💥💥💞💞欢迎来到本博客❤️❤️💥💥
🏆博主优势:🌞🌞🌞博客内容尽量做到思维缜密,逻辑清晰,为了方便读者。
⛳️座右铭:行百里者,半于九十。
📋📋📋本文目录如下:🎁🎁🎁
目录
💥1 概述
文献来源:
摘要:
本文考虑了基于逆瞬态波理论的充水管道中多次泄漏的识别问题。该问题的解析解决方案涉及各种泄漏之间的非线互项。本文从分析和数值上表明,这些非线性项的泄漏大小与幂二和的量级相同;因此,可以忽略不计。由于这种简化,制定并测试了分别识别泄漏位置和泄漏大小的最大似然 (ML) 方案。结果表明,ML估计方案在噪声方面具有高效和鲁棒性。此外,ML方法是一种超分辨率泄漏定位方案,因为它的可解析泄漏距离(大约0.15λ,其中λ是最小波长)低于奈奎斯特-香农采样定理极限(0.5λ)。此外,还推导了 Cramér-Rao 下界 (CRLB) 并用于显示 ML 方案估计的效率。ML估计器的方差近似于CRLB,证明ML方案属于泄漏定位方法的最佳无偏估计器类。
关键词:
泄漏检测 缺陷检测 瞬态波 最大似然性 超分辨率 Cramér-Rao下限
📚2 运行结果
部分代码:
% this figure shows the accuracy of linearization
figure;
plot(omega/omega_th,abs(h_x2)); hold on;
plot(omega/omega_th,abs(h_x2_approximation),'r--');
xlabel('\omega/\omega_{th}','Interpreter','Tex'); ylabel('Head magnitude [m]');
legend('Transfer matrix','Linearized model')
title('Frequency Response Diagram');
h_NL = zeros(length(omega),M); % Eq. (14)
for mm = 1:M
h_NL(:,mm) = -Z.'.*sinh(xM(mm)*mu.').*q_x1 + cos(xM(mm)*mu.').*h_x1;
end
noise_method = 2;
SNR = 10;
sigma = 1/(10^(SNR/20));
h = h_hydro - h_NL;
h_MFP = real(h)+sigma*35*randn(size(h)) + 1j*imag(h)+sigma*13*randn(size(h));
h_MFP = h_MFP(:);
h_xM_up_esti = real(h_xM_up)+sigma*mean(mean(abs(real(h))))*randn(size(h_xM_up))+1j*imag(h_xM_up)+sigma*mean(mean(abs(imag(h))))*randn(size(h_xM_up));
q_x1_esti = (-h_xM_up_esti.'./(Z.*sinh(xM_up*mu))).'; % Eq. (17)
%% leak localization using MFP (1D search)
% For MFP, please refer to: X. Wang and M. S. Ghidaoui, Pipeline leakage detection using the matched-field processing method, Journal of Hydraulic Engineering, 144(6), 04018030, 2018
xL_test = 1:1:L-1; % discrete points for plotting cost function along the pipe
cost_MFP = zeros(size(xL_test));
S_MFP = zeros(size(xL_test));
G_all = zeros(length(xL_test),length(omega)*M);
for xx = 1:length(xL_test)
G = zeros(length(omega),M);
for mm = 1:M
if xL_test(xx)>xM(mm)
G(:,mm) = zeros(J,1);
else
G(:,mm) = -sqrt(9.8/2/(H1-(H1-H2)/L*xL_test(xx)))*Z.'.*sinh((xM(mm)-xL_test(xx))*mu.').*(Z.'.*sinh(xL_test(xx)*mu.').*q_x1_esti);
end
end
G = G(:);
G_all(xx,:) = G;
% if noise_method == 3 % whitening blue noise
% G = 1./sigma_blue'.*G;
% end
cost_MFP(xx) = h_MFP'*G*G'*h_MFP/(G'*G); % Eq. (19) in the paper
S_MFP(xx) = G'*h_MFP/(G'*G); % Eq. (20) in the paper
end
CdAl_MFP = S_MFP;
[~,x_max] = max(cost_MFP);
if no_L==1
disp(['Conventional MFP: the estimated leak position is ',num2str(xL_test(x_max)),' m and the estimated leak size is CdAl=',num2str(abs(CdAl_MFP(x_max)))]);
disp(['Conventional MFP: the localization error is ',num2str(abs(xL_test(x_max)-xL(1))),' m ']);
end
figure;
plot(xL_test,abs(cost_MFP)); hold on; % plot(xL_test,abs(S_MFP),'r--');
for ii = 1:length(xL)
plot([xL(ii) xL(ii)],[min(abs(cost_MFP)) 1.1*max(abs(cost_MFP))],'r--');
end
plot(xM,ones(size(xM))*min(abs(cost_MFP)),'rx','MarkerSize',8,'Linewidth',1.5);
plot(xM_up,min(abs(cost_MFP)),'rx','MarkerSize',8,'Linewidth',1.5);
xlabel('x [m]'); ylabel('|B|^2','Interpreter','Tex')
ylim([min(abs(cost_MFP)) 1.1*max(abs(cost_MFP))]);
🎉3 参考文献
文章中一些内容引自网络,会注明出处或引用为参考文献,难免有未尽之处,如有不妥,请随时联系删除。