基于dsp和分层时序记忆齿轮箱故障诊断系统word格式论文.docxVIP

Gearbox Fault Diagnosis System Based on DSP and Hierarchical Temporal MemoryAbstractGearbox is a key part that widely used in many mechanical equipments. Performing monitoring and fault diagnosis on gearbox plays an important role in industry security and productivity effect. In this paper, an online monitoring and diagnosis system is developed based on DSP in order to realize online diagnosis of gearbox, and a new algorithm named Hierarchial Temporal Memory (HTM) is applied on the system for testing the diagnosis performance.Using embedded system makes it easy to intergrate data acquisition, signal processing, feature extraction and fault diagnosis on embedded platfom as well as automatic fault diagnosis and many communication protocols. The system takes DSP-TMS320F2812 as the core processor. The hardware is builded by various peripheral modules of TMS320F2812, which include data acquisition of vibration signal, rotation speed signal and digital signal. The system can communicate with the upper monitor by GSM protocol, TCP/IP protocol and CAN bus. A lot of signal processing algorithms are embedded in TMS320F2812 for feature extraction of the gearbox such as FIR digital filter, power spectrum, zoom power spectrum and wavelet envelope analysis, etc. In additon, feature extraction functions are also embedded on DSP to realize automatic feature extraction of gearbox faults.This paper proposes a scheme that using Hierarchial Temporal Memory on the gearbox fault detection. First, the structure and principle of Hierarchical Temporal Memory is introduced. Then we transform the feature vector that obtained from vibration signal of six testing points on gearbox into bitmap format in order to meet the request of a HTM region. 60 sample vectors of 6 working states of gearbox are selected for training the HTM region until the region gets stable sparse distribute representation. After that, a conditional probability matrix is calculated according to the stable sparse distribute