论文部分内容阅读
为了提高核电站松动件故障诊断的能力 ,结合时频变换理论和神经网络理论 ,提出了对松动件碰撞位置进行估计的新方法。通过对安装在反应堆压力壳上的多个加速度传感器的信号进行采集 ,并经过信号预处理、时频变换、神经网络计算等过程 ,实现对核电站松动件碰撞位置的定位。经过对 10 0多次实验数据的学习 ,发现神经网络计算结果能够满足核电站松动件定位精度要求。该方法适用于复杂连接结构、以及复杂几何曲面装置的松动件定位问题 ,减少了人为因素的影响 ,具有比较高的自动处理功能和定位精度
In order to improve the capability of fault diagnosis of loose parts in nuclear power plants, a new method for estimating the collision position of loose parts is proposed based on the time-frequency transformation theory and neural network theory. The signals of multiple accelerometers mounted on the pressure vessel of the reactor are collected and processed by signal preprocessing, time-frequency transformation and neural network calculation to locate the collision position of the loose parts of the nuclear power plant. After learning more than 10 experimental data, it is found that the neural network calculation results can meet the requirements of positioning accuracy of loose parts of nuclear power plants. The method is suitable for the problem of locating the loose parts of a complicated connection structure and a complicated geometric curved surface device, reducing the influence of human factors, and has relatively high automatic processing function and positioning accuracy