Research On Fault Diagnosis Of Cement Mill Gearbox Based On ICA-SVM Model

Grinding system plays an important role in the cement production process which is named “Milling thrice and burning once”.Vertical roller mill(called as vertical mill in the later)is a highly comprehensive grinding apparatus because of setting crushing process,grinding process,drying process in one.As many advantages exists in the vertical mill such as simple process,low power consumption,less covering and high efficiency,it has been widely used in the modern cement production line.The mill reducer not only transfer power to mill but also support the weight of materials,disc and roller mills.The failure probability of mill reducer is much higher because it always works in the hot,dusty,heavy-duty,high-torque and strong shock environment.In order to ensure the equipment working safely,stably and highly efficient,it's necessary to study the fault diagnosis of mill gearbox.Aiming at the purpose introduced above,taking mill gearbox as the research object,the method of “fault characteristics-fault classification” as the tool,fault diagnosis to the mill gearbox as the goal.The main research work are as follows:(1)The importance of vertical mill in the modern cement production line and significance of gearbox is introduced from the research background.After understanding the present research situation,the development direction of mill reducer's fault diagnosis and it's related fields.All methods are gathered together and compared,so a method that is fit for reducer's fault diagnosis is carrying out combined with the real environment of cement production line.(2)Study the vibration mechanism,model,common failure,reasons and single appearances of vertical mill gearbox's major components such as gears,bearings and shafts.(3)Choosing Independent component analysis as the method of feature extraction.Therefore,the principle of ICA,algorithm,sources number estimation under different situation are discussed.Two innovative methods are proposed.One is near singular value ratio based on Singular Value Decomposition which aims to estimate the number of the source signals.The other one is Singular Value Decomposition of Independent components which can be a tool of feature extraction.Both methods have been tested and verified by experiment.(4)Support Vector Machine is employed because of the reason that it is based on structural risk minimization and has a relatively complete theoretical basis.After the principles of classification are presented.Different failure modes are successfully classified by SVM.(5)Two experiments are designed to verify the feasibility and replicability of “ICA-SVM fault diagnosis model”.If the feasibility and replicability performs well,we can use the model to complete vertical mill gearbox's fault diagnosis.