BP预测.rar_Bp 故障_故障 预测_机械轴承的特征故障分类识别_轴承_轴承故障分类

close all; clear all; clc; N=10; [FileName, PathName] = uigetfile({'*.xls','Excel File(*.xls)';... '*.txt','TXT File(*.txt)';'*.*','All File(*.*)'},'选择文件'); str = [PathName, FileName]; % 获取文件的路径 train_x=[]; test_x=[]; for sheet=1:6 [data,date,f] = xlsread(str,sheet); % 读取所选文件 data1=data(:,2:end); train_x=[train_x; data1(1:8,:)]; test_x=[test_x; data1(9:end,:)]; end %% %(1,0,0,0,0,0)对应第一种故障0.06 %（0,0,0,0,0,1）对应最后一种故障0.03 train_y=[]; temp=eye(6); for i=1:6 train_y=[train_y;ones(8,1)*temp(i,:)]; end net=newff(minmax(train_x'),train_y',{5 5},{'purelin' 'purelin'},'trainlm'); %构造神经网络net=newff(threshold,[15,1],{'tansig','logsig'},'traingdx'); %上面这句话的意思是建立新的神经网络net，newff(输入，输出，{隐层数量及该层的节点个数}，{传输函数名}，训练函数)；{10}代表单隐层， %并且该层节点为10个，如果想显示多隐层，比如双隐层，{10,15}这样就行，传输函数也要相应增加，比如{'tansig','logsig'}节点数和传输函数名可以随意调整， %目的当然是让预测结果更精确 net.trainParam.show=50;% net.trainParam.goal = 0.00001; %设置精确度 net.trainParam.epochs = 500; %设置训练次数，5000次 [net,tr] = train(net,train_x',train_y'); %训练开始 net.iw{1,1}%隐层权值 net.b{1}%隐层阈值 net.lw{2,1}%输出层权值 net.b{2}%输出层阈值 %验证测试 test_idx=[]; for i=1:size(test_x,1) a=sim(net,test_x(i,:)') max_a=max(a); kk=find(a==max_a); test_idx=[test_idx;kk]; end %% shu=[]; for i=1:6 shu=[shu; i*ones(2,1)]; end disp('测试结果'); test_idx figure(1); plot(1:length(shu),shu,'r-o'); hold on; plot(1:length(shu),test_idx,'b-*'); %% xlabel('测试样本数量'); ylabel('所属类别'); inum=0; for i=1:12 if(shu(i)==test_idx(i)) inum=inum+1; end end rate=inum/length(test_idx)