Fault Diagnosis Method Of Cantilever Roadheader Based On Compound Network Topology

China's coal mining is the world's largest country,a long time,the sustainability of coal mining equipment and personnel work has been restricting the safety of coal mining industry development,to the country and the people's lives and property has brought a very important impact.In addition,intelligent coal mining technology innovation has gradually become the leading new coal mining methods,but if the roadway as the main equipment in the boring machine in intelligent development lagging behind,will inevitably lead to low efficiency,resulting in economic loss,impact Normal mining efficiency and economic and social benefits of coal mines.Therefore,it is an important task to ensure the safe production of coal mines in the process of tunneling,which is to ensure the fast,stable and effective excavation work.At present,the methods of condition monitoring and fault diagnosis of coal mine roadheader are mainly on-line threshold judgment,and lack of intelligent machine fault diagnosis theory and method.Online threshold is a traditional method,with the motor current limit alarm and leakage protection,etc.to determine the fault situation.Hydraulic component fault diagnosis is also determined by local data analysis afterwards.The shortcomings of this method is: local parts of the fault is often found to affect the whole machine work caused by shutdown shutdown conditions,are after maintenance,the threshold to determine the conditions are too single,fault chain propagation is long,difficult to determine and locate the specific point of failure.Fault diagnosis of complete machine system is widely used in fault diagnosis of complex system.Through data analysis of each observation point of the system,the fault propagation path of roadheader can be judged,and intelligent reasoning can predict the development trend of fault.Compared with the on-line threshold judgment,the whole system fault diagnosis is suitable for the operation evaluation of complex system,which is different from the similar component diagnosis and evaluation method,which has the advantages of economy,high efficiency,large prediction range and flexible reasoning.At present,the machine condition monitoring and fault diagnosis technology research status mainly include:?1?The development platform of state monitoring and fault diagnosis system of roadheader based on embedded hardware is a platform for data display and information threshold judgment.?2?Using fuzzy neural network to build the expert platform of fault diagnosis information of roadheader,the theoretical analysis lacks more effective means;?3?The specific fault diagnosis of hydraulic,electrical,mechanical and other parts of the boring machine is focused on the details,but lacks the fault diagnosis reasoning of the whole system.After summarizing the current situation of fault diagnosis of many boring machines,this paper presents the fault diagnosis technology of roadheader based on PSO-BP,fault tree,faulted Petri net and other complex network topologies.A fault diagnosis model of roadheader based on multiple network structures was established.A model based on PSO optimization algorithm was established to analyze the mechanism and path of roadheader fault propagation.By means of field test data The performance of the theoretical model and the analytical results are verified.The main research results and conclusions are as follows:?1?Fault diagnosis model of roadheader based on BP neural network.In this chapter,the eigenvalues of the fault of the TBM are determined by analyzing the vibration signals,and the eigenvalues are determined as the input vectors of the neural network.The BP neural network was established to test the performance of the network.According to the different working status of the TBM,the BP neural network was used to analyze the data and to accurately identify the cutting parameters of the neural network.BP neural network was built and the performance of the network was tested.Finally,the BP network was compared with the PSO-BP network by the PSO-BP neural network.The PSO-BP neural network was tested and the performance of the PSO-BP neural network was tested.Comparison shows the optimization effect.?2?Fault diagnosis model of roadheader based on dynamic fault treeBased on the analysis of the structure of TBM,a typical fault tree model of TBM is established,and the minimum cut set of the fault tree is determined by qualitative analysis of the fault tree,which leads to the failure of roadheader Of possible failure sources.The relative probability of the fault tree bottom of the TBM is obtained by analytic hierarchy process?AHP?.The fault tree of the roadheader is quantitatively analyzed,and the importance of each fault source in the TBM system is obtained qualitatively.Based on this result,the optimal route of the TBM fault is given,which provides the basis for the fast search of the fault source and the weak link of the system design.?3?Fault diagnosis model of roadheader based on faulted Petri netBased on the application of the traditional Petri nets in the field of fault diagnosis,this paper gives the different modeling classification and a theoretical expression of the fault diagnosis problem,and abstracts the three major scientific and technical problems in the field of fault diagnosis.On this basis,The definition and the realization of the diagnosis method of the TBM based on the faulty Petri net are discussed.Finally,the expression of the intelligent diagnosis function and the path of the fault propagation are verified and pointed out.At the same time,Intelligent reasoning consistent expression.?4?Research on Fault Diagnosis Optimization of Roadheader Based on Multi-networkTopologyWhen the target precision is 10^-3,the average number of training steps of BP neural network is 100.2.The average number of training steps of PSO-BP neural network is 4 steps.The average training step of PSO-BP neural network is 4 steps.;When the target precision is 10^-5,the average training step of BP neural network is 198.5 steps,and the average number of training steps is 7 steps in PSO-BP network.It can be seen that the PSO-BP network training can reach the target precision more quickly than the BP neural network,and the training steps of the PSO-BP network are much less than the BP neural network.Fault Tree-Petri Multiple Compound Network Topology Fault Diagnosis Model Research,Perfecting Mathematical Theory Method of Fault Diagnosis of Roadheader.