锅炉承压管泄漏双曲面定位的遗传算法优化

Boiler Tube Leakage Hyperboloidal Location Optimization via Genetic Algorithm

电站锅炉"四管"泄漏的精确定位对于火电机组的安全、经济运行具有极其重要的意义。该文建立了双曲面定位模型并给出炉膛内平面四元阵列的测点分布,泄漏定位实验表明通过极大似然(maximum likelihood,ML)广义互相关对能取得非常稳定和尖锐的时间估计峰值,频域信噪比函数能较好地抑制炉膛强背景噪声的影响。为了避免初值猜测与远场假设,采用遗传算法:十进制浮点编码,非均匀高斯变异算子,进行双曲面方程组定位解的全局性概率搜索。对比遗传算法和拟牛顿法的双曲面方程组优化,表明遗传算法的优化误差略小于拟牛顿法,具有稳健性和高精度,但其需要较长的收敛时间。时延估计误差对定位影响很大,时间迟延误差不大于0.1ms,平面四元阵列空间坐标定位误差保持在1m之内从而满足精确定位的要求。

The main requirement of a modern power plant is an efficient and reliable operation;hence,the source of boiler tube leakage must be precisely localized.In this study,the quaternary acoustic array was used in the furnace,a set of hyperboloidal equations was established,and the sensors distributed for capturing the leakage signal were given.The experiment on leakage localization shows that a stable and sharp peak is obtained through an approximation of the maximum likelihood estimator which signal-to-noise ratio function on frequency domain performs well in inhibition of the background noise.In order to avoid initial guess and distant source assumption,this study employed adaptive Guassian mutation operator,which was implemented in the genetic algorithm（GA）,to provide an explicit solution.Compared with the quasi-Newton method,GA is more robust and accuracy,although it requires more time to converge.The time delay estimator error greatly influences the accuracy of the localization.When the time delay estimator error is below 0.1 ms,the quaternary plane array localization error is within the permitted range of 1 m,which is derived through numerical calculation.