Study On Artificial Neural Network Modeling For Dynamic Measurement Errors And Experiment Research

With the development of technology, it is required higher on measurement technology to develop to a higher level. Dynamic measurement has gradually become the mainstream of modern measurement now. It, how to improve measurement accuracy, is always peoples' research focus in engineering measurement and instrument design. Great attention has been aroused in dynamic accuracy, which becomes an urgent need to address issues in theoretical research on accuracy.According neural network theory, we studied the method of modeling for Dynamic Error. Two kinds of methods, BP network and RBF network, were used to model and predict for the Dynamic Error.In a self-designed dynamic error experimental system, the standard signal (motor pulses) and the measured signal (grating pulses) were collected synchronously, and isolated the dynamic error. Through the design of the circuit of the experimental system, the signal was collected, counted and contrasted. Differential treatment was used in motor output signal with Am261s32;and counting was carried on with 74LS161 and 74HC573; and data was acquired with AC6651;Counting result was improved due to improved experimental device and HCTL-2020 with more counting median; and interface was designed with LABVIEW, with which data can be preserved real-time.Through the installing and commissioning for experimental device, we successfully separated the dynamic angular displacement measurement error. Based on BP and RBF neural network model, the dynamic error of experiment was modeled and forecasted.