The Research And Analysis Of Photoelectricity Torque Measurement System

In the mechanical transmission system, torque is the important performance index,which reflects the system performances of equipment. Now, micro、intelligence andNon-contact is the trend of torque sensor. The method for measuring torque based onphoto-electricity in principle has more advantages, such as, which are non-contact,non-intervention, anti-electromagnetic interference. These are the trends of torquemeasurement.Digital signal processor (DSP) has strong operation ability, which can achieve highspeed data acquisition and complex algorithm operations. But it is not suitable for logiccontrol and high real-time data collection. Field programmable gate array (FPGA) is richin resources, flexible interface, which can realize parallel operation and all of its controllogic has been completed by hardware, but it is not suitable for complex arithmetic. Inorder to improve the real-time dynamic measuring ability of torque, This paper usedphoto-electricity multiple code-tracks technology measure torque, analyzed the principleof torque measurement and designed a real-time multi-channel acquisition system basedon DSP and FPGA. In order to acquire the data rapidly and accurately，analysis andcomparison of the multi-channel signal feature and to compare the state of variation ofresults as the count basis through multi-channel square wave signal subdivision countcycle. This paper has used the combination of DSP and FPGA method on the hardware,The usage of FPGA can realize acquisition multi-channel signal, data cache and peripheralcontrol, use of DSP for signal processing and complete data transmission, the data istransmitted to the computer for storage, display and analysis, etc.Because of the system itself and external interference factors, the measurement oftorque has fluctuations of up and down, This paper used EMD threshold denoisingmethods denoising the vibration signal and extraction the stable signal waveform. Andused Hilbert-Huang transform for signal analysis and feature extraction in the same time, amethod which combines support vector regression (SVR) machine with mirror extensionwas proposed Because of Hilbert-Huang transform existed endpoint effect, Finally this paper make endpoint effect got restrain effectively.