Radiation immunity fault diagnosis device and method based on multi-source heterogeneous observation

Abstract

The application discloses a radiation interference fault diagnosis device and method based on multi-source heterogeneous observation, relates to the technical field of product quality detection, and comprises a radiation interference module, an energy collection module, a spectrum processing module, a signal conversion module and a fault diagnosis module; the interference test environment is built through the radiation interference module, the optical information of the equipment to be tested is collected by the energy collection module, is coupled, is subjected to spectrum separation through the spectrum processing module, target central wavelength synthesis is carried out, and target light intensity information is acquired; signal conversion is carried out through the signal conversion module, multi-source heterogeneous optical characteristics are extracted by the fault diagnosis module, fault identification is carried out, and a fault identification result is output; the problems that the anti-interference detection efficiency of the equipment to be tested is low and the accuracy is poor due to excessive dependence on subjective judgment are overcome, the fault diagnosis efficiency and accuracy in the optical information sequence are realized, and the anti-interference detection efficiency, accuracy and reliability of radiation are further improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of product quality detection, and particularly relates to a radiation immunity fault diagnosis device and method based on multi-source heterogeneous observation. BACKGROUND

[0002] With the continuous improvement of the intelligent level of electronic and electrical equipment, the equipment often has a state monitoring function and prompts the user in the form of an indicator light. This is also the most direct information source of the abnormal state of the equipment. The light intensity, color and flicker frequency of the indicator light are monitored to determine whether the indicator light can normally work under radiation immunity. At present, in electromagnetic compatibility tests such as radiation immunity tests, the working state of the indicator light is mainly determined by the way that a technician observes a video according to his own experience. Although the operation is simple, it is difficult to judge the rapid changes of the brightness, color and frequency of the indicator light, subjective errors are prone to occur, and the scene of simultaneous changes of multiple indicator lights of the equipment is difficult to deal with. This poses great challenges to the accuracy of the collected equipment and the concentration of the detection personnel during the test.

[0003] A kind of automatic detection device and method for electric energy meter radiation immunity based on residence time self-adaptation and pulse light flicker identification are disclosed in Chinese patent with publication number CN109946638B and publication date January 8, 2021. The frequency range and field strength of the interference electromagnetic field are determined according to the relevant information of the measured electric energy meter. The electric energy signals output by the measured electric energy meter and the standard electric energy meter are collected in real time, and the percentage error of the output signals of the measured electric energy meter and the standard electric energy meter under the specified verification number of turns is calculated. The residence time is determined, and the frequency of the interference magnetic field is changed. The percentage error frequency graph is automatically drawn based on the residence time and the frequency of the interference magnetic field to complete the detection of the electric energy meter under the current state in all frequency ranges. However, this scheme needs to collect the data generated by the electric energy meter during operation for analysis, and is only used for detection of the electric energy meter to improve the detection efficiency of the electric energy meter, which has poor applicability. SUMMARY

[0004] The application aims at the problem of low detection efficiency and poor effect caused by subjective judgment in existing radiation immunity detection, and provides a radiation immunity fault diagnosis device and method based on multi-source heterogeneous observation, which builds an interference test environment through a radiation interference module, collects optical information of a to-be-tested device through an energy collection module, couples the optical information, performs spectral separation through a spectrum processing module, synthesizes target central wavelengths, obtains target light intensity information, performs signal conversion through a signal conversion module, extracts multi-source heterogeneous optical features through a fault diagnosis module, performs fault recognition, and outputs fault recognition results.

[0005] To solve the above technical problems, according to a first aspect of an embodiment of the application, a radiation immunity fault diagnosis device based on multi-source heterogeneous observation is provided, which comprises:

[0006] A radiation interference module is configured to build an interference test environment for a to-be-tested device.

[0007] An energy collection module is configured to collect a plurality of optical information generated by the to-be-tested device in the interference test environment, and couple the plurality of optical information.

[0008] A spectrum processing module is configured to perform spectral separation on the coupled optical information, synthesize light intensity data with target central wavelengths, and obtain target light intensity information.

[0009] A signal conversion module is configured to perform signal conversion on the target light intensity information, and obtain light intensity digital signals.

[0010] A fault diagnosis module is configured to extract multi-source heterogeneous optical feature information based on the light intensity digital signals, perform fault recognition, and output fault recognition results.

[0011] In the scheme, the performance of the to-be-tested device under specific conditions can be evaluated by building an interference test environment of the to-be-tested device, to ensure the consistency and repeatability of the test environment, and to provide reliable data basis for subsequent fault diagnosis; the energy collection module can optically collect visible light information with a small caliber and a large field of view, and after efficient coupling of the collected optical information, spectral separation and signal conversion are performed, high-precision capture and demodulation of optical information such as light intensity, wavelength distribution and flicker frequency are realized, and the optical characteristics related to radiation interference can be more effectively identified, providing standardized data input for fault diagnosis, thereby improving the accuracy of fault frequency point diagnosis, and further assisting the detection personnel to effectively cope with problems such as inability to capture transient changes of indicator lights, loss of weak brightness change information, and mutual interference of multiple indicator light information in experiments, thereby significantly improving the detection efficiency, accuracy and reliability.

[0012] Preferably, the radiation interference module comprises an electric wave bearing device and an electric field generating device.

[0013] The electric wave bearing device is used to accommodate a semi-electric wave darkroom of the to-be-tested device, to shield the interference electromagnetic field in the test environment.

[0014] The electric field generating device is used to generate interference electromagnetic waves and act on the to-be-tested device.

[0015] Preferably, the electric field generating device at least comprises a signal generator, a power amplifier and a radio frequency cable.

[0016] The signal generator is used to produce a target radio frequency signal.

[0017] The power amplifier is used to amplify the target radio frequency signal and transmit it to the radio frequency cable, so that the radio frequency cable corresponding antenna group generates a standard interference signal corresponding to the target radio frequency signal.

[0018] Preferably, the energy collection module comprises an electromagnetic non-interference unit and an information coupling unit.

[0019] The electromagnetic non-interference unit comprises a plurality of electromagnetic non-interference channels made of electromagnetic wave non-interference materials, which collect optical information of visible wave bands emitted by the state indicator light component of the to-be-tested device through the electromagnetic non-interference channels, and obtain the optical information.

[0020] The information coupling unit is used to couple the optical information and send it to the spectral processing module.

[0021] Preferably, the spectral processing module comprises a spectral separation submodule and a spectral synthesis submodule.

[0022] The spectrum separation sub-module is configured to perform equal-ratio splitting on the optical information, perform band-pass filtering spectrum separation on the split signal, and extract optical intensity data of the split signal.

[0023] The spectrum synthesis sub-module is configured to synthesize the filter wavelength and the actual center wavelength of the optical intensity data, and take the wavelength-synthesized optical intensity data as target optical intensity information.

[0024] Preferably, the signal conversion module comprises a first converter and a second converter.

[0025] The first converter is configured to convert an optical signal sequence of the target optical intensity information into a target electrical signal sequence.

[0026] The second converter is configured to perform analog-digital conversion on the target electrical signal sequence to obtain an optical intensity digital signal.

[0027] Preferably, the fault diagnosis module comprises a first fault diagnosis unit and a second fault diagnosis unit.

[0028] The first fault diagnosis unit is configured to extract heterogeneous optical characteristic information according to the optical intensity digital signal.

[0029] The second fault diagnosis unit is configured to diagnose a fault position by taking the heterogeneous optical characteristic information as an input of a fault detection model.

[0030] Preferably, the signal generator comprises a key configuration unit for generating a target radio frequency signal.

[0031] The key configuration unit adopts an upper computer and control software to automatically control key parameters, wherein the key parameters at least include an output frequency, a level, and a modulation parameter.

[0032] According to another aspect of the embodiment of the present application, a radiation immunity fault diagnosis method based on multi-source heterogeneous observation is provided, comprising the following steps:

[0033] S1, constructing an interference test environment of a device to be tested;

[0034] S2, collecting optical information of the device to be tested in response to the interference test environment, and performing optical intensity coupling on the optical information to obtain an optical coupling information sequence;

[0035] S3, performing spectrum separation and target light wave synthesis on the optical coupling information sequence to obtain target optical intensity information;

[0036] S4, performing signal conversion on the target optical intensity information to obtain an optical intensity digital signal;

[0037] S5, diagnose the fault position by taking the light intensity digital signal as input of a fault detection model, and output a fault identification result.

[0038] In the scheme, the performance of the to-be-tested device under specific conditions can be evaluated by building an interference test environment of the to-be-tested device, so as to ensure consistency and repeatability of the test environment and provide reliable data basis for subsequent fault diagnosis; the energy collection module can optically collect visible light information with a small aperture and a large field of view, and after efficient coupling of the collected optical information, spectrum separation and signal conversion are performed, high-precision capture and demodulation of optical information such as light intensity, wavelength distribution and flicker frequency are realized, the optical characteristics related to radiation interference can be more effectively identified, standardized data input is provided for fault diagnosis, so as to improve the accuracy of fault frequency point diagnosis, and further assist the detection personnel to effectively cope with problems such as inability to capture transient changes of indicator lights, loss of weak brightness change information, mutual interference of multiple indicator light information and the like in experiments, and the detection efficiency, accuracy and reliability are significantly improved.

[0039] Preferably, the fault position diagnosis by taking the light intensity digital signal as input of the fault detection model comprises: performing multi-source heterogeneous optical feature extraction on the light intensity digital signal based on a long short-term memory neural network algorithm combined with an automatic encoder, and converting the multi-source heterogeneous optical features into time-based optical feature sequences;

[0040] The optical feature sequences are trained to construct a fault detection model;

[0041] Based on the fault detection model, the reconstruction error of the multi-source heterogeneous optical sample data of the to-be-tested device is calculated, the abnormal state of the sample is judged based on the calculation result, and the abnormal sample is integrated to identify the fault position.

[0042] The beneficial effects of the present application are:

[0043] 1. The electromagnetic non-perturbation channel can optically collect visible light information with a small aperture and a large field of view, and after efficient coupling of the collected optical information, spectrum separation is performed, high-precision capture and demodulation of optical information such as light intensity, wavelength distribution, flicker frequency and the like are realized;

[0044] 2. The long short-term memory neural network algorithm and the automatic encoder are combined to perform data dimension reduction and feature extraction on the optical information based on the fault diagnosis module, more effective multi-source heterogeneous optical features related to radiation interference are extracted from the light intensity digital signal, and the accuracy of fault diagnosis is improved;

[0045] 3. By calculating the reconstruction error of multi-source heterogeneous optical sample data, the fault frequency point can be quickly identified, and the corresponding interference frequency band can be quickly located. This makes it easier for testing personnel to reproduce the problem in the corresponding frequency band, identify the fault type, and analyze the cause of the fault. Attached Figure Description

[0046] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0047] Figure 1 This is a schematic diagram of the organizational structure of a radiation immunity fault diagnosis device based on multi-source heterogeneous observation, according to an embodiment of the present invention.

[0048] Figure 2 This is a flowchart of a radiation immunity fault diagnosis method based on multi-source heterogeneous observation, according to an embodiment of the present invention.

[0049] Figure 3 This is a schematic diagram of a data training process according to a specific embodiment of the present invention.

[0050] Figure 4 This is a schematic diagram of a fault diagnosis process according to a specific embodiment of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only one preferred embodiment of this invention and are only used to explain this invention. They do not limit the scope of protection of this invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0052] Example 1: As Figure 1 As shown, a radiation immunity fault diagnosis device based on multi-source heterogeneous observation includes: a radiation interference module, used to build an interference test environment for the device under test;

[0053] The energy harvesting module is used to collect several optical information generated by the device under test in the interference test environment, and to couple the several optical information.

[0054] The spectral processing module is used to perform spectral separation on the coupled optical information, acquire light intensity data, and synthesize it with the target center wavelength to obtain target light intensity information;

[0055] The signal conversion module is configured to perform signal conversion on the target light intensity information to obtain a light intensity digital signal.

[0056] The fault diagnosis module is configured to extract multi-source heterogeneous optical feature information based on the light intensity digital signal to perform fault recognition and output a fault recognition result.

[0057] Specifically, the radiation interference module comprises an electric wave bearing device and an electric field generating device.

[0058] The electric wave bearing device is configured to bear a semi-electric wave darkroom of the device under test to shield the interference electromagnetic field in the test environment.

[0059] The electric field generating device is configured to generate an interference electromagnetic wave and apply the interference electromagnetic wave to the device under test.

[0060] In this embodiment, the semi-electric wave darkroom is configured to shield the interference electromagnetic field to prevent the interference electromagnetic field from affecting the external environment (including the device and the test personnel).

[0061] The electric field generating device comprises at least a signal generator, a power amplifier and a radio frequency cable.

[0062] The signal generator is configured to generate a target radio frequency signal.

[0063] The power amplifier is configured to transmit the target radio frequency signal after amplification to the radio frequency cable, and generate a standard interference signal corresponding to the target radio frequency signal through an antenna group corresponding to the radio frequency cable.

[0064] Specifically, the signal generator comprises a key configuration unit configured to generate a target radio frequency signal.

[0065] The key configuration unit is configured to automatically control key parameters through an upper computer and control software, wherein the key parameters at least include an output frequency, a level and a modulation parameter.

[0066] As an implementation, the radiation interference module realizes automatic control on the electric field generating device through the host computer and control software; for example, the test product with multiple indicator lights is placed in a semi-electric wave darkroom, a signal source generates radio frequency signals in a frequency range of 80MHz-6GHz, the signals are amplified through a power amplifier and output to an antenna group through a coaxial cable, wherein a log-periodic antenna is used for 80MHz-1GHz and a horn antenna is used for 1GHz-6GHz, and then a field strength of not more than 30V / m after calibration is applied to the test product; the output frequency, level and modulation parameters of the signal generator are automatically controlled through the host computer and control software, and a standard interference signal is generated through the power amplifier and the transmitting antenna; during the test, after entering the automatic test interface through the host computer, the "vertical" or "horizontal" polarization is selected, the corresponding data of the reference calibration is called, and the full-band interference can be performed according to the frequency range, step, modulation mode, field strength and hardware settings set in the test template.

[0067] The energy collection module comprises an electromagnetic non-interference unit and an information coupling unit;

[0068] The electromagnetic non-interference unit comprises a plurality of electromagnetic non-interference channels made of electromagnetic wave non-interference material, and the visible waveband emitted by the state indicator light piece of the device under test is optically collected through the electromagnetic non-interference channels to obtain the optical information;

[0069] The information coupling unit is used for coupling the optical information and sending it to the optical spectrum processing module.

[0070] Further, the energy collection module can be made of all electromagnetic wave non-interference materials, without metal and electronic components, to ensure that the interference on the electromagnetic radiation test is as low as possible and the reliability of the test product under the radio frequency radiation interference test is improved; the energy collection module can also be flexibly placed near the state indicator light piece of the device under test to collect optical information.

[0071] As an implementation, the electromagnetic non-interference channels of the energy collection module adopt a simplified double-convex lens achromatic combination design on the optical system, realize high-flux light information collection in the visible light waveband range with a micro distance, a small aperture and a large field of view (field of view > 120°), and are matched with a multi-joint adjustable plastic support to be freely placed near the state indicator light piece of the device under test to collect light intensity information at different angles, efficiently couple the collected optical information into a large-diameter multi-mode optical fiber for information transmission, and the length of the optical fiber can reach 10-15m to facilitate lossless transmission of the information in the visible light waveband, and finally transmit the information to the corresponding module outside the darkroom, such as the optical spectrum processing module for optical spectrum separation.

[0072] The optical spectrum processing module comprises an optical spectrum separation submodule and an optical spectrum synthesis submodule;

[0073] The optical spectrum separation sub-module is configured to perform equal-ratio light splitting on the optical information, perform band-pass filtering spectrum separation on the split light, and extract light intensity data of the split light signal;

[0074] The optical spectrum synthesis sub-module is configured to synthesize the filter wavelength and the actual center wavelength of the light intensity data, and take the light intensity data after wavelength synthesis as target light intensity information.

[0075] As an implementation, the split light signal is subjected to spectrum separation by using blue, green and red band-pass filters with center wavelengths of 488 nm, 530 nm and 633 nm respectively, and the blue, green and red band-pass filtered light intensity (λa, λb, λc) is synthesized according to the filter wavelength distribution as the weight (ωa, ωb, ωc), so that the center wavelength resolution of better than 1 nm in the visible light spectrum range can be achieved, so as to more accurately identify and analyze the spectral characteristics of the input light signal.

[0076] Among them, by using band-pass filters with center wavelengths of 488 nm (blue light), 530 nm (green light) and 633 nm (red light), different spectral components in the input light signal can be accurately separated, and these filters can allow light signals in a specific wavelength range to pass through while blocking light signals of other wavelengths, thereby achieving accurate extraction of spectral components.

[0077] In this embodiment, by performing equal-ratio light splitting on the input light signal (i.e., the optical information collected by the collection module), the split optical information has the same splitting ratio and optical path length, which helps the signal conversion module to maintain stability when processing these information, and at the same time helps to optimize the fault diagnosis process, reduces the additional processing procedures and time caused by uneven splitting, and improves the fault diagnosis efficiency. By using band-pass filters of multiple wavelengths for spectrum separation, light intensity information of multiple wavebands can be extracted as the basis for fault diagnosis, and the robustness of fault diagnosis is enhanced, that is, even when the information of some wavebands is affected by interference or noise, accurate fault diagnosis can still be performed through the information of other wavebands.

[0078] The signal conversion module comprises a first converter and a second converter;

[0079] The first converter is configured to convert the optical signal sequence of the target light intensity information into a target electrical signal sequence;

[0080] The second converter is configured to perform analog-to-digital conversion on the target electrical signal sequence to obtain a light intensity digital signal.

[0081] Further, the first converter is an optoelectronic signal converter, and the second converter is an ADC converter. The optoelectronic signal converter can adopt a silicon-based photodetector to ensure photoelectric conversion of 380-780 nm, and through excellent device process, ensure that the photodetector has the characteristics of low noise and low dark current, and has excellent sensitivity when converting the optical signal into an electrical signal. Then, through a high-speed transimpedance amplification circuit with low bias current, the weak current signal can be amplified, and the wideband characteristics can be obtained to adapt to the collection and processing of high-frequency flicker optical signals. The ADC converter converts the optical signal into a high-precision digital signal, which can capture the weak intensity change and instantaneous change of the optical information.

[0082] In the embodiment, the optical signal is converted by the signal conversion module to obtain a stable signal, thereby improving the accuracy and reliability of the digital signal and providing more valuable data support for subsequent fault diagnosis.

[0083] The fault diagnosis module includes a first fault diagnosis unit and a second fault diagnosis unit.

[0084] The first fault diagnosis unit is configured to extract heterogeneous optical feature information according to the optical intensity digital signal.

[0085] The second fault diagnosis unit is configured to diagnose a fault position by taking the heterogeneous optical feature information as an input of a fault detection model.

[0086] In the embodiment, in the electromagnetic compatibility detection, most of the data is normal, and the fault data is relatively small, which is prone to overfitting, thereby reducing the generalization performance of the model. Therefore, in the fault diagnosis task, the optical information of the device when it is working normally is taken as unsupervised data, and the difference between the abnormal data and the normal data is found to discover the fault.

[0087] Embodiment 2, as shown in Figure 2 The radiation immunity fault diagnosis method based on multi-source heterogeneous observation includes steps S1-S5.

[0088] S1, constructing an interference test environment of a device to be tested;

[0089] S2, collecting optical information of the device to be tested in response to the interference test environment, and coupling the optical information to obtain an optical coupling information sequence;

[0090] S3, performing spectrum separation and target light wave synthesis on the optical coupling information sequence to obtain target optical intensity information;

[0091] S4, performing signal conversion on the target optical intensity information to obtain an optical intensity digital signal;

[0092] S5, diagnose the fault position based on the light intensity digital signal as the input of the fault detection model, and output a fault identification result.

[0093] Specifically, the step of diagnosing the fault position based on the light intensity digital signal as the input of the fault detection model comprises: performing multi-source heterogeneous optical feature extraction on the light intensity digital signal based on a long short-term memory neural network algorithm combined with an automatic encoder, and converting the multi-source heterogeneous optical features into time-based optical feature sequences.

[0094] Training the optical feature sequences to construct a fault detection model.

[0095] Based on the fault detection model, reconstructing error calculation is performed on the multi-source heterogeneous optical sample data of the to-be-tested equipment, the abnormal state of the sample is judged based on the calculation result, and the abnormal sample is integrated to identify the fault position.

[0096] As an embodiment, as shown in Figure 3 The LSTM network is composed of multiple LSTM units which work cooperatively to improve the accuracy of fault diagnosis by learning the long-term dependence in the optical information sequence data; the automatic encoder (AE) determines the optimal threshold based on the evaluation of the reconstruction loss rate of all optical information sequence data, thereby realizing the diagnosis of the fault position in the optical information sequence, and the specific steps are as follows:

[0097] A1, obtaining a plurality of multi-source heterogeneous optical data of a device during normal operation as an original data set, and processing it into a series of time sequences [X1, X2, X3,..., X n ], each sequence X contains a fixed length t time window data [x1, x2, x3,..., x t ], the length of t is determined according to the repetition frequency of the optical data, and x t represents the features at the corresponding time point t.

[0098] A2, the data is input into the encoder of the LSTM, and the features are converted into a time-based feature sequence batch. For example, t can be set to 8, and the input optical data of a single time window can be represented as an 8x1 vector, which is input into the encoder. The encoder first creates an LSTM network with 8 LSTM units in the first layer, and each LSTM unit processes one sample. The 8 LSTM units work in sequence, and the 1st LSTM unit passes the sample result to the 2nd LSTM unit, which decides whether to retain or forget the sample result of the 1st LSTM unit. If the 2nd LSTM unit decides to retain, it is written into long-term memory, and the sample information of the 1st LSTM unit and the processed sample feature information are passed to the 3rd LSTM unit. In this way, the last 1st LSTM unit, i.e., the 8th LSTM unit in the model, processes all the samples saved by the previous 7 LSTM units, and the last 1st LSTM unit outputs the information of all related samples. The output is now converted into a 1x16 vector as the encoded feature. To facilitate subsequent recovery, a replication layer is added in the 2nd layer to create a copy of the 1x16 vector equal to t. For example, when t = 8, the 2nd layer creates 8 copies of the encoded feature as a two-dimensional vector equal to 8x16.

[0099] A3, the sequence structure of the input data is recovered based on the LSTM decoder. For example, each 1x16 vector in the 8x16 two-dimensional vector represents a feature in the original time sequence, and these features are restored as the input of the LSTM decoder, as follows:

[0100] A31, a 3rd layer network containing 8 LSTM units is created, where the number of LSTM units represents the number of features in the sequence;

[0101] A32, each LSTM unit individually processes the input of a 1x16 encoded feature and outputs a result with the same vector size (i.e., 1x16), and the output 8x16 vector represents what is learned from the encoded features. A matrix multiplication is performed with a separate 1x16 convolutional layer to restore the result to a vector with a size of 8x1.

[0102] A4, the weights and parameters of the model are constantly adjusted during training data using the backpropagation strategy, so as to calculate the reconstruction loss Loss between the output and the input (as shown in Figure 4 , in this embodiment, the mean absolute error (MAE) algorithm is used as the reconstruction error loss function, and the formula is as follows:

[0103]

[0104] where n represents the total number of samples, x iis a representation of the original input to the encoder, is the output generated by the decoder.

[0105] A5, based on steps A1-A4, complete the training of the fault detection model, the optical information sequence of the device to be tested is taken as the detection sample to pass through the fault detection model, the different reconstruction error rates of all samples are obtained, the maximum reconstruction error rate is set as the threshold value, if the reconstruction error rate of the detection sample exceeds the threshold value, it is considered as an abnormal sample of fault occurrence, and the fault occurrence position is obtained by integrating the abnormal sample.

[0106] It can be understood that, through the fault detection mode of the unsupervised hybrid deep learning algorithm, the fault occurrence position can be diagnosed from the data sequence, but the corresponding interference frequency band of the fault occurrence cannot be known, therefore, in addition to the frame selection of the potential fault occurrence position, the time stamp of the fault position relative to the starting stage of the data sequence is also obtained, and the change result of the interference frequency with time is obtained, the output results of the two are compared by time stamp, and the interference frequency band value corresponding to the fault occurrence frequency band is obtained, so that the detection personnel can reproduce the corresponding frequency band problem, identify the fault type, and analyze the fault reason.

[0107] In the embodiment, by combining long short-term memory and automatic encoder, since the LSTM network is composed of multiple LSTM units, these units work cooperatively and can learn the long-term dependence relationship in the optical information sequence data, thereby improving the accuracy of fault diagnosis. Through the automatic encoder, even when processing unsupervised optical information sequence data with unpredictable data distribution, the optimal threshold value can be determined by evaluating the reconstruction loss rate of all optical information sequence data, thereby realizing the diagnosis of the fault position in the optical information sequence.

[0108] The beneficial effects of the embodiment: through the energy collection module, the visible light information collection of small caliber and large field of view is realized in optics, the module is manufactured by all electromagnetic wave undisturbed materials, and the interference on electromagnetic radiation test is reduced as much as possible. The collected optical information is coupled into the multimode optical fiber for information transmission, transmitted from the electromagnetic wave darkroom to the wavelength sensitive spectrum processing module and high bandwidth weak signal conversion module outside the darkroom, the high precision capture and demodulation of optical information such as light intensity, wavelength distribution and flicker frequency are realized. The unsupervised hybrid deep learning diagnosis module combines long short-term memory (LSTM) and automatic encoder (AE) to construct a fault detection model, learns the long-term dependence relationship in the optical information sequence data, improves the accuracy of fault diagnosis, and even when processing unsupervised optical information sequence data with unpredictable data distribution, the optimal threshold value can be determined by evaluating the reconstruction loss rate of all optical information sequence data, so as to realize the diagnosis of the fault position in the optical information sequence. Finally, the diagnosis result is compared with the time stamp of the test module frequency output result, the interference frequency band value corresponding to the fault frequency band is obtained, so as to assist the detection personnel to effectively cope with the problems such as the instantaneous change of the indicator light that cannot be captured in the experiment, the loss of weak brightness change information, the mutual interference of multiple indicator light information, and the like, and improve the efficiency, accuracy and reliability of the detection.

[0109] The above specific embodiments are the preferred embodiments of the present application, and the specific implementation range of the present application is not limited thereto. The scope of the present application includes but is not limited to the specific embodiments, and equivalent changes made in accordance with the shape, structure and method of the present application are within the scope of protection of the present application.

Claims

1. A radiation immunity fault diagnosis device based on multi-source heterogeneous observation, characterized in that: include: The radiated interference module is used to set up the interference test environment for the device under test. The energy harvesting module is used to collect several optical information generated by the device under test in the interference test environment, and to couple the several optical information. The spectral processing module is used to perform spectral separation on the coupled optical information, acquire light intensity data, and synthesize it with the target center wavelength to obtain target light intensity information; The signal conversion module is used to convert the target light intensity information into a digital light intensity signal. The fault diagnosis module is used to extract multi-source heterogeneous optical feature information based on the light intensity digital signal, so as to perform fault identification and output the fault identification result; The energy harvesting module includes an electromagnetic non-interference unit and an information coupling unit; The electromagnetic interference-free unit includes several electromagnetic interference-free channels made of electromagnetic wave interference-free materials. The electromagnetic interference-free channels are used to collect optical information in the visible band emitted by the status indicator of the device under test, and to obtain the optical information. The information coupling unit is used to couple the optical information and send it to the spectral processing module; The spectral processing module includes a spectral separation submodule and a spectral synthesis submodule; The spectral separation submodule is used to perform equal-splitting spectral processing on the optical information, perform bandpass filtering spectral separation on the spectral signal after spectral separation, and extract the light intensity data of the spectral signal. The spectral synthesis submodule is used to synthesize the filter wavelengths of the light intensity data to obtain target light intensity information; Specifically, the optical information is divided into equal parts, and a bandpass filter containing blue, green and red light is used to perform spectral separation on the divided optical information, and the light intensity data of the split signal of each channel is measured; the center wavelengths of blue, green and red light in the bandpass filter are used as weights respectively, and the light intensity data of each channel are weighted and merged with the corresponding weights to obtain the target light intensity information; The light intensity digital signal is subjected to multi-source heterogeneous optical feature extraction based on a long short-term memory neural network algorithm combined with an autoencoder, and the multi-source heterogeneous optical features are converted into a time-based optical feature sequence.

2. The radiation immunity fault diagnosis device based on multi-source heterogeneous observation according to claim 1, characterized in that: The radiation interference module includes a radio wave carrying device and an electric field generating device. The radio wave carrying device is used to receive the semi-anechoic chamber of the device under test in order to shield the interfering electromagnetic fields in the test environment. The electric field generating device is used to generate interfering electromagnetic waves and apply them to the device under test.

3. The radiation immunity fault diagnosis device based on multi-source heterogeneous observation according to claim 2, characterized in that: The electric field generating device includes at least a signal generator, a power amplifier, and radio frequency cables; The signal generator is used to produce the target radio frequency signal; The power amplifier is used to amplify the target radio frequency signal and transmit it to the radio frequency cable, and generate a standard interference signal corresponding to the target radio frequency signal through the antenna group corresponding to the radio frequency cable.

4. The radiation immunity fault diagnosis device based on multi-source heterogeneous observation according to claim 1, characterized in that: The signal conversion module includes a first converter and a second converter; The first converter is used to convert the optical signal sequence of the target light intensity information into a target electrical signal sequence; The second converter is used to perform analog-to-digital conversion on the target electrical signal sequence to obtain a digital light intensity signal.

5. The radiation immunity fault diagnosis device based on multi-source heterogeneous observation according to claim 4, characterized in that: The fault diagnosis module includes a first fault diagnosis unit and a second fault diagnosis unit; The first fault diagnosis unit is used to extract heterogeneous optical feature information based on the light intensity digital signal; The second fault diagnosis unit is used to diagnose the fault location by using the heterogeneous optical feature information as input to the fault detection model.

6. The radiation immunity fault diagnosis device based on multi-source heterogeneous observation according to claim 3, characterized in that: The signal generator includes a key configuration unit for producing the target radio frequency signal; The key configuration unit uses a host computer and control software to automatically control key parameters, which include at least output frequency, level and modulation parameters.

7. A radiated immunity fault diagnosis method based on multi-source heterogeneous observation, applicable to the radiated immunity fault diagnosis device based on multi-source heterogeneous observation as described in any one of claims 1-6, characterized in that: Includes the following steps: S1. Construct the interference test environment for the device under test; S2. In response to the interference test environment, acquire the optical information of the device under test, and perform optical intensity coupling on the optical information to obtain an optical coupling information sequence; S3. Perform spectral separation and target light wave synthesis on the optical coupling information sequence to obtain target light intensity information; S4. Perform signal conversion on the target light intensity information to obtain a digital light intensity signal; S5. Extract multi-source heterogeneous optical feature information based on the aforementioned digital light intensity signal, use it as input to the fault detection model for fault location diagnosis, and output the fault identification result; specifically including: The light intensity digital signal is subjected to multi-source heterogeneous optical feature extraction based on a long short-term memory neural network algorithm combined with an autoencoder, and the multi-source heterogeneous optical features are converted into a time-based optical feature sequence. The optical feature sequence is used to train a fault detection model; Based on the fault detection model, the reconstruction error of the multi-source heterogeneous optical sample data of the device under test is calculated. Based on the calculation results, the abnormal state of the sample is determined, and the abnormal sample is integrated to identify the fault location.

Images