multi-source feature model has higher load identification rate than single vibration feature model.Finally,based on the algorithm proposed in this thesis,the ball mill load identification system based on Lab VIEW platform is built.The software architecture of the whole system is introduced,and the software design of the vibration signal data collection,feature extraction,load identification,data storage and other functions is realized.By drawing on the industrial measured data,the test experiment of the load identification system is completed,and the effectiveness and accuracy of the method is verified.