| Rolling bearing as a key component of bridge cranes, directly determines the efficiency of cranes and the safety of workers. During the last decade, a variety of PC-based fault diagnosis of rolling bearing device has been developed at home and abroad, however, there are some shortcomings such as less flexibility in practical applications. For the diagnosis of small sample learning problems, a lot of good theory base and support vector machine multi-classification algorithm with the generalization ability have been put forward by some researchers, but most of these algorithms are still just verified by simulation and varely verified by other experimental platforms.This dissertation analyzes the domestic and international issues related to the research methods, according to the poor flexibility of PC-based diagnostic devices, based on ARM9core and FPGA was designed. Firstly, transplantation of the embedded systems was completed, including U-Boot, Linux kernel, root file system and Qtopia. Secondly, concerning the design of the algorithm, two types’Support Vector Machine classification algorithm of separability largest multi-class were used in rolling bearing fault diagnosis, completing the automatic identification of four states. Finally, this paper completed the simulation of two types’algorithm of separability largest multi-class.The embedded system was used in this dissertation, which makes the devices of the bearing fault diagnosis more and more exquisite, consequently, systemic resources could be customized by the desired function. The adopting FPGA program greatly promoted the ability of the fault diagnosis, which laid a solid foundation for the further data processing and fault identification. Finally, two types’support vector machine classification algorithm of separability largest multi-class were validated successfully by experimental results in the platform, and the success rate of diagnosis is up to94.5%. |
