The Research On The Fault Diagnosis Method For The Rotating System Of Postal Equipment

In recent years,China's logistics business has been developed by leaps and bounds. With the rapid development of the technology of internet of things and information processing, and the requirement of labor cost and delivery time for logistics business, all kinds of rotary system of postal equipment are widely used in logistics processing center. As the core component of the rotation system of postal equipment,the induction station can malfunction easily due to the long-term high speed and full load operation and it can bring great economic losses. So the the fault diagnosis is very important for the induction station.The post rotating equipment is affected by the complex vibration mode of the mechanical mechanism, the poor environment noise and uncertainty of fault criteria, so it is very difficult to accurately identify whether it is abnormal. However the existing detection method based on the electrical signal and expert information system is insufficient. This paper systematically researches the fault diagnosis method of the rotation system of postal equipment from the perspective of the coupled rigid-flexible multi-body dynamics, feature extraction and information fusion.Firstly, the mechanical fault definition and classification method for the mechanical failure of the induction station is proposed Four grades and criteria of mechanical failure are defined based on the working principle, structural characteristics and failure state of the induction station. Then, the failure modes, the fault contents and the failure mechanism of the three main rotating parts are studied. Finally, based on the four kinds of quantitative measurement indexes, the quantitative classification method is established. And use it to score the sixteen kinds of common mechanical faults of the induction station. The result is consistent with statistical experience. At the same time, it is found that the progressive evolution of bearing fault has a major impact on the other mechanical faults.Secondly, the induction station of rigid flexible coupling virtual prototype model is established based on the three dimensional parametric model of package platform. Under the premise of the comprehensive consideration of its operating mechanism and structural characteristics,the modal and natural frequencies of the model are analyzed by using the modal theory. The result shows that the theoretical operating frequency is much larger than that of the simulation result, which verifies that the model is able to avoid the occurrence of resonance. Also found Where is the biggest vibration is roller, Which consistent with the reality. Finally the flexible roller axis is carried out based on the modal neutral file, and the model is established in accordance with the actual structure of the induction station. What's more, the modeling method provides a basis for studying the vibration position and characteristics of other post equipment rotating system.Thirdly, the research method of fault collaborative simulation for the induction station is presented. And the motion control of the package, belt,roller bearings and other components is realized on the virtual prototype for the induction station based on the packet flow analysis and collaborative simulation data communication. Then the simulation is carried out under different fault conditions, and velocity, acceleration,vibration amplitude and other data are collected by the virtual sensor at Marker point. Finally, the model is compared with the experimental platform. The result shows that the simulation data is in good agreement with the experimental data, which verifies the accuracy of the simulation model of belt and roller bearing fault conditions and operation. What's more, the method proposed in this research can also be used to study the operation fault simulation of other post rotating equipment.Fourthly, the single fault diagnosis model based on Support Vector Data Description(SVDD) for induction station bearing is presented. To solve the problem of background noise interference, the wavelet threshold method is used to deal with the noise reduction to improve the signal-to-noise ratio. Then, the relative complete bearing fault feature space is constructed from the time domain, frequency domain and time-frequency domain. A method based on time-spectral kurtosis is presented to filter and reconstruct the bearing fault signal, which enhances identification of resonance frequency bands. Finally, establish a model of SVDD single fault classifier by constructing the least feature hyper sphere, and introduce the relaxation factor to realize the bilateral support of the super sphere. At the same time, the parameters such as kernel function, width parameter and penalty factor are optimized. The result shows that the recognition rates of the normal class and the fault class are increased by 5% and 8.34% respectively compared with the Support Vector Machine(SVM) method.Fifthly, aiming to solve the problem of the limitation of SVDD algorithm for single fault recognition, a multi class fault diagnosis model based on grey correlation improved information entropy for induction station bearing is constructed. The annotation of important attributes of information entropy is enhanced to extract the characteristic information,based on the grey relational degree between the sample and the standard model. Then the SVM multi fault classifier is used to identify the multi class faults of the induction station bearing. And the result shows that the proposed model can be used as a priority decision factor in the multi fault classification, and the classification accuracy is 95.83%.Sixthly, to solve the problems caused by otherness, one-sidedness and the conflicts between the multiple channels, a multi channel fault information fusion decision model based on the improved evidence combination rule is established. The model firstly constructs the basic probability assignment function of the information structure according to the output of BP neural network. And the momentum factor is introduced into BP neural network to improve the learning rate. Then the combination rule of evidence is improved based on the uniform strength and support. Finally, use the new rules to fuse the output information of BP neural network in each channel, and use the maximum BPAF method to make the final decision. The result shows that the accuracy of multi channel information fusion decision-making is increased to 99.5%compared with that of the single channel.