我是keras /张量流的新手，我曾尝试在StackOverFlow上使用一些答案，但没有一个对我有用。也许对我来说这是不正确的实现，但我进行了很多搜索，但似乎没有任何效果。
我的代码的基本目标是创建一个神经网络，以优化围绕升/降系数的机翼设计，该神经网络仅使用迎角值。为了说服，输入和除输出基本上是相同的，但是损失函数是唯一的，因为它测试了期望值和输出的升/降系数的alpha值，然后将两者相减，以查看是否有改善。使用改善值，损失等于改善的1分频，因为大的改善将意味着较小的损失。另外，如果结果为负，则我有一个if语句返回高价值损失。另外，我使用的是google colab，因此有时导入的代码位于不同的代码块中。
航空模拟：
%cd OpenAeroStruct
from __future__ import division, print_function
import numpy as np
from openaerostruct.geometry.utils import generate_mesh
from openaerostruct.integration.aerostruct_groups import AerostructGeometry, AerostructPoint
from openmdao.api import IndepVarComp, Problem, Group
from openaerostruct.utils.constants import grav_constant
# Create a dictionary to store options about the surface
mesh_dict = {'num_y' : 5,
'num_x' : 2,
'wing_type' : 'CRM',
'symmetry' : True,
'num_twist_cp' : 5}
mesh, twist_cp = generate_mesh(mesh_dict)
surface = {
# Wing definition
'name' : 'wing', # name of the surface
'symmetry' : True, # if true, model one half of wing
# reflected across the plane y = 0
'S_ref_type' : 'wetted', # how we compute the wing area,
# can be 'wetted' or 'projected'
'fem_model_type' : 'tube',
'thickness_cp' : np.array([.1, .2, .3]),
'twist_cp' : twist_cp,
'mesh' : mesh,
# Aerodynamic performance of the lifting surface at
# an angle of attack of 0 (alpha=0).
# These CL0 and CD0 values are added to the CL and CD
# obtained from aerodynamic analysis of the surface to get
# the total CL and CD.
# These CL0 and CD0 values do not vary wrt alpha.
'CL0' : 0.0, # CL of the surface at alpha=0
'CD0' : 0.015, # CD of the surface at alpha=0
# Airfoil properties for viscous drag calculation
'k_lam' : 0.05, # percentage of chord with laminar
# flow, used for viscous drag
't_over_c_cp' : np.array([0.15]), # thickness over chord ratio (NACA0015)
'c_max_t' : .303, # chordwise location of maximum (NACA0015)
# thickness
'with_viscous' : True,
'with_wave' : False, # if true, compute wave drag
# Structural values are based on aluminum 7075
'E' : 70.e9, # [Pa] Young's modulus of the spar
'G' : 30.e9, # [Pa] shear modulus of the spar
'yield' : 500.e6 / 2.5, # [Pa] yield stress divided by 2.5 for limiting case
'mrho' : 3.e3, # [kg/m^3] material density
'fem_origin' : 0.35, # normalized chordwise location of the spar
'wing_weight_ratio' : 2.,
'struct_weight_relief' : False, # True to add the weight of the structure to the loads on the structure
'distributed_fuel_weight' : False,
# Constraints
'exact_failure_constraint' : False, # if false, use KS function
}
# Create the problem and assign the model group
prob = Problem()
# Add problem information as an independent variables component
indep_var_comp = IndepVarComp()
indep_var_comp.add_output('v', val=248.136, units='m/s')
indep_var_comp.add_output('alpha', val=5., units='deg')
indep_var_comp.add_output('Mach_number', val=0.84)
indep_var_comp.add_output('re', val=1.e6, units='1/m')
indep_var_comp.add_output('rho', val=0.38, units='kg/m**3')
indep_var_comp.add_output('CT', val=grav_constant * 17.e-6, units='1/s')
indep_var_comp.add_output('R', val=11.165e6, units='m')
indep_var_comp.add_output('W0', val=0.4 * 3e5, units='kg')
indep_var_comp.add_output('speed_of_sound', val=295.4, units='m/s')
indep_var_comp.add_output('load_factor', val=1.)
indep_var_comp.add_output('empty_cg', val=np.zeros((3)), units='m')
prob.model.add_subsystem('prob_vars',
indep_var_comp,
promotes=['*'])
aerostruct_group = AerostructGeometry(surface=surface)
name = 'wing'
# Add tmp_group to the problem with the name of the surface.
prob.model.add_subsystem(name, aerostruct_group)
point_name = 'AS_point_0'
# Create the aero point group and add it to the model
AS_point = AerostructPoint(surfaces=[surface])
prob.model.add_subsystem(point_name, AS_point,
promotes_inputs=['v', 'alpha', 'Mach_number', 're', 'rho', 'CT', 'R',
'W0', 'speed_of_sound', 'empty_cg', 'load_factor'])
com_name = point_name + '.' + name + '_perf'
prob.model.connect(name + '.local_stiff_transformed', point_name + '.coupled.' + name + '.local_stiff_transformed')
prob.model.connect(name + '.nodes', point_name + '.coupled.' + name + '.nodes')
# Connect aerodyamic mesh to coupled group mesh
prob.model.connect(name + '.mesh', point_name + '.coupled.' + name + '.mesh')
# Connect performance calculation variables
prob.model.connect(name + '.radius', com_name + '.radius')
prob.model.connect(name + '.thickness', com_name + '.thickness')
prob.model.connect(name + '.nodes', com_name + '.nodes')
prob.model.connect(name + '.cg_location', point_name + '.' + 'total_perf.' + name + '_cg_location')
prob.model.connect(name + '.structural_mass', point_name + '.' + 'total_perf.' + name + '_structural_mass')
prob.model.connect(name + '.t_over_c', com_name + '.t_over_c')
# Set up the problem
prob.setup()
# Set the alpha in the problem and run analysi
alphas = np.linspace(-5,5)
# Loopo through each alpha
def testvalue(num):
alphas = np.linspace(num,num)
# Loopo through each alpha
for alpha in alphas:
prob['alpha'] = alpha
prob.run_model()
value = prob['AS_point_0.wing_perf.CL']/prob['AS_point_0.wing_perf.CD']
return(value[0])
break
测试一下是否可以使用机翼：
def testifitworks(totalarray):
try:
return testvalue(totalarray)
print(totalarray)
except:
print(totalarray)
return -100
keras部分：
# Artificial Neural Network
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import keras.backend as K
# Importing the Keras
import keras
import tensorflow as tf
# Importing the dataset
dataset = pd.read_csv('letstryadifferentname.csv')
X = dataset.iloc[:, :1].values
y = dataset.iloc[:, 0].values
# Encoding categorical data
# Splitting the dataset into the Training set and Test set
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 0)
# Feature Scaling
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)
def newLoss(layer):
def loss(y_true, y_pred):
print(y_true)
print(0)
if tf.contrib.framework.is_tensor(y_true) or tf.contrib.framework.is_tensor(y_pred):
print(1)
if((testifitworks(K.int_shape(y_pred)[0])-testifitworks(K.int_shape(y_true)[0]))<=0):
print(2)
return K.abs(100)
else:
print(3)
return K.abs(1/(testifitworks(K.int_shape(y_pred)[0])-testifitworks(K.int_shape(y_true)[0])))
else:
print(4)
if testifitworks((y_pred)[0])-testifitworks((y_true)[0])<=0:
print(5)
return 100
else:
print(6)
return K.abs(1/(testifitworks((y_pred)[0])-testifitworks((y_true)[0])))
#print(tf.keras.backend.get_value(y_pred))
#print(tf.shape(y_true)[0])
return loss
#returning a function
# Importing the Keras libraries and packages
#import keras
from keras.models import Sequential
from keras.layers import Dense
# Initialising the ANN
classifier = Sequential()
# Adding the input layer and the first hidden layer
classifier.add(Dense(units = 6, kernel_initializer = 'uniform', activation = 'relu', input_dim = 1))
# Adding the second hidden layer
classifier.add(Dense(units = 6, kernel_initializer = 'uniform', activation = 'relu'))
# Adding the output layer
classifier.add(Dense(units = 1, kernel_initializer = 'uniform', activation = 'sigmoid'))
# Compiling the ANN
classifier.compile(optimizer = 'adam', loss = newLoss(classifier), metrics = ['accuracy'])
# Fitting the ANN to the Training set
classifier.fit(X_train, y_train, batch_size = 10, epochs = 10)
我遇到以下错误。任何帮助将不胜感激。
Tensor("dense_6_target:0", shape=(?, ?), dtype=float32)
0
1
None
None
2
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py in _apply_op_helper(self, op_type_name, name, **keywords)
510 as_ref=input_arg.is_ref,
--> 511 preferred_dtype=default_dtype)
512 except TypeError as err:
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py in internal_convert_to_tensor(value, dtype, name, as_ref, preferred_dtype, ctx, accept_symbolic_tensors)
1174 if ret is None:
-> 1175 ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
1176
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py in _TensorTensorConversionFunction(t, dtype, name, as_ref)
976 "Tensor conversion requested dtype %s for Tensor with dtype %s: %r" %
--> 977 (dtype.name, t.dtype.name, str(t)))
978 return t
ValueError: Tensor conversion requested dtype int32 for Tensor with dtype float32: 'Tensor("dense_6_sample_weights:0", shape=(?,), dtype=float32)'
During handling of the above exception, another exception occurred:
TypeError Traceback (most recent call last)
<ipython-input-10-511534c1dbfe> in <module>()
74
75 # Compiling the ANN
---> 76 classifier.compile(optimizer = 'adam', loss = newLoss(classifier), metrics = ['accuracy'])
77
78 # Fitting the ANN to the Training set
/usr/local/lib/python3.6/dist-packages/keras/engine/training.py in compile(self, optimizer, loss, metrics, loss_weights, sample_weight_mode, weighted_metrics, target_tensors, **kwargs)
340 with K.name_scope(self.output_names[i] + '_loss'):
341 output_loss = weighted_loss(y_true, y_pred,
--> 342 sample_weight, mask)
343 if len(self.outputs) > 1:
344 self.metrics_tensors.append(output_loss)
/usr/local/lib/python3.6/dist-packages/keras/engine/training_utils.py in weighted(y_true, y_pred, weights, mask)
419 score_array = K.mean(score_array,
420 axis=list(range(weight_ndim, ndim)))
--> 421 score_array *= weights
422 score_array /= K.mean(K.cast(K.not_equal(weights, 0), K.floatx()))
423 return K.mean(score_array)
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py in binary_op_wrapper(x, y)
810 with ops.name_scope(None, op_name, [x, y]) as name:
811 if isinstance(x, ops.Tensor) and isinstance(y, ops.Tensor):
--> 812 return func(x, y, name=name)
813 elif not isinstance(y, sparse_tensor.SparseTensor):
814 try:
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py in _mul_dispatch(x, y, name)
1076 is_tensor_y = isinstance(y, ops.Tensor)
1077 if is_tensor_y:
-> 1078 return gen_math_ops.mul(x, y, name=name)
1079 else:
1080 assert isinstance(y, sparse_tensor.SparseTensor) # Case: Dense * Sparse.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_math_ops.py in mul(x, y, name)
5858 # Add nodes to the TensorFlow graph.
5859 _, _, _op = _op_def_lib._apply_op_helper(
-> 5860 "Mul", x=x, y=y, name=name)
5861 _result = _op.outputs[:]
5862 _inputs_flat = _op.inputs
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py in _apply_op_helper(self, op_type_name, name, **keywords)
545 "%s type %s of argument '%s'." %
546 (prefix, dtypes.as_dtype(attrs[input_arg.type_attr]).name,
--> 547 inferred_from[input_arg.type_attr]))
548
549 types = [values.dtype]
TypeError: Input 'y' of 'Mul' Op has type float32 that does not match type int32 of argument 'x'.
我相信问题在于访问张量中的值，但我现在还不确定。