Research On Data Mining Technology For The Bulk Port Handling Machinery Fault Diagnosis

With the continuous development of modern industrial technology,the industrial processes become increasingly complex.The devices continuous operation under high load,high power conditions,the failure will inevitably occur.The traditional fault diagnosis methods encountered a bottleneck in solving the current increasingly complex industrial mechanical failures.It is difficult to accurately locate the fault in a short time.Taking advantage of the unique advantage of data mining techniques in terms of knowledge discovery provides a new impetus to the development of fault diagnosis technology.In this thesis,the mechanical failure of port belt conveyor system as the research object,data mining techniques for bulk port handling machinery fault diagnosis are studied.The essence of fault diagnosis is the pattern recognition of faults.Fault pattern recognition using data mining techniques is to classify fault samples.There are many classification algorithms in data mining.Among them,decision tree method attracts widespread attention for the advantages of its high model training efficiency and high classification accuracy.It has been successfully used in pattern recognition and fault diagnosis and other fields.The thesis study the questions of using decision tree method to extract fault rules and to classify new fault samples.Aiming at the problems above,this thesis mainly has conducted the following work:(1)This thesis proposes an improved algorithm based on variable precision rough set model(VPWMR algorithm).For the problem of decision tree model for fault diagnosis based on the traditional decision tree method with extracted fault rules not simple and effective enough,low classification accuracy,poor noise immunity and other issues,this thesis improves the decision tree generation method based on rough set model(WMR algorithm),and proposes an improved algorithm based on variable precision rough set model.Improved decision tree construction method introduces an error parameter ?(<? ?5.00),allowing the existence of a certain classification error when dividing instances.Thus some small-scale instances with noise and inconsistency may be divided into large-scale instances of categories,thereby weakening the adverse effects on the decision tree classification process caused by a small number of instances.A case analysis and comparison experiments illustrate that the decision trees constructed by VPWMR algorithm are better than WMR algorithm in terms of classification accuracy,size and noise immunity,By comparing the results of different ? values under certain conditions,it indicates the effect of different ? values on classification accuracy of the data set.(2)Combining strong capabilities to deal with uncertainty and incomplete information of rough set theory and rapid classification advantage of VPWMR decision tree algorithm,this thesis proposes a rough set and decision tree intelligent fault diagnosis model(RSDT diagnosis model).A case analysis which respectively uses RSDT model and C4.5 model to build decision tree models and makes a comparative analysis,shows the feasibility and effectiveness of RSDT model.(3)The belt drive means of a bulk terminal as the main study object,a fault diagnosis system is devised.RSDT model is applied to the system,achieving the functions of the mechanical fault rules mining and fault diagnosis,which also shows the feasibility of the diagnostic method in actual systems.