西储大学轴承故障数据和往复压缩机数据

function [Out_FDispEn, npdf]=FDispEn(x,m,nc,tau) % % This function calculates fluctuation-based dispersion entropy (FDispEn) of a univariate % signal x, using different mapping approaches (MA) % % Inputs: % % x: univariate signal - a vector of size 1 x N (the number of sample points) % m: embedding dimension % nc: number of classes (it is usually equal to a number between 2 and 8 - we used c=5 in Ref. [1]) % MA: mapping approach, chosen from the following options (we used 'LOGSIG' in [1]): % 'LM' (linear mapping), % 'NCDF' (normal cumulative distribution function), % 'TANSIG' (tangent sigmoid), % 'LOGSIG' 'logarithm sigmoid', % 'SORT' (sorting method). % % tau: time lag (it is usually equal to 1) % % Outputs: % % Out_DispEn: scalar quantity - the DispEn of x % npdf: a vector of length nc^m, showing the normalized number of fluctuation-based disersion patterns of x % % Example: FDispEn(rand(1,200),2,6,'LOGSIG',1) % % Ref: % % [1] H. Azami and J. Escudero, "Amplitude- and Fluctuation-based Dispersion Entropy", Entropy, 2018. % % [2] M. Rostaghi and H. Azami, "Dispersion Entropy: A Measure for Time-Series Analysis", IEEE Signal Processing Letters. vol. 23, n. 5, pp. 610-614, 2016. % % If you use the code, please make sure that you cite the references [1] and [2]. % % Hamed Azami and Javier Escudero Rodriguez % [email protected] and [email protected] % % 15-January-2018 %% N=length(x); sigma=std(x); mu=mean(x); %% Mapping approaches y=normcdf(x,mu,sigma); y=mapminmax(y,0,1);%使用mapminmax函数将信号映射到[0,1]区间内，不同数据具有相同尺度和范围，方便比较和分析 y(y==1)=1-1e-10; y(y==0)=1e-10; z=round(y*nc+0.5); all_patterns=[1:2*nc-1]'; for f=2:m-1 temp=all_patterns; all_patterns=[]; j=1; for w=1:2*nc-1 [a,b]=size(temp); all_patterns(j:j+a-1,:)=[temp,w*ones(a,1)]; j=j+a; end end N_PDE=(2*nc-1)^(m-1); for i=1:N_PDE key(i)=0; for ii=1:m-1 key(i)=key(i)*100+all_patterns(i,ii); end end for i_h=1:m %嵌入维数从i_h到m，当前的嵌入维数 %ind(i_h,:)=[(i_h-1)*tau+1:N-(m-1)*tau+(i_h-1)*tau];%起始索引为(i_h-1)tau+1，结束索引为N-(m-1)tau+(i_h-1)*tau。将这个范围存储在ind矩阵的第i_h行中。 ind =hankel(1:N-(m-1)*tau, N-(m-1)*tau:N)';%这两代码实现作用一样 end %ind =hankel(1:N-(m-1)*tau, N-(m-1)*tau:N)'; embd2 = z(ind(:, 1:tau:end)); dembd2=diff(embd2)'+nc; emb=zeros(N-(m-1)*tau,1); for i_e=m-1:-1:1 emb=dembd2(:,i_e)*100^(i_e-1)+emb; end pdf=zeros(1,N_PDE); for id=1:N_PDE [R,C]=find(emb==key(id)); pdf(id)=length(R); end npdf=pdf/(N-(m-1)*tau); p=npdf(npdf~=0); Out_FDispEn = -sum(p .* log(p));