论文部分内容阅读
研究在仪器仪表中利用分形插值和人工神经网络算法改善测试精度和响应时间的方法。在仪表检测中有时会遇到检测的分辨率与实时性相冲突的情况 ,此时对数据进行插值是一个很好的解决办法。利用分形和人工神经网络算法插值是一种可以进行多点数插值的优越方法 ,因为它可以通过训练学习不断修正网络的权值 ,使检测误差的方差控制在预定的范围。还研究了利用神经网络做谱分析来求取主频的方法 ,它在运算速度和分辨率方面都优于 FFT。文章还提出了一些减少人工神经网络学习训练时间的方法。
Research in instrumentation using fractal interpolation and artificial neural network algorithm to improve the test accuracy and response time of the method. In the detection of instruments sometimes encountered in the detection of resolution and real-time conflict, the data interpolation is a good solution. Interpolation using fractal and artificial neural network algorithm is a superior method for multi-point interpolation, because it can constantly modify the weight of the network through training and learning to control the variance of detection error within a predetermined range. Also studied using neural network spectral analysis to find the frequency method, which is superior to FFT speed and resolution. The article also put forward some ways to reduce the training time of artificial neural network learning.