工业大数据竞赛(基于深度学习的模型)-python黑洞网

本站消息

站长简介/公众号

出租广告位,需要合作请联系站长

齐天大圣

1103

文章

891018

访问

+关注

分类

暂无分类

日期归档

2023-05(3)

2023-06(1)

工业大数据竞赛(基于深度学习的模型)

发布于2019-08-22 16:33 阅读(124) 评论(0) 点赞(24) 收藏(2)


#数据处理代码
 
import pandas as pd
import numpy as np
 
#提取M1数据
pinlv=[]
#M_number=['M1','M2','M3','M4','M5']
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端X':[[]],'压缩机非联端Y':[[]],'压缩机联端X':[[]],'压缩机联端Y':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(7):
        file_count=len(os.listdir('D:\训练数据\M1\M1%s\%s'%(M_xilie[h],boxing[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(7):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\训练数据\M1\M1%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[x[:4] for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\训练数据\M1\M1%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        #import array
        #int16重要!
        #pinlv=np.array(pinlv,dtype=np.int16)
        #import wave
        #import numpy as np
        #import scipy.signal as signal
        #import scipy.io.wavfile
 
        #scipy.io.wavfile.write('D:\训练数据\M1\M1a\%s.wav' %(boxing[h]),44100,pinlv)   #7.2
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\训练数据\M1.csv')
import pandas as pd
import numpy as np
 
#提取M2数据
pinlv=[]
#M_number=['M1','M2','M3','M4','M5']
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端X':[[]],'压缩机非联端Y':[[]],'压缩机联端X':[[]],'压缩机联端Y':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(6):
        file_count=len(os.listdir('D:\训练数据\M2\M2%s\%s'%(M_xilie[h],boxing[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(6):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\训练数据\M2\M2%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[x[:4] for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\训练数据\M2\M2%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        #import array
        #int16重要!
        #pinlv=np.array(pinlv,dtype=np.int16)
        #import wave
        #import numpy as np
        #import scipy.signal as signal
        #import scipy.io.wavfile
 
        #scipy.io.wavfile.write('D:\训练数据\M1\M1a\%s.wav' %(boxing[h]),44100,pinlv)   #7.2
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\训练数据\M2.csv')
import pandas as pd
import numpy as np
 
#提取M3数据
pinlv=[]
#M_number=['M1','M2','M3','M4','M5']
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端X':[[]],'压缩机非联端Y':[[]],'压缩机联端X':[[]],'压缩机联端Y':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(7):
        file_count=len(os.listdir('D:\训练数据\M3\M3%s\%s'%(M_xilie[h],boxing[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(7):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\训练数据\M3\M3%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[x[:4] for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\训练数据\M3\M3%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        #import array
        #int16重要!
        #pinlv=np.array(pinlv,dtype=np.int16)
        #import wave
        #import numpy as np
        #import scipy.signal as signal
        #import scipy.io.wavfile
 
        #scipy.io.wavfile.write('D:\训练数据\M1\M1a\%s.wav' %(boxing[h]),44100,pinlv)   #7.2
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\训练数据\M3.csv')
import pandas as pd
import numpy as np
 
#提取M4数据
pinlv=[]
#M_number=['M1','M2','M3','M4','M5']
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端A','压缩机非联端B','压缩机联端A','压缩机联端B','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端A':[[]],'压缩机非联端B':[[]],'压缩机联端A':[[]],'压缩机联端B':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(7):
        file_count=len(os.listdir('D:\训练数据\M4\M4%s\%s'%(M_xilie[h],boxing[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(7):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\训练数据\M4\M4%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[x[:3] for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\训练数据\M4\M4%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        #import array
        #int16重要!
        #pinlv=np.array(pinlv,dtype=np.int16)
        #import wave
        #import numpy as np
        #import scipy.signal as signal
        #import scipy.io.wavfile
 
        #scipy.io.wavfile.write('D:\训练数据\M1\M1a\%s.wav' %(boxing[h]),44100,pinlv)   #7.2
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\训练数据\M4.csv')
import pandas as pd
import numpy as np
 
#提取M5数据
pinlv=[]
#M_number=['M1','M2','M3','M4','M5']
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端X':[[]],'压缩机非联端Y':[[]],'压缩机联端X':[[]],'压缩机联端Y':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(7):
        file_count=len(os.listdir('D:\训练数据\M5\M5%s\%s'%(M_xilie[h],boxing[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(7):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\训练数据\M5\M5%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[x[:4] for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\训练数据\M5\M5%s\%s\wave_%s.csv' %(M_xilie[g],boxing[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        #import array
        #int16重要!
        #pinlv=np.array(pinlv,dtype=np.int16)
        #import wave
        #import numpy as np
        #import scipy.signal as signal
        #import scipy.io.wavfile
 
        #scipy.io.wavfile.write('D:\训练数据\M1\M1a\%s.wav' %(boxing[h]),44100,pinlv)   #7.2
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\训练数据\M5.csv')
def xiaoshudian(x):
    t=0
    for i in range(len(x)):
        if x[i]=='.':
            t=t+1
        if t==2:
            return x[:i]
        if i==len(x)-1:
            return x
#xiaoshudian('1.1.23456789')
import pandas as pd
import numpy as np
 
#提取决赛数据
pinlv=[]
M_xilie=['_1','_2','_3','_4','_5']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)']
 
boxing2=['汽轮机非联端A','汽轮机非联端B','汽轮机联端A','汽轮机联端B','汽轮机位移A(波形)','汽轮机位移B(波形)']
 
fenlei=pd.DataFrame({'压缩机非联端X':[[]],'压缩机非联端Y':[[]],'压缩机联端X':[[]],'压缩机联端Y':[[]],'压缩机轴位移A(波形)':[[]],\
                     '压缩机轴位移B(波形)':[[]],'压缩机轴位移C(波形)':[[]],'label':[[]]})
 
file_all_count=[]
k=0
import os
for h in range(5):
    for i in range(6):
        file_count=len(os.listdir('D:\决赛数据\M18\M18%s\%s'%(M_xilie[h],boxing2[i])))
        file_all_count.append(file_count)
 
for g in range(5):
    for h in range(6):
        for i in range(file_all_count[k]):
            df=pd.read_csv('D:\决赛数据\M18\M18%s\%s\wave_%s.csv' %(M_xilie[g],boxing2[h],(i+1)),skiprows=5)
            df=df.drop(['wave'],axis=1)    
            #df.loc[i+1]=list(data)
            listdata=list(df)
            listdata=[xiaoshudian(x) for x in listdata]
            listdata=[float(x) for x in listdata]
 
            df2=pd.read_csv('D:\决赛数据\M18\M18%s\%s\wave_%s.csv' %(M_xilie[g],boxing2[h],(i+1)),nrows=1,header=None)
            freq=df2.loc[0,1]
 
            listdata2=[x*freq for x in listdata]
 
            pinlv=pinlv+listdata2    
 
        fenlei.loc[g,['%s'%(boxing[h])]]=[pinlv]
        pinlv=[]
        k=k+1
fenlei.to_csv('d:\决赛数据\M18.csv')
#转化为 .wav文件
import librosa
import numpy as np
import pandas as pd
import scipy
df=pd.read_csv('d:\决赛数据\M17.csv')
df2=df.drop(['Unnamed: 0'],axis=1)
 
M_xilie=['_1','_2','_3','_4','_5']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)']
 
for h in range(5):
    for i in range(6):
        df2.iloc[h,i]=df2.iloc[h,i][1:-1]
        pinlv=df2.iloc[h,i].split(',')
        pinlv=[float(x) for x in pinlv]
        pinlv=np.array(pinlv,dtype=np.int16)
        scipy.io.wavfile.write('D:\决赛数据\M17%s%s.wav' %(M_xilie[h],boxing[i]),44100,pinlv)


#模型的训练和检验
 
#建立模型一: 区分机器是否坏掉
#读入数据 X y
import librosa
import numpy as np
 
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
for g in range(5):
    for h in range(5):
        for i in range(7):
            
            if g==1 and i==6:
                break
                
            file_name='D:\训练数据\M%s%s%s.wav'%(g+1,M_xilie[h],boxing[i])  
            X, sample_rate = librosa.load(file_name, res_type='kaiser_fast') 
            mfccs = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=40).T,axis=0)
 
            if g==0 and h==0 and i==0:
                tezhen=mfccs
            else:
                tezhen=np.vstack((tezhen,mfccs))
 
from sklearn.preprocessing import LabelEncoder
from keras.utils import np_utils
 
X=tezhen
y=np.array([['a'],['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],['a'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            
            ['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],\
            ['a'],['a'],['a'],['a'],['a'],['a'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],\
            
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g'],\
            ['g'],['g'],['g'],['g'],['g'],['g'],['g']])
 
lb = LabelEncoder()
 
y = np_utils.to_categorical(lb.fit_transform(y))
#训练模型 1
import numpy as np
from keras.models import Sequential
from keras.layers import Dense,Dropout,Activation,Flatten,LSTM
from keras.layers import Convolution2D,MaxPooling2D
from keras.optimizers import Adam
from sklearn import metrics
 
num_labels = y.shape[1]
filter_size = 2
 
# build model
model=Sequential()
 
model.add(Dense(256,input_shape=(40,)))
model.add(Activation('relu'))
model.add(Dropout(0.1))
 
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.1))
 
model.add(Dense(num_labels))
model.add(Activation('softmax'))
 
model.compile(loss='binary_crossentropy',metrics=['accuracy'],optimizer='adam')
 
model.fit(X,y,batch_size=34,epochs=25)
 
 
#获取全组决赛数据
import librosa
import numpy as np
 
M_xilie=['_1','_2','_3','_4','_5']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)']
 
for g in range(8):
    for h in range(5):
        for i in range(6):
       
            file_name='D:\决赛数据\M%s%s%s.wav'%(g+11,M_xilie[h],boxing[i])
            X, sample_rate = librosa.load(file_name, res_type='kaiser_fast')
            mfccs = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=40).T,axis=0)
 
            if g==0 and h==0 and i==0:
                tezhen_juesai=mfccs
            else:
                tezhen_juesai=np.vstack((tezhen_juesai,mfccs))
X_juesai=tezhen_juesai
print(X_juesai.shape)
(240, 40)
#预测决赛全组数据
ar=model.predict_classes(X_juesai)
for i in np.arange(0,240,6):
    print(ar[i:i+6])
    if (i+1)%5==0:
        print('')
 
#建立模型二: 区分机器具体时间坏掉
#读入数据 X y
import librosa
import numpy as np
 
M_xilie=['a','b','c','d','e']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)','压缩机轴位移C(波形)']
 
for g in range(2):
    for h in range(5):
        for i in range(7):
            
            if g==1 and i==6:
                break
                
            file_name='D:\训练数据\M%s%s%s.wav'%(g+1,M_xilie[h],boxing[i])  
            X, sample_rate = librosa.load(file_name, res_type='kaiser_fast') 
            mfccs = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=40).T,axis=0)
 
            if g==0 and h==0 and i==0:
                tezhen=mfccs
            else:
                tezhen=np.vstack((tezhen,mfccs))
 
from sklearn.preprocessing import LabelEncoder
from keras.utils import np_utils
 
X2=tezhen
y2=np.array([['a'],['a'],['a'],['a'],['a'],['a'],['a'],\
            ['b'],['b'],['b'],['b'],['b'],['b'],['b'],\
            ['c'],['c'],['c'],['c'],['c'],['c'],['c'],\
            ['d'],['d'],['d'],['d'],['d'],['d'],['d'],\
            ['e'],['e'],['e'],['e'],['e'],['e'],['e'],\
            
            ['a'],['a'],['a'],['a'],['a'],['a'],\
            ['b'],['b'],['b'],['b'],['b'],['b'],\
            ['c'],['c'],['c'],['c'],['c'],['c'],\
            ['d'],['d'],['d'],['d'],['d'],['d'],\
            ['e'],['e'],['e'],['e'],['e'],['e']])
 
lb = LabelEncoder()
 
y2 = np_utils.to_categorical(lb.fit_transform(y2))
#训练模型 2
import numpy as np
from keras.models import Sequential
from keras.layers import Dense,Dropout,Activation,Flatten,LSTM
from keras.layers import Convolution2D,MaxPooling2D
from keras.optimizers import Adam
from sklearn import metrics
 
num_labels = y2.shape[1]
filter_size = 2
 
# build model
model2=Sequential()
 
model2.add(Dense(256,input_shape=(40,)))
model2.add(Activation('relu'))
model2.add(Dropout(0.1))
 
model2.add(Dense(64))
model2.add(Activation('relu'))
model2.add(Dropout(0.1))
 
model2.add(Dense(num_labels))
model2.add(Activation('softmax'))
 
model2.compile(loss='categorical_crossentropy',metrics=['accuracy'],optimizer='adam')
 
model2.fit(X2,y2,batch_size=34,epochs=25)
 
#获取故障组决赛数据
import librosa
import numpy as np
 
M_number=['11','16','17','18']
M_xilie=['_1','_2','_3','_4','_5']
boxing=['压缩机非联端X','压缩机非联端Y','压缩机联端X','压缩机联端Y','压缩机轴位移A(波形)','压缩机轴位移B(波形)']
 
for g in range(4):
    for h in range(5):
        for i in range(6):
       
            file_name='D:\决赛数据\M%s%s%s.wav'%(M_number[g],M_xilie[h],boxing[i])
            X, sample_rate = librosa.load(file_name, res_type='kaiser_fast')
            mfccs = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=40).T,axis=0)
 
            if g==0 and h==0 and i==0:
                tezhen_juesai=mfccs
            else:
                tezhen_juesai=np.vstack((tezhen_juesai,mfccs))
X_juesai=tezhen_juesai
print(X_juesai.shape)
(120, 40)
#预测故障组决赛数据
ar=model2.predict_classes(X_juesai)
for i in np.arange(0,120,6):
    print(ar[i:i+6])
    if (i+1)%5==0:
        print('')