forked from schwxd/LSTM-Keras-CMAPSS
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathCMAPSSDataset.py
180 lines (149 loc) · 8.68 KB
/
CMAPSSDataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
import tensorflow as tf
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.preprocessing import StandardScaler
# column names of CMAPSS Dataset
# CMAPSS数据集列名
columns = ['id', 'cycle', 'setting1', 'setting2', 'setting3', 's1', 's2', 's3','s4', 's5', 's6', 's7', 's8',
's9', 's10', 's11', 's12', 's13', 's14', 's15', 's16', 's17', 's18', 's19', 's20', 's21']
feature_columns = ['setting1', 'setting2', 'setting3', 's1', 's2', 's3','s4', 's5', 's6', 's7', 's8',
's9', 's10', 's11', 's12', 's13', 's14', 's15', 's16', 's17', 's18', 's19', 's20', 's21', 'cycle_norm']
class CMAPSSDataset():
def __init__(self, fd_number, batch_size, sequence_length):
super(CMAPSSDataset).__init__()
self.batch_size = batch_size
self.sequence_length = sequence_length
self.train_data = None
self.test_data = None
self.train_data_encoding = None
self.test_data_encoding = None
# \s+ 匹配一个或多个空格
data = pd.read_csv("C-MAPSS-Data\\train_FD00" + fd_number + ".txt", delimiter="\s+", header=None)
data.columns = columns
# 计算该数据集包含的engine数目
self.engine_size = data['id'].unique().max()
# 计算每一行的剩余cycle
rul = pd.DataFrame(data.groupby('id')['cycle'].max()).reset_index()
rul.columns = ['id', 'max']
data = data.merge(rul, on=['id'], how='left')
data['RUL'] = data['max'] - data['cycle']
#data.drop(['cycle', 'setting1', 'setting2', 'setting3'], axis=1, inplace=True)
data.drop(['max'], axis=1, inplace=True)
# 将id之外的列正规化
self.std = StandardScaler()
data['cycle_norm'] = data['cycle']
cols_normalize = data.columns.difference(['id', 'cycle', 'RUL'])
norm_data = pd.DataFrame(self.std.fit_transform(data[cols_normalize]),
columns=cols_normalize, index=data.index)
join_data = data[data.columns.difference(cols_normalize)].join(norm_data)
self.train_data = join_data.reindex(columns=data.columns)
# 读取测试数据集并执行相同操作
# 测试集engine完整的rul还需要包括RUL_FD00x.txt文件中的部分
test_data = pd.read_csv("C-MAPSS-Data\\test_FD00" + fd_number + ".txt", delimiter="\s+", header=None)
test_data.columns = columns
truth_data = pd.read_csv("C-MAPSS-Data\\RUL_FD00" + fd_number + ".txt", delimiter="\s+", header=None)
truth_data.columns = ['truth']
truth_data['id'] = truth_data.index + 1
test_rul = pd.DataFrame(test_data.groupby('id')['cycle'].max()).reset_index()
test_rul.columns = ['id', 'elapsed']
test_rul = test_rul.merge(truth_data, on=['id'], how='left')
test_rul['max'] = test_rul['elapsed'] + test_rul['truth']
test_data = test_data.merge(test_rul, on=['id'], how='left')
test_data['RUL'] = test_data['max'] - test_data['cycle']
test_data.drop(['max'], axis=1, inplace=True)
test_data['cycle_norm'] = test_data['cycle']
norm_test_data = pd.DataFrame(self.std.fit_transform(test_data[cols_normalize]),
columns=cols_normalize, index=test_data.index)
join_test_data = test_data[test_data.columns.difference(cols_normalize)].join(norm_test_data)
self.test_data = join_test_data.reindex(columns=test_data.columns)
def get_train_data(self):
return self.train_data
def get_test_data(self):
return self.test_data
def get_feature_slice(self, input_data):
# Reshape the data to (samples, time steps, features)
def reshapeFeatures(input, columns, sequence_length):
data = input[columns].values
num_elements = data.shape[0]
#print(num_elements)
for start, stop in zip(range(0, num_elements-sequence_length), range(sequence_length, num_elements)):
yield(data[start:stop, :])
feature_list = [list(reshapeFeatures(input_data[input_data['id'] == i], feature_columns, self.sequence_length))
for i in range(1, self.engine_size + 1) if len(input_data[input_data['id'] == i]) > self.sequence_length]
##for i in range(len(feature_list)):
## print(np.array(feature_list[i]).shape)
feature_array = np.concatenate(list(feature_list), axis=0).astype(np.float32)
length = len(feature_array) // self.batch_size
return feature_array[:length*self.batch_size]
#
# get the engine id of dataset
# In VAE the encoded dataset need to be reshape again (using sliding window within each engine)
# so the engine id need to be reserved
# #
def get_engine_id(self, input_data):
def reshapeLabels(input, sequence_length, columns=['id']):
data = input[columns].values
num_elements = data.shape[0]
return(data[sequence_length:num_elements, :])
label_list = [reshapeLabels(input_data[input_data['id'] == i], self.sequence_length)
for i in range(1, self.engine_size+1)]
label_array = np.concatenate(label_list).astype(np.int8)
length = len(label_array) // self.batch_size
return label_array[:length*self.batch_size]
def get_label_slice(self, input_data):
def reshapeLabels(input, sequence_length, columns=['RUL']):
data = input[columns].values
num_elements = data.shape[0]
return(data[sequence_length:num_elements, :])
label_list = [reshapeLabels(input_data[input_data['id'] == i], self.sequence_length)
for i in range(1, self.engine_size+1)]
label_array = np.concatenate(label_list).astype(np.float32)
length = len(label_array) // self.batch_size
return label_array[:length*self.batch_size]
# 每个engine只取最后一个sequence_length（如果该engine的数据条目数大于sequence_length的话，否则就舍弃）
# 用于最后的evaluation
def get_last_data_slice(self, input_data):
num_engine = input_data['id'].unique().max()
test_feature_list = [input_data[input_data['id'] == i][feature_columns].values[-self.sequence_length:]
for i in range(1, num_engine+1) if len(input_data[input_data['id'] == i]) >= self.sequence_length]
test_feature_array = np.asarray(test_feature_list).astype(np.float32)
length_test = len(test_feature_array) // self.batch_size
test_label_list = [input_data[input_data['id'] == i]['RUL'].values[-1:]
for i in range(1, num_engine+1) if len(input_data[input_data['id'] == i]) >= self.sequence_length]
test_label_array = np.asarray(test_label_list).astype(np.float32)
length_label = len(test_label_array) // self.batch_size
return test_feature_array[:length_test*self.batch_size], test_label_array[:length_label*self.batch_size]
#
def set_test_data_encoding(self, test_data_encoding):
self.test_data_encoding = test_data_encoding
def set_train_data_encoding(self, train_data_encoding):
self.train_data_encoding = train_data_encoding
if __name__ == "__main__":
datasets = CMAPSSDataset(fd_number='1', batch_size=10, sequence_length=50)
train_data = datasets.get_train_data()
train_feature_slice = datasets.get_feature_slice(train_data)
train_label_slice = datasets.get_label_slice(train_data)
train_engine_id = datasets.get_engine_id(train_data)
print("train_data.shape: {}".format(train_data.shape))
print("train_feature_slice.shape: {}".format(train_feature_slice.shape))
print("train_label_slice.shape: {}".format(train_label_slice.shape))
print("train_engine_id.shape: {}".format(train_engine_id.shape))
test_data = datasets.get_test_data()
print("test_data.shape: {}".format(test_data.shape))
test_feature_slice = datasets.get_feature_slice(test_data)
test_label_slice = datasets.get_label_slice(test_data)
test_engine_id = datasets.get_engine_id(test_data)
print("test_feature_slice.shape: {}".format(test_feature_slice.shape))
print("test_label_slice.shape: {}".format(test_label_slice.shape))
print("test_engine_id.shape: {}".format(test_engine_id.shape))
"""
np.savetxt('train_engine_id.txt', train_engine_id, fmt='%d')
np.savetxt('test_engine_id.txt', test_engine_id, fmt='%d')
data_batch = datasets.get_train_dataset_batch()
print(type(data_batch))
print(np.array(data_batch).shape)
data_batch_tensor = tf.convert_to_tensor(data_batch)
print(data_batch_tensor.shape)
"""