CN106301610A - The adaptive failure detection of a kind of superhet and diagnostic method and device - Google Patents
The adaptive failure detection of a kind of superhet and diagnostic method and device Download PDFInfo
- Publication number
- CN106301610A CN106301610A CN201610757171.8A CN201610757171A CN106301610A CN 106301610 A CN106301610 A CN 106301610A CN 201610757171 A CN201610757171 A CN 201610757171A CN 106301610 A CN106301610 A CN 106301610A
- Authority
- CN
- China
- Prior art keywords
- superheterodyne receiver
- grnn
- residual error
- fault
- output signal
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 97
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 58
- 238000002405 diagnostic procedure Methods 0.000 title abstract 2
- 238000000034 method Methods 0.000 claims abstract description 47
- 238000013528 artificial neural network Methods 0.000 claims description 46
- 238000012549 training Methods 0.000 claims description 45
- 238000003745 diagnosis Methods 0.000 claims description 30
- 238000012545 processing Methods 0.000 claims description 28
- 238000000605 extraction Methods 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000007257 malfunction Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 9
- 238000004088 simulation Methods 0.000 description 9
- 239000013598 vector Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 210000002569 neuron Anatomy 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 230000009977 dual effect Effects 0.000 description 5
- 238000002347 injection Methods 0.000 description 5
- 239000007924 injection Substances 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002955 isolation Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 230000010355 oscillation Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000002194 synthesizing effect Effects 0.000 description 2
- 241000219322 Dianthus Species 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/20—Monitoring; Testing of receivers
- H04B17/26—Monitoring; Testing of receivers using historical data, averaging values or statistics
Landscapes
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Electromagnetism (AREA)
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention discloses adaptive failure detection and diagnostic method and the device of a kind of superhet, method includes: inputs input signal to superhet, and is processed described input signal by described superhet；Utilize the one-level GRNN observer trained, the output signal of described input signal and described superhet previous moment is processed, obtains estimated output signal；According to described estimated output signal and the output signal of described superhet current time, obtain residual error；Utilize the two grades of GRNN observers trained, described input signal and described residual error are processed, obtains adaptive threshold；According to adaptive threshold described in described residual sum, determine whether described superhet breaks down.The present invention can detect whether superhet breaks down, and can be diagnosed to be fault type.
Description
Technical Field
The invention relates to the field of fault detection of a superheterodyne receiver, in particular to a self-adaptive fault detection and diagnosis method and device of a superheterodyne receiver.
Background
The superheterodyne receiver is widely applied to an electronic information system as equipment for receiving information by electronic equipment, and is small as a radio and large as a radar receiver, and the superheterodyne receiver plays a key role in corresponding equipment. However, as an important component of national defense and military construction, electronic information systems are in complex electromagnetic environments due to the devices themselves, and therefore, the direct and obvious benefit of researching fault detection of a typical device, namely a superheterodyne receiver, under the action of the complex electromagnetic environment is achieved.
The superheterodyne receiver is a typical module in an electronic information product, is commonly used in remote communication, and has the characteristics of high frequency resolution, high sensitivity, wide dynamic range, large frequency measurement range and the like. The superheterodyne receiver has relatively simple structure and high reliability, and is a necessary frequency measurement receiver in electronic information investigation. In the using process of military electronic information products, once the superheterodyne receiver fails, the communication of information is greatly influenced, so that the method has important significance for fault detection and diagnosis of the superheterodyne receiver. However, there is currently little research related to superheterodyne receiver fault detection and diagnosis.
Therefore, the invention researches the superheterodyne receiver and provides method support for fault detection and diagnosis of the superheterodyne receiver.
Disclosure of Invention
The technical problem solved according to the technical scheme provided by the embodiment of the invention is how to detect the fault of the superheterodyne receiver and diagnose the fault reason.
The self-adaptive fault detection and diagnosis method for the superheterodyne receiver provided by the embodiment of the invention comprises the following steps:
inputting an input signal to a superheterodyne receiver, and processing the input signal by the superheterodyne receiver;
processing the input signal and an output signal of the superheterodyne receiver at the previous moment by using a trained primary generalized regression neural network GRNN to obtain an estimated output signal;
obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current moment;
processing the input signal and the residual error for fault detection by using the trained secondary GRNN to obtain a self-adaptive threshold;
and determining whether the superheterodyne receiver fails according to the residual error for fault detection and the adaptive threshold.
Preferably, the step of obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current time includes:
and subtracting the estimated output signal output by the primary GRNN with the output signal of the superheterodyne receiver at the current moment to obtain the residual error of the superheterodyne receiver at the moment.
Preferably, the step of determining whether the superheterodyne receiver fails according to the residual error for failure detection and the adaptive threshold includes:
and comparing the residual error for fault detection with the adaptive threshold, and determining that the superheterodyne receiver has a fault if the residual error for fault detection is greater than the adaptive threshold.
Preferably, the trained one-stage GRNN is obtained by the following steps:
inputting an input signal to a superheterodyne receiver in a normal working state, and processing the input signal by the superheterodyne receiver in the normal working state;
and taking the input signal and the output signal of the superheterodyne receiver in the normal working state at the previous moment as the input of the primary GRNN, taking the output signal of the superheterodyne receiver in the normal working state at the current moment as the output of the primary GRNN, and training the primary GRNN to obtain the trained primary GRNN.
Preferably, the trained two-stage GRNN is obtained by:
obtaining a residual error for training a secondary GRNN according to the estimated output signal of the trained primary GRNN and the output signal of the superheterodyne receiver in the normal working state at the current moment;
and taking the input signal and the residual error for training the secondary GRNN as the input of the secondary GRNN, taking an expected threshold value as the output of the secondary GRNN, and training the secondary GRNN to obtain the trained secondary GRNN.
Preferably, the method further comprises the step of diagnosing a fault of the superheterodyne receiver.
Preferably, the step of diagnosing the fault of the superheterodyne receiver includes:
extracting the characteristics of the residual error for fault detection to obtain the time domain and frequency domain characteristics of the residual error for fault detection;
and processing the time domain and frequency domain characteristics of the residual error for fault detection by using the trained probabilistic neural network PNN, and determining the fault type of the superheterodyne receiver.
Preferably, the trained PNN is obtained by:
performing characteristic extraction on the residual error of the superheterodyne receiver with the known fault type to obtain time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type;
and training the PNN for fault classification by using the time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type to obtain the PNN after training.
According to a storage medium provided by an embodiment of the present invention, it stores a program for implementing the above-described adaptive fault detection and diagnosis method for a superheterodyne receiver.
The self-adaptive fault detection and diagnosis device for the superheterodyne receiver provided by the embodiment of the invention comprises the following components:
the superheterodyne receiver is used for processing an input signal;
the primary GRNN observer is used for processing the input signal and an output signal of the superheterodyne receiver at the previous moment by utilizing a trained primary generalized regression neural network GRNN to obtain an estimated output signal;
the calculation module is used for obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current moment;
the secondary GRNN observer is used for processing the input signal and the residual error for fault detection by utilizing the trained secondary GRNN to obtain a self-adaptive threshold;
and the judging module is used for determining whether the superheterodyne receiver has a fault according to the residual error for fault detection and the adaptive threshold.
Preferably, the apparatus further comprises:
the characteristic extraction module is used for extracting the characteristics of the residual error for fault detection to obtain the time domain and frequency domain characteristics of the residual error for fault detection;
and the PNN classifier is used for processing the time domain and frequency domain characteristics of the residual error for fault detection by using the trained probabilistic neural network PNN to determine the fault type of the superheterodyne receiver.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
the embodiment of the invention can judge whether the superheterodyne receiver has a fault through the self-adaptive threshold and the residual error generated by the GRNN-based double-stage observer, and diagnose the fault reason of the superheterodyne receiver by using the PNN, thereby filling the blank of fault detection and diagnosis of the superheterodyne receiver.
Drawings
Fig. 1 is a schematic block diagram of an adaptive fault detection and diagnosis method for a superheterodyne receiver according to an embodiment of the present invention;
fig. 2 is a block diagram of an adaptive fault detection and diagnosis apparatus of a superheterodyne receiver according to an embodiment of the present invention;
FIG. 3 is a flow chart of adaptive fault detection and diagnosis for a superheterodyne receiver based on observer and residual analysis according to an embodiment of the present invention;
fig. 4 is a diagram of a GRNN neural network architecture provided by an embodiment of the present invention;
FIG. 5 is a diagram of the establishment of a fault observer provided by an embodiment of the present invention;
FIG. 6 is a probabilistic neural network model provided by an embodiment of the present invention;
FIG. 7 is a superheterodyne receiver simulation model provided by an embodiment of the present invention;
FIG. 8 illustrates a neural network residual and adaptive thresholds trained in a normal state according to an embodiment of the present invention;
FIG. 9 shows the detection result of the electronic amplifier in failure according to the embodiment of the present invention;
fig. 10 shows a detection result when the if filter provided by the embodiment of the present invention is faulty;
fig. 11 is a detection result when a local oscillation fails according to the embodiment of the present invention;
fig. 12 is a schematic diagram of a predicted classification result of the PNN classifier according to the embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described below are only for the purpose of illustrating and explaining the present invention, and are not to be construed as limiting the present invention.
Fig. 1 is a schematic block diagram of a method for adaptive fault detection and diagnosis of a superheterodyne receiver according to an embodiment of the present invention, as shown in fig. 1, the method includes the steps of:
step S101: an input signal is input to a superheterodyne receiver, and the input signal is processed by the superheterodyne receiver. Wherein the input signal may be a sinusoidal signal.
Step S102: and processing the input signal and an output signal of the superheterodyne receiver at the previous moment by using the trained primary GRNN to obtain an estimated output signal.
Step S103: and obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current moment.
And obtaining a residual error of the superheterodyne receiver at the moment by subtracting the estimated output signal output by the primary GRNN from the output signal of the superheterodyne receiver at the moment.
Step S104: and processing the input signal and the residual error for fault detection by using the trained secondary GRNN to obtain a self-adaptive threshold.
Step S105: and determining whether the superheterodyne receiver fails according to the residual error for fault detection and the adaptive threshold.
And comparing the residual error for fault detection with the adaptive threshold, if the residual error for fault detection is greater than the adaptive threshold, determining that the superheterodyne receiver has a fault, otherwise, indicating that the superheterodyne receiver is normal.
When it is determined in step S105 that the superheterodyne receiver has a fault, the method further includes a step of diagnosing the fault of the superheterodyne receiver, that is, a step of detecting a fault type, which specifically includes:
step S106: and (4) performing feature extraction on the residual error for fault detection obtained in the step (S103) to obtain time domain and frequency domain features of the residual error for fault detection.
Step S107: inputting the time domain and frequency domain characteristics of the residual error for fault detection to the PNN after training, and processing the time domain and frequency domain characteristics of the residual error for fault detection by using the PNN after training to determine the fault type of the superheterodyne receiver.
The first-stage GRNN after training is obtained through the following steps: firstly, inputting an input signal to a superheterodyne receiver in a normal working state, processing the input signal by the superheterodyne receiver in the normal working state, then taking the input signal and an output signal of the superheterodyne receiver in the normal working state at the previous moment as the input of a primary GRNN, taking an output signal of the superheterodyne receiver in the normal working state at the current moment as the output of the primary GRNN, and training the primary GRNN to obtain the trained primary GRNN.
The trained two-stage GRNN is obtained through the following steps: firstly, obtaining a residual error for training a secondary GRNN according to the estimated output signal of the trained primary GRNN and the output signal of the superheterodyne receiver in the normal working state, then taking the input signal and the residual error for training the secondary GRNN as the input of the secondary GRNN, taking an expected threshold value as the output of the secondary GRNN, and training the secondary GRNN to obtain the trained secondary GRNN.
The PNN after training is obtained through the following steps: firstly, extracting the characteristics of the residual error of the superheterodyne receiver with the known fault type to obtain the time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type, and then training the PNN for fault classification by using the time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type to obtain the trained PNN.
To sum up, the embodiment of the present invention first performs each neural network training, specifically including: training the primary GRNN by using an input signal and output signals of the superheterodyne receiver in a normal working state at the previous moment and the current moment to obtain a trained primary GRNN; training the secondary GRNN by using the input signal and the residual signal of the superheterodyne receiver in a normal working state to obtain a trained secondary GRNN; and training the PNN by utilizing the residual error of the superheterodyne receiver with the known fault type to obtain the PNN after training. And then, carrying out fault detection by using the trained primary GRNN and secondary GRNN, wherein the fault detection specifically comprises the following steps: processing an input signal, an output signal of a superheterodyne receiver in an unknown state at the previous moment and the current moment by using the trained primary GRNN to obtain an estimated output signal at the current moment; and processing the input signal and a residual signal (namely a residual for fault detection) of the superheterodyne receiver in an unknown state by using the trained two-stage GRNN to obtain an adaptive threshold, and thus determining whether the superheterodyne receiver has a fault or not by comparing the residual of the superheterodyne receiver in the unknown state at the current moment with the adaptive threshold. And finally, when the superheterodyne receiver is determined to have a fault, processing the time-frequency domain characteristics of the residual signals of the superheterodyne receiver in a known state by using the trained PNN, and determining the fault type of the superheterodyne receiver.
It will be understood by those skilled in the art that all or part of the steps in the method according to the above embodiments may be implemented by a program, which may be stored in a computer-readable storage medium, and includes steps S101 to S107 when the program is executed. The storage medium may be ROM/RAM, magnetic disk, optical disk, etc.
Fig. 2 is a block diagram of an adaptive fault detection and diagnosis apparatus of a superheterodyne receiver according to an embodiment of the present invention, as shown in fig. 2, including:
the superheterodyne receiver 10 processes an input signal.
And the primary GRNN observer 20 is configured to process the input signal and an output signal of the superheterodyne receiver 10 at a previous time by using the trained primary GRNN to obtain an estimated output signal.
A calculating module 30, configured to obtain a residual error for fault detection according to the estimated output signal of the first-stage GRNN observer 20 and the output signal of the superheterodyne receiver 10 at the current time.
And the secondary GRNN observer 40 is configured to process the input signal and a residual error for fault detection obtained by the calculation module 30 by using the trained secondary GRNN to obtain an adaptive threshold.
A decision module 50, configured to determine whether the superheterodyne receiver fails according to the residual error for fault detection obtained by the calculation module 30 and the adaptive threshold obtained by the secondary GRNN observer 40.
A feature extraction module 60, configured to perform feature extraction on the residual error for fault detection obtained by the calculation module 30, so as to obtain time domain and frequency domain features of the residual error for fault detection.
And a PNN classifier 70, configured to process, by using the trained probabilistic neural network PNN, the time domain and frequency domain features of the residual error for fault detection obtained by the feature extraction module 60, and determine a fault type of the superheterodyne receiver.
In practical application, firstly, training each neural network, specifically including: the primary GRNN observer 20 trains a primary GRNN by using an input signal, an output signal of a superheterodyne receiver in a normal working state at a previous time and an output signal of a current time to obtain a primary GRNN after training; the secondary GRNN observer 40 trains a secondary GRNN by using the input signal and a residual signal of the superheterodyne receiver in a normal operating state to obtain a trained secondary GRNN; the PNN classifier 70 trains the PNN using the residual of the superheterodyne receiver of the known fault type to obtain a trained PNN. And then, carrying out fault detection by using the trained primary GRNN and secondary GRNN, wherein the fault detection specifically comprises the following steps: the primary GRNN observer 20 processes an input signal, an output signal of a superheterodyne receiver in an unknown state at a previous time and an output signal of a current time by using the trained primary GRNN to obtain an estimated output signal of the current time; the secondary GRNN observer 40 processes the input signal and a residual signal (i.e., a residual for fault detection) of the superheterodyne receiver in an unknown state by using the trained secondary GRNN to obtain an adaptive threshold, so that the decision module 50 determines whether the superheterodyne receiver fails by comparing the residual of the superheterodyne receiver in the unknown state at the current time with the adaptive threshold. Finally, when it is determined that the superheterodyne receiver fails, the PNN classifier 70 processes the time-frequency domain characteristics of the residual signal of the superheterodyne receiver in a known state by using the trained PNN, and determines the type of the failure of the superheterodyne receiver.
The following describes the embodiments of the present invention with reference to fig. 3 to 6.
The methods of fault detection and diagnosis may be classified into physical model-based methods and data-driven-based methods. Physical model-based methods require a clear understanding of the structural mechanisms of the study subject, requiring the determination of an accurate mathematical model. Due to the complex structure of the superheterodyne receiver, it is difficult to accurately obtain the physical model thereof. The method based on data driving does not need an accurate system model, can detect and diagnose faults only by using monitoring data, and is widely applied to complex systems in recent years. Huwenjin et al utilize a strong tracking filter to perform real-time fault diagnosis on the monitoring of a hydraulic servo system; the Luaachen et al utilizes a deep learning neural network to diagnose a fault of a rolling bearing; in the early 60 s, in order to achieve state feedback or other needs for control systems, d.g. luenberg, r.w. bara, and j.e. berberland et al proposed the concept and construction of state observers, which in recent years were used on top of fault detection for various control systems, and Jayakumar and Das proposed an observer-based aircraft control system fault detection, isolation and reconstruction method in 2006. Lixiong and the like provide a method for fault isolation and reconstruction of an aircraft control system by establishing a set of steady adaptive observers; and the Dianthus haichiensis and the like carry out fault detection on the aircraft environment control system by constructing a fault observer. It can be seen that the observer finds good application in complex systems/equipment. Therefore, the invention utilizes the observer to perform the final fault classification of the superheterodyne receiver.
The neural network method is a data-driven-based method, and can realize nonlinear and robust fault detection, isolation and health assessment. In the neural network method, the Generalized Regression Neural Network (GRNN) has the characteristics of stronger network approximation capability and higher learning speed and convergence speed when processing large sample data, can avoid generating a local minimum value, is used for accurately tracking and controlling the change of a system model, and can adaptively change the parameters of the neural network, thereby realizing the fault detection and diagnosis of the superheterodyne receiver. Meanwhile, because the superheterodyne receiver is interfered by an electromagnetic environment, a fault threshold value is influenced by system input, random interference, parameter drift, modeling errors, random noise and the like, and the traditional fault detection and diagnosis method based on the fixed threshold value cannot meet the requirements of practical engineering application. Therefore, the good learning capability and robustness of the GRNN neural network are utilized to generate the self-adaptive threshold of the fault so as to establish a self-adaptive threshold fault observer to detect the fault of the superheterodyne receiver. In the aspect of fault diagnosis of the superheterodyne receiver, the established fault observer can output a difference value between actual output and estimated output of the superheterodyne receiver in a fault state, namely a residual error, wherein the residual error contains information of a fault mode, so that the fault diagnosis is performed by performing feature extraction on the residual error. The Probabilistic Neural Network (PNN) is widely applied to the field of pattern recognition and pattern classification by taking Bayes decision and density function estimation as a theoretical basis, and the main idea is to separate a decision space in a multidimensional input space by using a Bayes decision rule, namely, the expected risk of wrong classification is minimum. The main advantage of PNN compared to BP networks is represented by the training and classification speed, so the present invention utilizes PNN for final fault classification of superheterodyne receivers.
Fig. 3 is a flowchart of adaptive fault detection and diagnosis of a superheterodyne receiver based on an observer and residual error analysis according to an embodiment of the present invention, and as shown in fig. 3, a dual stage GRNN adaptive fault observer (i.e., a GRNN dual stage observer) is first established to perform fault detection on the superheterodyne receiver, then residual error data of each fault mode is obtained through the established dual stage GRNN adaptive fault observer, time domain and time-frequency domain features are extracted from the residual error data, and finally, a PNN is used to perform fault diagnosis on the superheterodyne receiver.
GRNN-based self-adaptive fault observer establishment
GRNN neural network
The neural network has the characteristic of approximating any nonlinear function, and can be used for state recognition of a nonlinear system without establishing a state recognition model based on an actual mathematical model of the system. The GRNN neural network takes nonlinear regression analysis as a theoretical basis, and realizes the nonlinear regression analysis of the independent variable Y to the independent variable x by calculating Y with the maximum probability value.
FIG. 4 is the present inventionThe GRNN neural network structure diagram provided in the embodiment of the present invention is, as shown in fig. 4, different from the radial basis function neural network, that the GRNN neural network has a faster learning speed and an approaching speed, and is more advantageous in processing a large amount of nonlinear data. Corresponding network input X ═ X1,x2,...,xn]TThe corresponding network output is Y ═ Y1,y2,...,yk]T。
2. Establishing an observer
The residual is the difference between the actual output value and the expected output value of the superheterodyne receiver, and can be defined as the following equation (1):
wherein, γiIs the residual value, uiIs the actual output value of the output signal,is the desired output value.
As can be seen from equation (1), the residual magnitude reflects the difference between the actual output value and the desired output value. When a superheterodyne receiver is abnormal, the difference between the actual and expected outputs becomes large, i.e., the residual value becomes large. When the superheterodyne receiver fails, the residual error will exceed a set residual error threshold, that is, when the residual error exceeds a predetermined threshold, it can be determined that the system has failed. Therefore, whether the system fails can be judged by comparing the relative size relationship between the residual error of the current state and the residual error threshold value.
In practical application, the output of the system is not only affected by the input command of the system, but also affected by some uncertain factors such as parameter drift, modeling error, random noise and the like, system state, working condition and the like. Ignoring the effects of these non-fault factors will result in a high false alarm rate or a low fault detection rate. To solve this problem, the present invention uses an adaptive thresholding method to eliminate the effect of non-fault factors on the residual.
Each observer in the present invention contains two neural networks, one GRNN neural network for estimating the desired output of the system to produce the residual, and the other GRNN neural network for producing the adaptive threshold.
3. Residual sum threshold generator
Fig. 5 is a diagram illustrating the establishment of a fault observer according to an embodiment of the present invention, such as the dual stage GRNN neural network observer shown in fig. 5.
(1) And establishing a primary GRNN neural network residual error generator.
Because the superheterodyne receiver is an integral system and is difficult to acquire parameter values of internal components, Z is added in front of the input end of the GRNN neural network observer-1A link, so as to simulate a hysteresis link of the output hysteresis function of the real control assembly; control signal c of actuator superheterodyne receiver in normal working state is collectedi(k) Output signal u of superheterodyne receiver at previous timei(k-1) and the output signal u at the current timei(k)，k＝1,2,3,…,n。
Superheterodyne receiver control signal c to be acquiredi(k) And the output signal u of the previous time superheterodyne receiveri(k-1) synthesizing a vector as a training input sample of the GRNN neural network observer, and outputting a signal u of the superheterodyne receiver at the current momenti(k) As output samples of observer training.
Normalizing the input and output samples for network training to be between [ -1,1], setting the basic parameters of the GRNN neural network, and training the network. And after the training is finished, storing the trained GRNN neural network observer.
Then the control signal ci(k) And the output signal u of the previous time superheterodyne receiveri(k-1) synthesizing a vector and normalizing the vector to be used as the input vector of the trained GRNN neural network observer, so that the estimated value output by the superheterodyne receiver at the moment can be obtained(i.e., the desired output value), k is 1,2,3, …, n.
Estimating value output by GRNN observerWith the collected superheterodyne receiver true output value ui(k) Taking the difference, the residual signal r (k) at the moment of the superheterodyne receiver can be obtained, where k is 1,2,3, …, n.
(2) And establishing a two-stage GRNN neural network adaptive threshold generator.
The adaptive threshold value is that the fault threshold value changes along with the change of system input instructions and working conditions, and can be obtained through the synchronization of a trained GRNN neural network. A residual signal r (k) in a normal state of the superheterodyne receiver acquired by the primary observer, where k is 1,2,3, …, n, is defined as a reference residual th.
Control signal c with superheterodyne receiveri(k) And the residual r (k) in the normal state is a training input vector of the secondary GRNN adaptive threshold generator, and the expected adaptive threshold is used as an output vector of the network. The desired adaptive threshold is defined as the following equation (2):
wherein,to the desired threshold, r (k) is the residual, and b is the correction factor. After training is completed, the trained GRNN adaptive threshold generator is saved.
After the training of the residual generator GRNN neural network is completed, the building of the dual stage observer is completed, and can be used for fault detection. Firstly, inputting test data into a GRNN residual error generator to obtain a residual error value at the moment; and then inputting the residual value and a control command into a secondary GRNN adaptive threshold generator to obtain an adaptive threshold. And comparing the residual error with the adaptive threshold, and if the residual error is larger than the adaptive threshold, indicating that the superheterodyne receiver has failed.
Fault diagnosis based on residual error analysis of GRNN observer
1. Feature extraction
After residual errors of all fault modes of the superheterodyne receiver are obtained through the constructed GRNN fault observer, feature extraction is carried out on the obtained residual errors, wherein the feature extraction comprises time domain features and frequency domain features.
The time domain characteristics are characteristic information of the signals which are relatively intuitive, and the time domain characteristics contain a large amount of information which can reflect the basic characteristics of the signals, so the time domain characteristics of the residual error of the superheterodyne receiver are extracted to be used as fault characteristic values. The temporal feature extraction is shown in table 1.
TABLE 1 time domain characterization formula
The frequency domain characteristics can reflect the change of the signal on the frequency structure, but for the non-stationary nonlinear signal, the frequency domain analysis cannot reflect the change situation of the instantaneous frequency of the signal along with the time, so the time-frequency domain characteristics are extracted from the residual error of the superheterodyne receiver. Wavelet packet decomposition can provide a more refined analysis method for signals, frequency bands are divided in a multi-level mode, high-frequency parts which are not subdivided in multi-resolution analysis are further decomposed, and the maximum value of energy of each node after wavelet packet decomposition can be selected as the time-frequency domain characteristics of original residual signals according to the characteristics of the analyzed signals. The components of the signal X (n) after wavelet packet transformation and reconstruction can form a matrix D, the singular value of the matrix is solved, the singular value is combined with the information entropy theory, and the wavelet packet singular entropy is calculated through the following formula (3):
where λ is the singular value of the matrix D.
The wavelet packet singular entropy can reflect the complexity of the signal, and the more complex the signal is, the larger the wavelet singular entropy is. The average value of the wavelet packet decomposition coefficients is also calculated as the characteristic of the residual signal.
2. Probabilistic neural network
The Probabilistic Neural Network (PNN) is a neural network model combining a statistical method and a feedforward neural network, has the characteristics of simple structure, quick training and the like, and is very widely applied. In the mode classification, the method has the advantages that the linear learning algorithm can be used for finishing the work of the conventional nonlinear algorithm, and the high-precision characteristic of the nonlinear algorithm can be kept. Fig. 6 is a structural model of a PNN according to an embodiment of the present invention, and as shown in fig. 6, the PNN generally includes 4 layers, i.e., an input layer, a mode layer, a summation layer, and an output layer, where each layer is composed of a plurality of units.
The input layer passes the training samples to the network. The mode layer calculates the matching relation between the input feature vector and each mode in the training set, and sends the matching relation to the Gaussian activation function according to the distance to obtain the output of the mode layer, wherein the number of neurons in the normal mode layer is the same as the number of input sample vectors (namely the number of training samples). The summation layer simply sums the inputs from the pattern layers corresponding to the same class in the training samples to obtain the maximum likelihood that the input sample belongs to that class, regardless of the outputs of the pattern layer units belonging to other classes. Thus, each class has only one summation layer element, which is connected to the mode layer elements belonging only to its class, but not to other elements in the mode layer. Various types of probability estimation can be obtained through normalization processing of the output layer. The output layer neuron is a type of competitive neuron, and each neuron corresponds to a data type. The output layer neuron number is equal to the class number of the training sample data, which receives the various classes of probability density functions output from the summation layer. The neuron with the largest probability density function outputs 1, that is, the corresponding class is the sample pattern class to be identified, and the outputs of other neurons are all 0.
Examples provided by embodiments of the present invention are verified below with reference to fig. 7 to 12.
1. Generation of test data
In this embodiment, a MATLAB/Simulink simulation model is used to generate experimental data of a superheterodyne receiver. Fig. 7 is a superheterodyne receiver simulation model according to an embodiment of the present invention, and as shown in fig. 7, the superheterodyne receiver simulation model includes a signal generating system, a mixer, an intermediate frequency filter, a low pass filter, an intermediate frequency amplifier, and the like.
When the fault injection of the simulation model of the superheterodyne receiver system is performed, the fault injection is performed by selecting the fault of an electronic amplifier, the band-pass reduction fault of an intermediate frequency filter and the local oscillation frequency deviation fault, and the fault injection mode is shown in table 2. In the simulation process, the simulation time is set to be 0.1s, the simulation sampling rate is 60kHz, and 6000 sampling points are obtained for each group of fault data.
TABLE 2 simulation model Fault injection
| Type of failure | Failure mode | Fault injection method |
| Failure of electronic discharger | Gain factor reduction | The gain factor is changed from 20 to 17 |
| Fault of intermediate frequency discharge filter | Intermediate frequency filter bandpass reduction | The band pass is changed from 465 +/-6 kHz to 465 +/-0.5 kHz |
| Local oscillator failure | Local oscillator frequency deviation | The local oscillation frequency is changed from 465kHz to 460kHz |
2. Adaptive fault detection for superheterodyne receivers
Under normal conditions, the GRNN neural network is trained with the input signal and the output from the previous time as inputs, and a threshold parameter b is set to 0.01. Fig. 8 shows the training effect of the neural network residual and the adaptive threshold trained in the normal state according to the embodiment of the present invention, as shown in fig. 8, the upper gray line in the graph represents the adaptive threshold, and the lower black line represents the residual, and it can be seen from fig. 8 that when the superheterodyne receiver is in the normal state, the fault threshold (i.e., the adaptive threshold) is always higher than the fault residual (i.e., the residual). And taking the neural network model trained in the normal state as a network model of the fault observer, and performing fault detection on the superheterodyne receiver by inputting fault data.
When the electronic amplifier is failed, a normal control signal and an output signal of the failure at the previous time are input into an observer as input data, and a detection result is obtained. Fig. 9 is a detection result of the electronic amplifier provided by the embodiment of the present invention when it is faulty, and it can be seen from fig. 9 that the observer input residual exceeds the adaptive threshold, so that the superheterodyne receiver is detected to be faulty.
When the intermediate frequency filter is in fault, a normal control signal and an output signal of the fault at the previous moment are input into an observer as input data, and a detection result is obtained. Fig. 10 is a detection result of the if filter provided by the embodiment of the present invention when it is faulty, and it can be seen from fig. 10 that the observer input residual exceeds the adaptive threshold, so that the superheterodyne receiver is detected to be faulty.
When the local oscillator is in fault, normal control signals and output signals of faults at the previous moment are input into an observer as input data, and a detection result is obtained. Fig. 11 is a detection result when the local oscillator fails according to the embodiment of the present invention, and it can be seen from fig. 11 that the input residual of the observer exceeds the adaptive threshold, so that it is detected that the superheterodyne receiver fails.
3. Superheterodyne receiver fault diagnosis
After the superheterodyne receiver is subjected to fault detection, it is detected that the superheterodyne receiver is in fault, but it is not known what kind of fault occurs, so that it needs to be subjected to further fault diagnosis to determine the fault type thereof. When the GRNN-based self-adaptive fault observer model is established to detect the faults of the superheterodyne receiver, the residual errors of different fault types of the superheterodyne receiver are obtained, and therefore when the fault diagnosis of the superheterodyne receiver is carried out, the residual errors of different fault types of the superheterodyne receiver are used as data to be processed.
One of the key steps of fault diagnosis is feature extraction, so after residual errors of fault modes of the superheterodyne receiver are obtained, feature extraction is performed on residual error data of the fault modes. According to the above, the time domain characteristics and the time-frequency domain characteristics of the residual signal are used as the fault characteristics, and then fault diagnosis is performed.
During feature extraction, 465 points of residual data of each fault mode are selected as a group of data each time, 275 groups of data are selected for each fault mode to extract time domain and time-frequency domain features respectively, therefore, 1100 groups of fault features are selected in a normal state, under the condition of gain faults of an electronic amplifier, medium-frequency filters and local oscillator deviation faults, the first 220 groups of fault features are selected as training data under each fault mode, a PNN classifier is trained, and then the last 55 groups of fault feature data under each group of fault modes are used as test data to be tested. Fig. 12 is a schematic diagram of a predicted classification result of the PNN classifier provided in the embodiment of the present invention, and as shown in fig. 12, the calculated classification accuracy is 96.36%.
In summary, the present invention can perform adaptive fault detection and diagnosis on the superheterodyne receiver, and firstly, a GRNN-based two-stage observer is established according to the superheterodyne receiver, wherein the first-stage GRNN neural network observer is used to estimate an expected output of a system and obtain a residual error, the second-stage GRNN neural network is used to generate an adaptive threshold, and whether the superheterodyne receiver has a fault is determined by the generated adaptive threshold and the residual error. Because the GRNN-based double-stage observer can only detect whether the superheterodyne receiver has faults and cannot judge which faults occur, the method extracts the characteristics of the residual errors in various fault modes, and then utilizes the PNN to diagnose the faults of the superheterodyne receiver, and test results show that the method can well detect and diagnose the faults of the superheterodyne receiver.
Although the present invention has been described in detail hereinabove, the present invention is not limited thereto, and various modifications can be made by those skilled in the art in light of the principle of the present invention. Thus, modifications made in accordance with the principles of the present invention should be understood to fall within the scope of the present invention.
Claims (10)
1. A method for adaptive fault detection and diagnosis of a superheterodyne receiver, comprising:
inputting an input signal to a superheterodyne receiver, and processing the input signal by the superheterodyne receiver;
processing the input signal and an output signal of the superheterodyne receiver at the previous moment by using a trained primary generalized regression neural network GRNN to obtain an estimated output signal;
obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current moment;
processing the input signal and the residual error for fault detection by using the trained secondary GRNN to obtain a self-adaptive threshold;
and determining whether the superheterodyne receiver fails according to the residual error for fault detection and the adaptive threshold.
2. The method of claim 1, wherein the step of deriving a residual error for fault detection based on the estimated output signal and an output signal of the superheterodyne receiver at a current time comprises:
and subtracting the estimated output signal output by the primary GRNN with the output signal of the superheterodyne receiver at the current moment to obtain the residual error of the superheterodyne receiver at the moment.
3. The method of claim 1, wherein determining whether the superheterodyne receiver is malfunctioning based on the residual error for failure detection and the adaptive threshold comprises:
and comparing the residual error for fault detection with the adaptive threshold, and determining that the superheterodyne receiver has a fault if the residual error for fault detection is greater than the adaptive threshold.
4. A method according to any of claims 1-3, wherein the trained one-stage GRNN is obtained by:
inputting an input signal to a superheterodyne receiver in a normal working state, and processing the input signal by the superheterodyne receiver in the normal working state;
and taking the input signal and the output signal of the superheterodyne receiver in the normal working state at the previous moment as the input of the primary GRNN, taking the output signal of the superheterodyne receiver in the normal working state at the current moment as the output of the primary GRNN, and training the primary GRNN to obtain the trained primary GRNN.
5. A method according to any of claims 1-3, wherein the trained two-stage GRNN is obtained by:
obtaining a residual error for training a secondary GRNN according to the estimated output signal of the trained primary GRNN and the output signal of the superheterodyne receiver in the normal working state at the current moment;
and taking the input signal and the residual error for training the secondary GRNN as the input of the secondary GRNN, taking an expected threshold value as the output of the secondary GRNN, and training the secondary GRNN to obtain the trained secondary GRNN.
6. A method according to any of claims 1-3, characterized in that the method further comprises the step of diagnosing a malfunction of the superheterodyne receiver.
7. The method of claim 6, wherein the step of diagnosing a failure of the superheterodyne receiver comprises:
extracting the characteristics of the residual error for fault detection to obtain the time domain and frequency domain characteristics of the residual error for fault detection;
and processing the time domain and frequency domain characteristics of the residual error for fault detection by using the trained probabilistic neural network PNN, and determining the fault type of the superheterodyne receiver.
8. The method of claim 6, wherein the trained PNN is obtained by:
performing characteristic extraction on the residual error of the superheterodyne receiver with the known fault type to obtain time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type;
and training the PNN for fault classification by using the time domain and frequency domain characteristics of the residual error of the superheterodyne receiver with the known fault type to obtain the PNN after training.
9. An adaptive fault detection and diagnostic apparatus for a superheterodyne receiver, comprising:
the superheterodyne receiver is used for processing an input signal;
the primary GRNN observer is used for processing the input signal and an output signal of the superheterodyne receiver at the previous moment by utilizing a trained primary generalized regression neural network GRNN to obtain an estimated output signal;
the calculation module is used for obtaining a residual error for fault detection according to the estimated output signal and the output signal of the superheterodyne receiver at the current moment;
the secondary GRNN observer is used for processing the input signal and the residual error for fault detection by utilizing the trained secondary GRNN to obtain a self-adaptive threshold;
and the judging module is used for determining whether the superheterodyne receiver has a fault according to the residual error for fault detection and the adaptive threshold.
10. The apparatus of claim 9, further comprising:
the characteristic extraction module is used for extracting the characteristics of the residual error for fault detection to obtain the time domain and frequency domain characteristics of the residual error for fault detection;
and the PNN classifier is used for processing the time domain and frequency domain characteristics of the residual error for fault detection by using the trained probabilistic neural network PNN to determine the fault type of the superheterodyne receiver.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610757171.8A CN106301610A (en) | 2016-08-29 | 2016-08-29 | The adaptive failure detection of a kind of superhet and diagnostic method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610757171.8A CN106301610A (en) | 2016-08-29 | 2016-08-29 | The adaptive failure detection of a kind of superhet and diagnostic method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106301610A true CN106301610A (en) | 2017-01-04 |
Family
ID=57675993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610757171.8A Pending CN106301610A (en) | 2016-08-29 | 2016-08-29 | The adaptive failure detection of a kind of superhet and diagnostic method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106301610A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110008548A (en) * | 2019-03-12 | 2019-07-12 | 宁波大学 | A kind of fault detection method based on GRNN distributing modeling strategy |
| CN111680356A (en) * | 2020-05-06 | 2020-09-18 | 北京航空航天大学 | Method for evaluating real-time safety performance of unmanned aerial vehicle actuator based on double-stage observer |
| CN111771042A (en) * | 2017-10-11 | 2020-10-13 | 英国石油勘探运作有限公司 | Detecting events using acoustic frequency domain features |
| CN113419226A (en) * | 2021-08-24 | 2021-09-21 | 四川锦美环保股份有限公司 | Radar troubleshooting system |
| CN115913993A (en) * | 2022-12-12 | 2023-04-04 | 千寻位置网络有限公司 | GNSS receiver observation data fault modeling and simulation method and device thereof |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0165698A1 (en) * | 1984-05-17 | 1985-12-27 | Honeywell Inc. | Servomotor feedback fault detector |
| EP0743577A1 (en) * | 1995-05-15 | 1996-11-20 | The Boeing Company | Signal selection and fault detection apparatus and method |
| JPH10124134A (en) * | 1996-10-21 | 1998-05-15 | Osaka Gas Co Ltd | Diagnostic system |
| KR20020028626A (en) * | 2000-10-11 | 2002-04-17 | 윤종용 | Fault diagnosis apparatus of nonlinear system based on dynamic threshold and method thereof |
| US20040181329A1 (en) * | 2002-08-01 | 2004-09-16 | Tseng Hongtei Eric | System and method for detecting roll rate sensor fault |
| CN102510044A (en) * | 2011-11-04 | 2012-06-20 | 上海电力学院 | Excitation inrush current identification method based on wavelet transformation and probabilistic neural network (PNN) |
| CN102705303A (en) * | 2012-05-16 | 2012-10-03 | 北京航空航天大学 | Fault location method based on residual and double-stage Elman neural network for hydraulic servo system |
| CN103259485A (en) * | 2013-06-14 | 2013-08-21 | 重庆交通大学 | Method of improving identification precision of speedless sensor under condition of unbalanced network voltage |
| CN103353752A (en) * | 2013-04-12 | 2013-10-16 | 北京航空航天大学 | Aircraft environment control system control component fault diagnosis method based on four-level RBF neural network |
| CN103778467A (en) * | 2014-01-16 | 2014-05-07 | 天津大学 | Power system transient state stability estimation inputted characteristic quantity selection method |
| CN104200269A (en) * | 2014-09-09 | 2014-12-10 | 江南大学 | Real-time fault diagnosis method based on online learning minimum embedding dimension network |
| CN104390776A (en) * | 2014-12-10 | 2015-03-04 | 北京航空航天大学 | Fault detection, diagnosis and performance evaluation method for redundant aileron actuator |
| CN104850889A (en) * | 2014-11-19 | 2015-08-19 | 北京航空航天大学 | Airplane rotation actuator drive unit adaptive fault detection, isolation and confidences assessment method |
| CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
| CN105003372A (en) * | 2015-06-15 | 2015-10-28 | 天津大学 | Oil spraying quantity fault diagnosis method based on rail pressure waveform characteristic parameter observation |
| CN105129109A (en) * | 2015-09-30 | 2015-12-09 | 北京航空航天大学 | Method for evaluating health of aircraft aileron actuator system based on multi-fractal theory and self-organizing map (SOM) network |
| CN105758645A (en) * | 2014-12-20 | 2016-07-13 | 哈尔滨智晟天诚科技开发有限公司 | Probabilistic neural network based engine fault diagnosis system |
-
2016
- 2016-08-29 CN CN201610757171.8A patent/CN106301610A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0165698A1 (en) * | 1984-05-17 | 1985-12-27 | Honeywell Inc. | Servomotor feedback fault detector |
| EP0743577A1 (en) * | 1995-05-15 | 1996-11-20 | The Boeing Company | Signal selection and fault detection apparatus and method |
| JPH10124134A (en) * | 1996-10-21 | 1998-05-15 | Osaka Gas Co Ltd | Diagnostic system |
| KR20020028626A (en) * | 2000-10-11 | 2002-04-17 | 윤종용 | Fault diagnosis apparatus of nonlinear system based on dynamic threshold and method thereof |
| US20040181329A1 (en) * | 2002-08-01 | 2004-09-16 | Tseng Hongtei Eric | System and method for detecting roll rate sensor fault |
| CN102510044A (en) * | 2011-11-04 | 2012-06-20 | 上海电力学院 | Excitation inrush current identification method based on wavelet transformation and probabilistic neural network (PNN) |
| CN102705303A (en) * | 2012-05-16 | 2012-10-03 | 北京航空航天大学 | Fault location method based on residual and double-stage Elman neural network for hydraulic servo system |
| CN103353752A (en) * | 2013-04-12 | 2013-10-16 | 北京航空航天大学 | Aircraft environment control system control component fault diagnosis method based on four-level RBF neural network |
| CN103259485A (en) * | 2013-06-14 | 2013-08-21 | 重庆交通大学 | Method of improving identification precision of speedless sensor under condition of unbalanced network voltage |
| CN103778467A (en) * | 2014-01-16 | 2014-05-07 | 天津大学 | Power system transient state stability estimation inputted characteristic quantity selection method |
| CN104200269A (en) * | 2014-09-09 | 2014-12-10 | 江南大学 | Real-time fault diagnosis method based on online learning minimum embedding dimension network |
| CN104850889A (en) * | 2014-11-19 | 2015-08-19 | 北京航空航天大学 | Airplane rotation actuator drive unit adaptive fault detection, isolation and confidences assessment method |
| CN104390776A (en) * | 2014-12-10 | 2015-03-04 | 北京航空航天大学 | Fault detection, diagnosis and performance evaluation method for redundant aileron actuator |
| CN105758645A (en) * | 2014-12-20 | 2016-07-13 | 哈尔滨智晟天诚科技开发有限公司 | Probabilistic neural network based engine fault diagnosis system |
| CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
| CN105003372A (en) * | 2015-06-15 | 2015-10-28 | 天津大学 | Oil spraying quantity fault diagnosis method based on rail pressure waveform characteristic parameter observation |
| CN105129109A (en) * | 2015-09-30 | 2015-12-09 | 北京航空航天大学 | Method for evaluating health of aircraft aileron actuator system based on multi-fractal theory and self-organizing map (SOM) network |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111771042A (en) * | 2017-10-11 | 2020-10-13 | 英国石油勘探运作有限公司 | Detecting events using acoustic frequency domain features |
| CN110008548A (en) * | 2019-03-12 | 2019-07-12 | 宁波大学 | A kind of fault detection method based on GRNN distributing modeling strategy |
| CN110008548B (en) * | 2019-03-12 | 2023-02-21 | 宁波大学 | Fault detection method based on GRNN distributed modeling strategy |
| CN111680356A (en) * | 2020-05-06 | 2020-09-18 | 北京航空航天大学 | Method for evaluating real-time safety performance of unmanned aerial vehicle actuator based on double-stage observer |
| CN111680356B (en) * | 2020-05-06 | 2023-10-10 | 北京航空航天大学 | Unmanned aerial vehicle actuator real-time safety performance evaluation method based on two-stage observer |
| CN113419226A (en) * | 2021-08-24 | 2021-09-21 | 四川锦美环保股份有限公司 | Radar troubleshooting system |
| CN115913993A (en) * | 2022-12-12 | 2023-04-04 | 千寻位置网络有限公司 | GNSS receiver observation data fault modeling and simulation method and device thereof |
| CN115913993B (en) * | 2022-12-12 | 2024-06-04 | 千寻位置网络有限公司 | GNSS receiver observation data fault modeling and simulation method and device thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106301610A (en) | The adaptive failure detection of a kind of superhet and diagnostic method and device | |
| WO2019080367A1 (en) | Method for evaluating health status of mechanical device | |
| CN111611718B (en) | Actuator fault estimation method based on Bayesian learning | |
| CN103018660B (en) | Multi-fault intelligent diagnosing method for artificial circuit utilizing quantum Hopfield neural network | |
| CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
| CN108804720B (en) | Oil pumping machine fault diagnosis method based on improved traceless Kalman filtering and RBF neural network | |
| CN106405384A (en) | Simulation circuit health state evaluation method | |
| CN109740687A (en) | A kind of fermentation process fault monitoring method based on DLAE | |
| CN111324110A (en) | Fermentation process fault monitoring method based on multiple shrinkage automatic encoders | |
| Li et al. | Signal frequency domain analysis and sensor fault diagnosis based on artificial intelligence | |
| CN110516742A (en) | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model | |
| CN104634603A (en) | Early fault diagnosis method for complex equipment | |
| CN114266197A (en) | Method for diagnosing equipment fault of hydropower station | |
| CN116595319A (en) | Prediction method and system applied to rail transit motor health state evaluation | |
| Desmet et al. | Leak detection in compressed air systems using unsupervised anomaly detection techniques | |
| CN108804721B (en) | Oil pumping machine fault diagnosis method based on self-adaptive unscented Kalman filtering and RBF neural network | |
| CN109635430A (en) | Grid power transmission route transient signal monitoring method and system | |
| Kahlen et al. | Improving Machine Learning Diagnostic Systems with Model-Based Data Augmentation―Part B: Application | |
| CN115047853B (en) | Minor fault detection method based on recursion standard variable residual error and kernel principal component analysis | |
| CN115729200B (en) | Unmanned aerial vehicle steering engine fault detection model construction method and device, and unmanned aerial vehicle steering engine fault detection method and device | |
| CN116227172A (en) | Rolling bearing performance degradation evaluation method based on convolutional neural network | |
| Xu et al. | A graph convolutional shrinkage network-based fault diagnosis method for industrial process | |
| Gomez-Gil et al. | Intelligent identification of induction motor conditions at several mechanical loads | |
| CN102830341A (en) | Online intelligent fault prediction method for power electronic circuit based on RS-CMAC (rough sets and cerebellar model articulation controller) | |
| Fomin et al. | Construction of diagnostic features space using Volterra kernels moments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20200103 |