Research On Intelligent Fault Identification Method For Transmission System Of Passenger Ropeway Under Severe Conditions

Passenger ropeway,as a transportation that can adapt to various complex terrain,needs to be in high load operation for a long time.According to statistics,nearly half of the passenger ropeways need to keep operating in the cold environment in winter.Long-term exposure in low temperature,strong wind,ice and snow,strong magnetic interference,variable speed,variable load and other complex conditions greatly aggravate the risk of passenger ropeway failure.According to the characteristics of passenger ropeway operation,fault identification methods based on signal analysis theory are often complicated and require analysis experience,which can’t meet the requirements of passenger ropeway operation safety guarantee.Therefore,this thesis focuses on the fault monitoring and detection of ropeway under complex operating conditions,which researches on the construction of health monitoring indicators and adaptive fault warning,intelligent fault identification under small samples,and intelligent quantitative identification of ropeway faults.They aim to provide effective decision-making basis for the maintenance and management of the ropeway system,and achieve the practical engineering application results.Aiming at the problem that the existing monitoring indicators and models are difficult to describe the complex nonlinear degradation process of ropeway equipment,an adaptive fault warning method based on fusion indicators and recurrent neural network is proposed.This method constructs optimal sensitive monitoring indicators by extracting relevant features from time domain,frequency domain and time-frequency domain.Furthermore,a heterogeneous Bi-directional GRU(Gated Recurrent Unit)model is built based on the fusion index,which only utilizes normal data for training to realize the adaptive fault warning of the ropeway.Multiple bearing acceleration degradation experiments and measured data of passenger ropeway in Winter Olympics are used to verify it effectiveness.The results indicate that the prediction accuracy of the proposed method is significantly higher than conventional models,especially in the early stage of failure development,which can provide strong support for early failure warning.Aiming at the problem of insufficient fault samples and incomplete failure modes of the ropeway,a metric-based meta-learning network with small samples is proposed.This method pre-trains the network by sufficient samples with various categories,which can make the network adaptively learn a pseudo-distance measurement.In specific application process,a small amount of labelled samples and pre-trained optimal weights are used to directly identify unknown data.Rotor bearing fault data and gearbox reducer fault data are utilized for verification,and the results show that the proposed method has high generalization.The optimal pre-trained weight can be applied to the identification of unknown data with multiple fault categories at the same time and obtains good recognition results.Aiming at the problem of fault damage-degree quantitative identification with high similarity between different damage-degree data of the same fault type,a fault quantitative identification network based on multi-scale parallel convolution structure is proposed.This method is based on the principle of inner product matching,multi-scale kernels are constructed in the same layer,so that the effective information of entire samples can be completely extracted.Finally,the information extracted by multiple-scale kernels is integrated to obtain the most discriminative features,which is utilized for quantitative fault diagnosis.Motor bearing fault experiments and measured data of passenger ropeway in Mount Hua validate its feasibility,and the results show that the proposed method can effectively realize the quantitative identification of ropeway transmission system operation faults.