CN109255441A - Spacecraft fault diagnosis method based on artificial intelligence - Google Patents
Spacecraft fault diagnosis method based on artificial intelligence Download PDFInfo
- Publication number
- CN109255441A CN109255441A CN201811216219.XA CN201811216219A CN109255441A CN 109255441 A CN109255441 A CN 109255441A CN 201811216219 A CN201811216219 A CN 201811216219A CN 109255441 A CN109255441 A CN 109255441A
- Authority
- CN
- China
- Prior art keywords
- neural network
- output value
- training
- network
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000013473 artificial intelligence Methods 0.000 title claims description 7
- 238000013528 artificial neural network Methods 0.000 claims abstract description 78
- 238000012549 training Methods 0.000 claims abstract description 70
- 238000012360 testing method Methods 0.000 claims abstract description 38
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 210000002569 neuron Anatomy 0.000 claims description 29
- 230000007613 environmental effect Effects 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 238000004891 communication Methods 0.000 abstract description 2
- 230000003044 adaptive effect Effects 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 8
- 238000013178 mathematical model Methods 0.000 description 7
- 238000013507 mapping Methods 0.000 description 6
- 239000000872 buffer Substances 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of spacecraft fault diagnosis methods based on backpropagation BP neural network, mainly solve the problems, such as strong to priori knowledge dependence in the prior art.Using spacecraft environment data training backpropagation BP neural network, the probability data that Spacecraft anomaly event occurs is fitted the present invention.Its implementation are as follows: determine backpropagation BP neural network model；Obtain the training set data and test set data of backpropagation BP neural network；Initialize the parameter of backpropagation BP neural network；Backpropagation BP neural network is trained；Trained backpropagation BP neural network is optimized；Fault diagnosis is carried out to new spacecraft data using the BP neural network optimized.The present invention takes full advantage of the adaptive ability and learning ability of BP neural network, and the probability data that Spacecraft anomaly event can occur well is fitted, and reduces the dependence to priori knowledge, can be used for wireless communication field.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to a spacecraft fault diagnosis method which can be used in the field of wireless network communication.
Background
In recent years, with the continuous development of the aerospace field and the continuous breakthrough of the aerospace technology, the human aerospace activities are continuously increased, and the health status of the spacecraft is also widely valued by technicians and experts engaged in the design of the spacecraft. Due to the complexity of the space environment, the limitation of a spacecraft ground test system and the improvement of the complexity of the current spacecraft system, the problems of abnormal operation of the spacecraft and system faults are inevitable, the reliability of the spacecraft is correspondingly reduced, and serious loss is caused if the spacecraft system cannot be timely and accurately diagnosed and repaired. Therefore, whether the spacecraft can carry out intelligent fault diagnosis plays an important role in completing the whole space flight mission.
The existing spacecraft fault diagnosis methods mainly comprise three methods, namely a signal processing-based method, a mathematical model-based method and a knowledge-based method. Wherein:
the method based on signal processing is the earliest fault diagnosis technology and is the basis for fault diagnosis by other methods. The method can identify and detect the system fault by analyzing measurable signal characteristics such as time domain, amplitude, frequency domain and the like without taking a mathematical model of the system as a basis. The spacecraft fault diagnosis technology of the method often cannot achieve the expected effect due to poor real-time performance and autonomy.
The fault diagnosis method based on the mathematical model is the basis of the development of the modern fault diagnosis technology and is also a method which is developed most mature and applied most widely. The core of the method is that a residual error is generated by a plurality of methods such as parameter estimation, state estimation and the like based on an analysis system mathematical model, and then the residual error is analyzed and processed with the next fault by a threshold value or other limiting criteria. Although the method can accurately and efficiently finish the fault diagnosis of the spacecraft system, the establishment of an accurate mathematical model is very difficult for the spacecraft with a complex system structure and the unpredictable space environment, and even if the mathematical model is established, the mathematical model is difficult to ensure that the spacecraft system is not interfered by uncertain factors.
The fault diagnosis method based on knowledge acquires occurrence signs or judgment principles of fault diagnosis through a direct or indirect method, so that the fault occurrence condition of the system can be known more intuitively, and accurate judgment can be made in time to complete the fault diagnosis of the system. However, due to limited knowledge coverage, the uncertainty factor of the spacecraft system is more, and the lack of experience technology causes the method to have certain limitation.
The patent of Beijing spacecraft general design department in its application, "spacecraft on-orbit abnormity warning and fault diagnosis system" (patent inventor: Wang Huan Qin Wei, et al; patent application No.: CN201510608592.X, publication No.: 105159286B) provides a spacecraft on-orbit abnormity warning and fault diagnosis system, in which: inputting compiled alarm diagnosis knowledge by a knowledge editor; the data buffer area buffers the input original telemetering data; the data area extracts the telemetering data or telemetering instructions from the data buffer area when the inference controller carries out logic matching operation and stores the telemetering data or telemetering instructions; the rule area loads the compiled alarm diagnosis knowledge from the knowledge editor, and each alarm diagnosis knowledge in the rule area is called as a rule; the inference controller carries out logic matching operation on the original telemetering data in the data area and the alarm diagnosis knowledge in the rule area to obtain a diagnosis result; selecting a diagnosis result to be output and outputting the diagnosis result to a blackboard; the blackboard stores the diagnosis result obtained by the reasoning controller through logic matching operation; and the result buffer area buffers the diagnosis result selected and output by the reasoning controller and sends the diagnosis result to the client, and the user replies confirmation information after checking the diagnosis result through the client. The method has the following limitations: the dependency on the prior knowledge is strong, if the upper and lower limits of parameters, fault knowledge and a system model need to be predetermined, so that the method is insufficient in the aspects of self-adaption capability, learning capability, inaccurate information processing and the like.
Disclosure of Invention
The invention aims to provide a spacecraft fault diagnosis method based on a BP neural network aiming at the defects in the prior art, so as to reduce the dependence on prior knowledge, improve the self-learning and self-adaptive capabilities and effectively realize the intelligent diagnosis of spacecraft abnormity early warning and faults.
The technical scheme of the invention is realized as follows:
first, technical principle
In recent years, artificial intelligence technology has been widely used in various fields such as computer vision, speech recognition, natural language processing, medical automatic diagnosis, and finance. As one of the development trends of spacecraft fault diagnosis technology research, the fault diagnosis technology based on artificial intelligence is the key point of the field development, is the most widely applied and researched direction at the present stage, and is an effective method for solving the problems because an accurate model is not needed.
The back propagation BP neural network as an important tool in the field of artificial intelligence is a multi-layer feedforward network trained according to an error inverse propagation algorithm and proposed by a group of scientists including Rumelhart and McCelland in 1986. The back propagation BP network can learn and store a large number of input and output mapping relations without knowing a mathematical equation describing the mapping relations in advance, and has excellent nonlinear mapping capacity, generalization capacity and fault tolerance capacity.
The method utilizes spacecraft environment data to train the back propagation BP neural network, fits probability data of occurrence of spacecraft abnormal events, and uses the trained back propagation BP network to represent a complex nonlinear mapping relation between a spacecraft environment influence factor and a spacecraft abnormal mode.
Second, technical scheme
According to the principle, the technical scheme of the invention comprises the following steps:
(1) determining a back propagation BP neural network model:
defining the BP neural network structure as: the spacecraft anomaly detection system comprises an input layer, two hidden layers and an output layer, wherein preprocessed space environment data are used as input of a back propagation BP neural network, and the output of the back propagation BP neural network represents the probability of occurrence of certain abnormal events of a spacecraft;
(2) acquiring training set data and test set data of a back propagation BP neural network:
taking 70% of data in the space environment data set as a training sample set and the rest 30% of data as a test sample set, and normalizing the training sample set and the test sample set to obtain a normalized training sample set XrAnd test sample set Xt；
Extracting labels corresponding to the training sample setLabels corresponding to test sample setsWhereinIndicating the label value corresponding to the kth training sample,the label value corresponding to the nth test sample is represented, wherein k is 1, 2.
(3) Initializing parameters of the back propagation BP neural network:
weighting matrix W of BP neural network(l)And an offset vector b(l)Initializing the number of the neurons into random numbers between 0 and 1, and setting the number of the neurons of the input layer and the hidden layer as slThe number of neurons in an output layer is 1, wherein L is 1,2, and L-1, and L represents the total number of layers of the BP neural network;
(4) training a back propagation BP network:
(4a) the current iteration number of the back propagation BP network training is represented by M, the target error value of the network training is represented by Acc, and the error value of the network training is represented by err, wherein M is 1,2,3max，MmaxSetting the maximum iteration times of the training BP network;
(4b) training set X after normalizationrRandomly selecting a sample, inputting the sample into a back propagation BP neural network, and calculating the actual output value y of the BP neural network corresponding to the single sample in a training set(k)；
(4c) Adopting a back propagation algorithm based on a gradient descent method, and utilizing the actual output value y of the network corresponding to the kth sample in the training set(k)And expected output valueCarrying out fine adjustment on the BP network to realize the update of network parameters;
(4d) after the parameters are updated, a minimum mean square error evaluation method is adopted, and the actual output values Y of the networks corresponding to all the training samples are utilizedrp＝{y(1),y(2),...,y(k),...,y(p)And the desired output value YrCalculate BP spiritGlobal training error Acc over networkrAnd determining Accr<If err is true, Acc will be determined if err is truerAssigning err, and executing (4e), otherwise, executing (4e) directly;
(4e) repeating (4b) to (4d) with M ═ M +1, and making M ═ MmaxThe BP neural network corresponding to the chronorr is used as a trained BP neural network;
(5) optimizing the trained back propagation BP neural network:
test set XtInputting the actual output value Y of the network corresponding to the test set into the trained BP neural networkp＝{d(1),d(2),...,d(n),...,d(q)}；
The actual output value Y through the network is evaluated by the minimum mean square errorpAnd the desired output value YtCalculating the error Acc of the trained BP neural networkpWherein d is(n)Representing the actual network output value corresponding to the nth test sample, wherein n is 1,2p>Whether Acc is established or not, if yes, M is adjusted, the step returns to the step (4), and if not, the step (6) is executed;
(6) and inputting the new normalized space environment data set V into the trained BP neural network to obtain a fault diagnosis result Y, namely the probability of occurrence of certain abnormal events of the spacecraft.
Compared with the prior art, the invention has the following advantages:
the invention fully utilizes the self-adaptive capacity and the learning capacity of the back propagation BP neural network, can learn and store a large number of input and output mapping relations through the BP neural network, does not need to know a mathematical equation describing the mapping relation in advance, and reduces the dependence on prior knowledge;
simulation results show that the method can well fit probability data of spacecraft abnormal events, can timely find abnormal conditions from a large amount of historical environmental data, and intelligently and effectively perform early warning and fault diagnosis on spacecraft abnormity.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a schematic diagram of the structure of a BP neural network employed in the present invention;
fig. 3 is a diagram of simulation results for spacecraft fault diagnosis using the present invention.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
Referring to fig. 1, the implementation steps of the invention are as follows:
step 1, determining a back propagation BP neural network model.
Defining a back propagation BP neural network structure comprises: the input layer, the two hidden layers and the output layer are shown in fig. 2, the input of the back propagation BP neural network is preprocessed space environment data, and the output of the back propagation BP neural network represents the probability of occurrence of some abnormal event of the spacecraft.
And 2, acquiring training set data and test set data of the back propagation BP neural network.
Taking 70% of data in the space environment data set as a training sample set and the rest 30% of data as a test sample set, and normalizing the training sample set and the test sample set to obtain a normalized training sample set XrAnd test sample set Xt；
Extracting labels corresponding to the training sample setLabels corresponding to test sample setsWhereinIndicating the label value corresponding to the kth training sample,the label value corresponding to the nth test sample is represented, wherein k is 1, 2.
The data set in this embodiment is 1080 pieces of space environment data acquired in a certain space environment.
And 3, initializing parameters of the back propagation BP neural network.
Weighting matrix W of BP neural network(l)And an offset vector b(l)Initializing the number of the neurons into random numbers between 0 and 1, and setting the number of the neurons of the input layer and the hidden layer as slThe number of neurons in an output layer is 1, wherein L is 1,2, and L-1, and L represents the total number of layers of the BP neural network; in this example, take s1＝144，s2＝s3＝64。
And 4, training the back propagation BP network.
4a) Let M be the current iteration number of the back propagation BP network training, where M is 1,2,3max，MmaxSetting the maximum iteration times of the training BP network; setting a target error value of network training as Acc and setting an error value of network training as err; in this example, take Mmax＝5000，err＝1.0，Acc＝0.0005；
4b) Preprocessing the training set by adopting a maximum and minimum normalization method to obtain a normalized training set Xr；
4c) In normalized training set XrRandomly selecting a sample, inputting the sample into a back propagation BP neural network, and calculating and trainingCentralizing the actual output value y of BP neural network corresponding to single sample(k)；
4c1) Random selection of normalized training set XrInputting the kth sample into BP neural network to obtain weighted sum of each neuronAnd output value of each neuron
Wherein:
slindicating the number of neurons in the L-th layer, wherein L is 1,2, L-1, and L indicates the total number of layers of the BP neural network;
represents the weighted sum of the input of the r-th neuron at the l + 1-th layer;
represents an output value of an r-th neuron of the l + 1-th layer;
representing the connection weight between the ith neuron of the l layer and the t neuron of the l +1 layer;
representing bias terms corresponding to the t-th neuron of the l +1 th layer;
f (-) represents the activation function used to calculate the output value of each neuron.
4c2) Using the weighted sum of each neuron obtained in step 4c1)And an output valueCalculating the actual output value y of the BP neural network corresponding to the kth sample in the training set(k)：
Wherein,and the output value of the neuron element of the L-th layer is represented, namely the actual output value of the network corresponding to the k-th sample.
4d) Adopting a back propagation algorithm based on a gradient descent method, and utilizing the actual output value y of the network corresponding to the kth sample in the training set(k)And expected output valueCarrying out fine adjustment on the BP network to realize the update of network parameters;
4d1) calculating an error function J corresponding to the kth sample:
wherein, y(k)Representing the actual output value of the network corresponding to the kth environmental sample data;a tag representing kth environmental sample data;
4d2) weighting matrix W of BP neural network by using error function J of kth sample(l)And an offset vector b(l)The two parameters are updated to obtain an updated weight matrixAnd an offset vectorThe parameters are as follows:
wherein α is the learning rate;
represents a weight matrix corresponding to the l +1 th layer, where r is 1,2l,t＝1,2,...,sl+1；
Denotes the offset vector of the l +1 th layer, where t 1,2l+1。
4e) After the parameters are updated, a minimum mean square error evaluation method is adopted, and the actual output values Y of the networks corresponding to all the training samples are utilizedrp＝{y(1),y(2),...,y(k),...,y(p)And the desired output value YrCalculating the global training error Acc of the BP neural networkrAnd determining Accr<If err is true, Acc will be determined if err is truerAssigning err, executing step 4f), otherwise, directly executing step 4f), wherein AccrCalculated by the following formula:
wherein AccrRepresenting the average value of the square of the error between the actual output value and the expected output value of the BP neural network corresponding to all the training samples;
y(k)representing an actual output value obtained by inputting the kth environmental sample data in the training set into a network;
a label representing the kth environmental sample data corresponding to the training set;
p represents the number of all training samples.
4f) Repeating steps 4c) to 4e) by making M ═ M +1, and making M ═ MmaxAnd the BP neural network corresponding to the chronorr is used as the trained BP neural network.
And 5, optimizing the trained BP neural network.
5a) Test set XtInputting the actual output value Y of the network corresponding to the test set into the trained BP neural networkp＝{d(1),d(2),...,d(n),...,d(q)In which d is(n)Representing the actual network output value corresponding to the nth test sample, wherein n is 1, 2.
5b) The actual output value Y through the network is evaluated by the minimum mean square errorpAnd the desired output value YtCalculating the error Acc of the trained BP neural networkp：
Wherein d is(n)Representing an actual output value obtained by inputting nth environmental sample data into a network in the test set;a label representing nth environment sample data corresponding to the test set; q represents the number of all test samples.
5c) Determine Accp>And (6) if Acc is established, if so, adjusting M, returning to the step 4, otherwise, obtaining the optimized BP neural network, and executing the step 6.
And 6, carrying out fault diagnosis on the new spacecraft data.
Inputting the new normalized space environment data set V into the BP neural network optimized in the step 5) to obtain a fault diagnosis result Y, namely the probability of occurrence of certain abnormal events of the spacecraft.
The technical effects of the invention are further explained by simulation experiments as follows:
1. simulation conditions are as follows:
the simulation of the fault diagnosis of the aerospace environment data set is realized by programming a tensierflow frame in python3.6, 1080 aerospace environment sample data are selected from the acquired aerospace environment data set, 70% of the acquired aerospace environment sample data are randomly selected as a training sample set, and the rest 30% of the acquired aerospace environment sample data are selected as a testing sample set.
2. Emulated content
The method of the invention is used for simulating spacecraft fault diagnosis on the test sample set to obtain a fitting effect graph of an actual output value and an expected output value of the BP neural network, and the result is shown in figure 3. Wherein the horizontal axis represents a test sample sequence, and the vertical axis represents the actual output value and the expected output value of the network of the spacecraft fault diagnosis method, as can be seen from fig. 3, the spacecraft fault diagnosis method can well fit spacecraft environmental data.
Claims (5)
1. A spacecraft fault diagnosis method based on artificial intelligence is characterized by comprising the following steps:
(1) determining a back propagation BP neural network model:
defining a back propagation BP neural network structure comprises: the input of the back propagation BP neural network is preprocessed space environment data, and the output of the back propagation BP neural network represents the probability of occurrence of certain abnormal events of the spacecraft;
(2) acquiring training set data and test set data of a back propagation BP neural network:
taking 70% of data in the space environment data set as a training sample set and the rest 30% of data as a test sample set, and normalizing the training sample set and the test sample set to obtain a normalized training sample set XrAnd test sample set Xt；
Extracting labels corresponding to the training sample setLabels corresponding to test sample setsWhereinIndicating the label value corresponding to the kth training sample,the label value corresponding to the nth test sample is represented, wherein k is 1, 2.
(3) Initializing parameters of the back propagation BP neural network:
weighting matrix W of BP neural network(l)And an offset vector b(l)Initializing the number of the neurons into random numbers between 0 and 1, and setting the number of the neurons of the input layer and the hidden layer as slThe number of neurons in an output layer is 1, wherein L is 1,2, and L-1, and L represents the total number of layers of the BP neural network;
(4) training a back propagation BP network:
(4a) let M be the current iteration number of the back propagation BP network training, where M is 1,2,3max，MmaxSetting the maximum iteration times of the training BP network; setting a target error value of network training as Acc and setting an error value of network training as err; in this example, take Mmax＝5000，err＝1.0，Acc＝0.0005；
(4b) Preprocessing the training set by adopting a maximum and minimum normalization method to obtainNormalized training set Xr；
(4c) In normalized training set XrRandomly selecting a sample, inputting the sample into a back propagation BP neural network, and calculating the actual output value y of the BP neural network corresponding to the single sample in a training set(k)；
(4d) Adopting a back propagation algorithm based on a gradient descent method, and utilizing the actual output value y of the network corresponding to the kth sample in the training set(k)And expected output valueCarrying out fine adjustment on the BP network to realize the update of network parameters;
(4e) after the parameters are updated, a minimum mean square error evaluation method is adopted, and the actual output values Y of the networks corresponding to all the training samples are utilizedrp＝{y(1),y(2),...,y(k),...,y(p)And the desired output value YrCalculating the global training error Acc of the BP neural networkrAnd determining Accr<If err is true, Acc will be determined if err is truerAssigning err, and executing (4f), otherwise, executing (4f) directly;
(4f) repeating (4c) to (4e) with M ═ M +1, and making M ═ MmaxThe BP neural network corresponding to the chronorr is used as a trained BP neural network;
(5) optimizing the trained back propagation BP neural network:
(5a) test set XtInputting the actual output value Y of the network corresponding to the test set into the trained BP neural networkp＝{d(1),d(2),...,d(n),...,d(q)In which d is(n)Representing the actual network output value corresponding to the nth test sample, wherein n is 1, 2.
(5b) The actual output value Y through the network is evaluated by the minimum mean square errorpAnd the desired output value YtCalculating the error Acc of the trained BP neural networkp；
(5c) Determine Accp>If Acc is true, adjusting M and returning to (4) if Acc is true, otherwise, optimizingThe BP neural network of (6);
(6) and inputting the new normalized space environment data set V into the trained BP neural network to obtain a fault diagnosis result Y, namely the probability of occurrence of certain abnormal events of the spacecraft.
2. The method of claim 1, wherein (4c) the actual output value y of the BP neural network corresponding to a single sample in the training set is calculated(k)It is implemented as follows:
(4c1) random selection of training set XrInputting the kth sample into BP neural network to obtain weighted sum of each neuronAnd output value of each neuron
Wherein:
slindicating the number of neurons in the L-th layer, wherein L is 1,2, L-1, and L indicates the total number of layers of the BP neural network;
represents the weighted sum of the input of the r-th neuron at the l + 1-th layer;
represents an output value of an r-th neuron of the l + 1-th layer;
representing the connection weight between the ith neuron of the l layer and the t neuron of the l +1 layer;
representing bias terms corresponding to the t-th neuron of the l +1 th layer;
f (-) represents the activation function used to calculate the output value of each neuron.
(4c2) Using the weighted sum of each neuron obtained in (4c1)And an output valueCalculating the actual output value y of the BP neural network corresponding to the kth sample in the training set(k)：
Wherein,and the output value of the neuron element of the L-th layer is represented, namely the actual output value of the network corresponding to the k-th sample.
3. The method of claim 1, wherein the actual output value y of the net corresponding to the kth sample is utilized in (4d)(k)And expected output valueThe BP network is finely adjusted, and the realization method comprises the following steps:
(4d1) calculating an error function J corresponding to the kth sample:
wherein, y(k)Representing the actual output value of the network corresponding to the kth environmental sample data;a tag representing kth environmental sample data;
(4d2) weighting matrix W of BP neural network by using error function J of kth sample(l)And an offset vector b(l)The two parameters are updated to obtain an updated weight matrixAnd an offset vectorThe parameters are as follows:
wherein α is the learning rate;
represents a weight matrix corresponding to the l +1 th layer, where r is 1,2l,t＝1,2,...,sl+1；
Denotes the offset vector of the l +1 th layer, where t 1,2l+1。
4. The method according to claim 1 or 3, wherein (4e) is usedActual output value Y of network corresponding to all training samplesrp＝{y(1),y(2),...,y(k),...,y(p)And the desired output value YrCalculating the global training error Acc of the BP neural networkrCalculated by the following formula:
wherein AccrRepresenting the average value of the square of the error between the actual output value and the expected output value of the BP neural network corresponding to all the training samples;
y(k)representing an actual output value obtained by inputting the kth environmental sample data in the training set into a network;
a label representing the kth environmental sample data corresponding to the training set;
p represents the number of all training samples.
5. A method according to claim 1 or 3, wherein the actual output value Y through the network in (5b) isp＝{d(1),d(2),...,d(n),...,d(q)And the desired output value YtCalculating the error Acc of the trained BP neural networkpCalculated by the following formula:
wherein AccpRepresenting the average value of the square of the error between the actual output value and the expected output value of the BP neural network corresponding to all the training samples;
d(n)representing an actual output value obtained by inputting nth environmental sample data into a network in the test set;
a label representing nth environment sample data corresponding to the test set;
q represents the number of all test samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811216219.XA CN109255441A (en) | 2018-10-18 | 2018-10-18 | Spacecraft fault diagnosis method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811216219.XA CN109255441A (en) | 2018-10-18 | 2018-10-18 | Spacecraft fault diagnosis method based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109255441A true CN109255441A (en) | 2019-01-22 |
Family
ID=65045696
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811216219.XA Pending CN109255441A (en) | 2018-10-18 | 2018-10-18 | Spacecraft fault diagnosis method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109255441A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109951357A (en) * | 2019-03-18 | 2019-06-28 | 西安电子科技大学 | Network application recognition methods based on multilayer neural network |
| CN110163389A (en) * | 2019-05-27 | 2019-08-23 | 上海电气泰雷兹交通自动化系统有限公司 | Apply to the Neural Networks Learning Algorithm in predictive maintenance management |
| CN110333692A (en) * | 2019-07-04 | 2019-10-15 | 南京农业大学 | The automatic monitoring diagnosis system of pig fever based on thermal infrared |
| CN110942401A (en) * | 2019-11-21 | 2020-03-31 | 黑龙江电力调度实业有限公司 | Intelligent communication method for power Internet of things |
| CN111026087A (en) * | 2019-12-20 | 2020-04-17 | 中国船舶重工集团公司第七一九研究所 | Weight-containing nonlinear industrial system fault detection method and device based on data |
| CN112067289A (en) * | 2020-08-21 | 2020-12-11 | 天津电气科学研究院有限公司 | Motor shaft and transmission shaft abnormal vibration early warning algorithm based on neural network |
| CN112287452A (en) * | 2020-10-12 | 2021-01-29 | 哈尔滨工业大学 | Intelligent modeling method for maintainability of spacecraft |
| CN113255546A (en) * | 2021-06-03 | 2021-08-13 | 成都卡莱博尔信息技术股份有限公司 | Diagnosis method for aircraft system sensor fault |
| CN113421176A (en) * | 2021-07-16 | 2021-09-21 | 昆明学院 | Intelligent abnormal data screening method |
| CN113673143A (en) * | 2020-05-15 | 2021-11-19 | 中国石油化工股份有限公司 | Memory, biochemical oxygen demand soft measuring method, system and device |
| CN113722902A (en) * | 2021-08-23 | 2021-11-30 | 西安电子科技大学 | Neural network-based optimal fit parameter estimation method for antenna of shaped reflector |
| CN113867307A (en) * | 2021-08-20 | 2021-12-31 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep neural network |
| CN114461629A (en) * | 2022-02-10 | 2022-05-10 | 电子科技大学 | Temperature calibration method and device for aircraft engine and storage medium |
| CN115081625A (en) * | 2022-07-21 | 2022-09-20 | 常安集团有限公司 | Intelligent control method and system for miniature circuit breaker |
| CN116305531A (en) * | 2023-01-13 | 2023-06-23 | 北京控制工程研究所 | Spacecraft health evolution model modeling method, device, equipment and medium |
| CN116502516A (en) * | 2023-01-15 | 2023-07-28 | 北京控制工程研究所 | Identification method and device for degradation stage of spacecraft component |
| CN118070003A (en) * | 2024-04-19 | 2024-05-24 | 中国西安卫星测控中心 | Spacecraft telemetry data interpolation method based on neural network |
-
2018
- 2018-10-18 CN CN201811216219.XA patent/CN109255441A/en active Pending
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109951357A (en) * | 2019-03-18 | 2019-06-28 | 西安电子科技大学 | Network application recognition methods based on multilayer neural network |
| CN110163389A (en) * | 2019-05-27 | 2019-08-23 | 上海电气泰雷兹交通自动化系统有限公司 | Apply to the Neural Networks Learning Algorithm in predictive maintenance management |
| CN110333692A (en) * | 2019-07-04 | 2019-10-15 | 南京农业大学 | The automatic monitoring diagnosis system of pig fever based on thermal infrared |
| CN110942401A (en) * | 2019-11-21 | 2020-03-31 | 黑龙江电力调度实业有限公司 | Intelligent communication method for power Internet of things |
| CN110942401B (en) * | 2019-11-21 | 2023-12-19 | 黑龙江电力调度实业有限公司 | Intelligent communication method for electric power Internet of things |
| CN111026087A (en) * | 2019-12-20 | 2020-04-17 | 中国船舶重工集团公司第七一九研究所 | Weight-containing nonlinear industrial system fault detection method and device based on data |
| CN113673143A (en) * | 2020-05-15 | 2021-11-19 | 中国石油化工股份有限公司 | Memory, biochemical oxygen demand soft measuring method, system and device |
| CN112067289A (en) * | 2020-08-21 | 2020-12-11 | 天津电气科学研究院有限公司 | Motor shaft and transmission shaft abnormal vibration early warning algorithm based on neural network |
| CN112287452A (en) * | 2020-10-12 | 2021-01-29 | 哈尔滨工业大学 | Intelligent modeling method for maintainability of spacecraft |
| CN113255546B (en) * | 2021-06-03 | 2021-11-09 | 成都卡莱博尔信息技术股份有限公司 | Diagnosis method for aircraft system sensor fault |
| CN113255546A (en) * | 2021-06-03 | 2021-08-13 | 成都卡莱博尔信息技术股份有限公司 | Diagnosis method for aircraft system sensor fault |
| CN113421176A (en) * | 2021-07-16 | 2021-09-21 | 昆明学院 | Intelligent abnormal data screening method |
| CN113421176B (en) * | 2021-07-16 | 2022-11-01 | 昆明学院 | Intelligent screening method for abnormal data in student score scores |
| CN113867307A (en) * | 2021-08-20 | 2021-12-31 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep neural network |
| CN113867307B (en) * | 2021-08-20 | 2022-10-21 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep neural network |
| CN113722902A (en) * | 2021-08-23 | 2021-11-30 | 西安电子科技大学 | Neural network-based optimal fit parameter estimation method for antenna of shaped reflector |
| CN114461629A (en) * | 2022-02-10 | 2022-05-10 | 电子科技大学 | Temperature calibration method and device for aircraft engine and storage medium |
| CN115081625A (en) * | 2022-07-21 | 2022-09-20 | 常安集团有限公司 | Intelligent control method and system for miniature circuit breaker |
| CN116305531B (en) * | 2023-01-13 | 2023-09-15 | 北京控制工程研究所 | Spacecraft health evolution model modeling method, device, equipment and medium |
| CN116305531A (en) * | 2023-01-13 | 2023-06-23 | 北京控制工程研究所 | Spacecraft health evolution model modeling method, device, equipment and medium |
| CN116502516A (en) * | 2023-01-15 | 2023-07-28 | 北京控制工程研究所 | Identification method and device for degradation stage of spacecraft component |
| CN116502516B (en) * | 2023-01-15 | 2023-12-01 | 北京控制工程研究所 | Identification method and device for degradation stage of spacecraft component |
| CN118070003A (en) * | 2024-04-19 | 2024-05-24 | 中国西安卫星测控中心 | Spacecraft telemetry data interpolation method based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109255441A (en) | Spacecraft fault diagnosis method based on artificial intelligence | |
| CN111539515B (en) | Complex equipment maintenance decision method based on fault prediction | |
| CN111428789A (en) | Network traffic anomaly detection method based on deep learning | |
| US8306931B1 (en) | Detecting, classifying, and tracking abnormal data in a data stream | |
| Zhang et al. | Data-core-based fuzzy min–max neural network for pattern classification | |
| CN109141847B (en) | Aircraft system fault diagnosis method based on MSCNN deep learning | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN107544904B (en) | Software reliability prediction method based on deep CG-LSTM neural network | |
| CN112131212A (en) | Hybrid cloud scene-oriented time sequence data anomaly prediction method based on ensemble learning technology | |
| CN109800875A (en) | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine | |
| CN110702418A (en) | Aircraft engine fault prediction method | |
| CN105095918A (en) | Multi-robot system fault diagnosis method | |
| CN113923104B (en) | Network fault diagnosis method, equipment and storage medium based on wavelet neural network | |
| CN110110809B (en) | Fuzzy automaton construction method based on machine fault diagnosis | |
| CN106656357B (en) | Power frequency communication channel state evaluation system and method | |
| CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
| CN114580288A (en) | Intelligent fault diagnosis method based on DSECJAN | |
| CN112560948A (en) | Eye fundus map classification method and imaging method under data deviation | |
| Xing-yu et al. | RNN-based method for fault diagnosis of grinding system | |
| CN116415505A (en) | System fault diagnosis and state prediction method based on SBR-DBN model | |
| CN114417242B (en) | Big data detection system for livestock and poultry activity information | |
| CN114154612A (en) | Intelligent agent behavior model construction method based on causal relationship inference | |
| Dang et al. | seq2graph: Discovering dynamic non-linear dependencies from multivariate time series | |
| Zhao et al. | Software reliability growth model based on fuzzy wavelet neural network | |
| CN112560252A (en) | Prediction method for residual life of aircraft engine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190122 |
|
| WD01 | Invention patent application deemed withdrawn after publication |