www.gusucode.com > matlab神经网络原理与实例精解 本书源文件 > 第13章 神经网络应用实例/基于概率神经网络的柴油机故障诊断/diagnose.m

    % diagnose.m
% 柴油机故障诊断

%% 清空工作空间
clear,clc
close all

%% 定义训练样本和测试样本
% 故障1
pro1 = [1.97,9.5332,1.534,16.7413,12.741,8.3052;
    1.234,9.8209,1.531,18.3907,13.988,9.1336]';
% 故障2
pro2 = [0.7682,9.5489,1.497,14.7612,11.497,7.68;
    0.7053,9.5317,1.508,14.3161,11.094,7.3552]';
% 故障3
pro3 = [0.8116,8.1302,1.482,14.3171,11.1105,7.4967;
    0.816,9.0388,1.497,15.0079,11.6242,7.7604]';
% 故障4
pro4 = [1.4311,8.9071,1.521,15.746,12.0088,7.8909;
    1.4136,8.6747,1.53,15.3114,11.6297,7.5984]';
% 故障5
pro5 = [1.167,8.3504,1.51,12.8119,9.8258,6.506;
    1.3392,9.0865,1.493,15.0798,11.6764,7.8209]';
% 正常运转
normal = [1.1803,10.4502,1.513,20.0887,15.465,10.2193;
    1.2016,12.4476,1.555,20.6162,15.755,10.1285]';

% 训练样本
trainx = [pro1, pro2, pro3, pro4, pro5, normal];
% 训练样本的标签
trlab = 1:6;
trlab = repmat(trlab, 2, 1);
trlab = trlab(:)';

%% 样本的归一化,s为归一化设置
[x0,s] = mapminmax(trainx);

%% 创建概率神经网络
tic;
spread = 1;
net = newpnn(x0, ind2vec(trlab), spread);
toc

%% 测试
% 测试样本
testx = [0.7854,8.7568,1.4915,14.4547,11.1971,7.5071;
         1.1833,11.8189,1.5481,20.2626,15.5814,10.0646;
         0.661,8.8735,1.508,13.598,10.5171,6.9744;
         1.3111,7.9501,1.4915,14.9174,10.7511,7.7127;
         1.2394,9.6018,1.5366,18.219,13.851,9.0142;    
         1.2448,8.3654,1.5413,15.2558,11.5643,7.503]';
     
% 测试样本标签（正确类别）
testlab = [3,6,2,5,1,4];

% 测试样本归一化
xx = mapminmax('apply',testx, s);

% 将测试样本输入模型
s = sim(net,xx);

% 将向量形式的分类结果表示为标量
res = vec2ind(s);

%% 显示结果
strr = cell(1,6);
for i=1:6
   if res(i) == testlab(i)
       strr{i} = '正确';
   else
       strr{i} = '错误';
   end
end

diagnose_ = {'第一缸喷油压力过大','第一缸喷油压力过小', '第一缸喷油器针阀磨损',...
    '油路堵塞', '供油提前角提前 ','正常'};

fprintf('诊断结果：\n');
fprintf('  样本序号    实际类别    判断类别      正/误       故障类型 \n');
for i =1:6
   fprintf('     %d           %d         %d          %s      %s\n',...
       i, testlab(i), res(i), strr{i},  diagnose_{res(i)}); 
end