Technology Study On Fault Characteristics Extraction And Classification Of High-speed Ring-spinning Spindle

Spindle is one of the main equipments in textile industry, its dynamic character is concerned with yarn's quality, and also it reflects the development of the equipments of the textile. So it is necessary to study the performance of the spindle.With a complex structure, spindle consists of a lot of parts. Every part can influence the whole spindle's performance. The construction of the spindle is introduced briefly in this paper. In the experiment of spindle, the acceleration sensor is used to receive the signal of the vibration.Theory of wavelet transformation and Multiresolution Analysis is introduced in this paper. Also it introduced the character, model and the learning methods of the neural networks, particularly the BP networks and also the algorithm of Levenberg-Marquardt and so on.The diagnosis system include two parts, spindle testing and vibration analysis. The testing system consists of acceleration sensor, charge amplifier, sampling card, computer and a set of software, the analysis system consists of signal analysis and neural networks. In Windows environment, the software designed with Visual C⁺⁺ can perform sampling processing, diagnosing and saving in real time. As to the vibration of spindle from different producers, a lot of experiments are done. A series of acceleration signal is gotten from spindle under different rotary speed. By processing both in time domain and in frequency domain, the information to evaluate the dynamic character becomes available. The cause to the violent vibration is found.A lot of experiments are done, and a lot of data are processed. After wavelet transformation of the vibration signal of the spindle Model DFG2 with normal and abnormal vibration character, in the third vibration spectrum, it is found that abnormally vibrating spindle has an outstanding spectrum peak and some a little lower ones which are consistent with spindle's rotary frequency and its some times frequency. It evaluated that the deformation of blade and wharve's assemblage and the loose of the upper bearing have enhanced the vibration. As to the spindle Model D1203, in the same way, the third vibration spectrum shows that a frequency increases with the rising of the spindle rotation speed. This frequency is near the inherent frequency of the spindle bolster, so the defects of the spindle may be here. The upper bearing and the lower bearing are not in the same axis caused the rotary friction.After the wavelet transformation and signal analysis of the spindle vibration, the characteristics of the spindle vibration can be extracted. These characteristics are permuted sequentially to form a vector. The neural network is trained by the vector of the spindle with typical faults,and then the previous spindle with abnormal vibration can be simulatedby the neural network which is trained already. The results of thesimulation and the previous analysis are very similar.Finally, a summary is made to the research work, and some ideas areput forwarded on the further research.