Aircraft arc fault detection method and device

Abstract

The invention belongs to the technical field of aviation power supply and distribution systems and electrical safety detection, and provides an aircraft arc fault detection method and device which are used for improving the interference immunity of an algorithm, improving the accuracy of arc detection and reducing the false alarm rate. The method comprises the following steps: S1, a data acquisition module acquires analog quantity information in a power supply and distribution network main circuit in real time, and transmits analog data to a data conditioning module; s2, a data conditioning module conditions the analog data and transmits the conditioned data to an arc detection processor; and S3, the arc detection processor detects whether an arc fault exists in the power supply and distribution network in real time through a built-in arc detection algorithm. The method is used for arc fault identification.

Description

Technical Field

[0001] The present invention belongs to the field of aviation power supply and distribution system and electrical safety detection technology, and in particular relates to an aircraft arc fault detection method and device. Background Art

[0002] Arc is a gas ionization discharge phenomenon. Under normal conditions, the gas has good insulation and no arc is generated. However, when the electric field at both ends of the gas gap is strong enough, the air or its insulating medium is ionized, resulting in a continuous discharge phenomenon, which is the arc. The arc phenomenon will form obvious light and high temperature, and the temperature can reach tens of thousands of degrees Celsius, which is very harmful to the surrounding electrical equipment. In particular, it poses a great threat to the safe flight of aircraft, so arc detection and protection are needed.

[0003] In existing arc fault identification, features are often extracted from arc data and a fixed feature threshold is set for arc detection. This protection strategy has a poor effect on the complex electrical network in the aircraft and the false detection rate remains high. Summary of the invention

[0004] The purpose of the present invention is to provide an aircraft arc fault detection method and device, so as to improve the anti-interference ability of the algorithm, improve the accuracy of arc detection, and reduce the false alarm rate.

[0005] The technical solution of the present invention:

[0006] An aircraft arc fault detection device, comprising:

[0007] Data acquisition module 1, data conditioning module 2, data storage module 3, arc detection processor 4, communication module 5 and action module 6;

[0008] The data acquisition module 1 collects analog quantity information in the main circuit of the power supply and distribution network in real time, and transmits the analog quantity information to the data conditioning module 2;

[0009] The data conditioning module 2 conditions the analog information and transmits the conditioned data to the arc detection processor 4;

[0010] The arc detection processor 4 detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm;

[0011] The arc detection processor 4 controls the action module 6 through control instructions, thereby realizing on-off control of the main circuit of the power supply and distribution network.

[0012] The arc detection processor 4 communicates with the external device through the communication module 5, receives the communication command of the external device, and reports its own status information.

[0013] The status information reported by the communication module 5 includes fault information.

[0014] The analog information collected by the data collection module 1 is current data, voltage data or combined data of current and voltage in the power supply and distribution network.

[0015] The data acquisition module 1 acquires analog quantity information in the following manners: acquiring analog quantity information through a voltage sensor, a current sensor or a sampling resistor.

[0016] The data conditioning module 2 performs analog information conditioning including amplification and filtering.

[0017] The arc detection processor 4 collects the analog data transmitted by the data processing module in real time and periodically at a specific sampling frequency by means of ADC sampling.

[0018] The data storage module 3 receives arc information from the arc detection processor 4 so as to be called by the arc detection processor 4 .

[0019] A method for detecting arc faults in an aircraft, comprising:

[0020] S1, the data acquisition module collects analog information in the main circuit of the power supply and distribution network in real time, and transmits the analog data to the data conditioning module;

[0021] S2, the data conditioning module conditions the analog data and transmits the conditioned data to the arc detection processor;

[0022] S3, the arc detection processor detects in real time whether there is an arc fault in the power supply and distribution network through the built-in arc detection algorithm.

[0023] Step S3 includes:

[0024] (1) The arc detection processor collects analog data transmitted by the data processing module in real time at a specific sampling frequency and converts the analog data into periodic discrete data;

[0025] (2) The arc detection processor performs a data validity check on the periodic discrete data to verify the validity of the data. If the data is valid, it proceeds to step (3). If the data is invalid, it proceeds to the next period of data collection and processing;

[0026] (3) After the arc detection processor verifies that the periodic discrete data is valid, it performs data preprocessing on the periodic discrete data to filter out interference information in the data;

[0027] (4) After the arc detection processor preprocesses the periodic discrete data, it performs feature extraction on the periodic discrete data and extracts a periodic feature vector matrix;

[0028] (5) The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier for arc fault determination.

[0029] Step S3 also includes:

[0030] (6) If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence:

[0031] (i) issuing control instructions to the action mechanism to control the action mechanism to perform arc protection action;

[0032] (ii) communicating with external equipment through a communication module to report arc fault information;

[0033] (iii) Storing arc fault information in a data storage module.

[0034] S4, the arc detection processor controls the action module through control instructions to achieve on-off control of the main circuit of the power supply and distribution network;

[0035] S5, the arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own status information.

[0036] Step (2) The arc detection processor performs data validity verification on the periodic discrete data using a weak signal detection method, as follows:

[0037] (21) Determine periodic discrete data [M 1 ,M n ] whether each point exceeds the set weak signal filtering threshold M C ;

[0038] (22) More than M in the statistical period C The number of weak signal count variables is counted as: M ;

[0039] (23) Determine the weak signal count variable Count M Is it greater than the weak signal count threshold Count? MF If it exceeds, the cycle is considered not to be a weak cycle and the process goes to step (3); if it does not exceed, the cycle is considered to be a weak cycle and the process goes directly to the next cycle analog data processing.

[0040] Step (3) Data preprocessing uses the Kalman filter method to reconstruct and smooth the data, and then extract the periodic eigenvector matrix after data reconstruction.

[0041] The arc fault classifier uses a pre-trained extreme learning machine, and the training method is as follows:

[0042] 1. Set the input layer, hidden layer, and output layer nodes of the neural network, and set the learning rate and activation function;

[0043] 2. Perform feature extraction through the pre-collected arc data, normal data, load data and interference data, and extract the feature vector matrix [T 1 ,T 2 ,T 3 ] the corresponding eigenvector matrices of the four types of data are obtained, and the corresponding analog arc eigenvector matrix, analog normal eigenvector matrix, load analog eigenvector matrix and interference analog eigenvector matrix are obtained. The array composed of the four types of eigenvector matrices is used as a training set for neural network training, and the output results of the neural network are arc fault and non-arc fault.

[0044] Beneficial effects of the present invention:

[0045] This scheme filters the interference data in the analog quantity through data validity verification and Kalman filtering method, which improves the anti-interference ability of the algorithm, improves the accuracy of arc detection, and reduces the false alarm rate.

[0046] The arc detection terminal adopts an intelligent algorithm and does not need to set the arc feature threshold, which avoids the restriction of a single arc feature threshold on the arc detection rate, reduces the false alarm rate and improves the detection rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings are still within the scope of the present invention.

[0048] Figure 1 It is a composition diagram of an aircraft arc fault detection device;

[0049] Figure 2 is an identification flow chart of an aircraft arc fault detection method;

[0050] Figure 3 This is a composition diagram of an aircraft arc fault detection device according to Embodiment 1 of the present invention;

[0051] Figure 4 This is an identification flow chart of an aircraft arc fault detection method according to Embodiment 1 of the present invention.

[0052] Figure 5 is a composition diagram of an aircraft arc fault detection device according to a second embodiment of the present invention;

[0053] Figure 6 This is an identification flow chart of an aircraft arc fault detection method according to Embodiment 3 of the present invention.

[0054] Figure 7 1 is a composition diagram of an aircraft arc fault detection device according to a fourth embodiment of the present invention;

[0055] Figure 8 This is an identification flow chart of an aircraft arc fault detection method according to a fourth embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] The basic principle of the technical solution of the device of the present invention is as follows Figure 1 shown.

[0058] An aircraft arc fault detection device, comprising:

[0059] Data acquisition module 1, data conditioning module 2, data storage module 3, arc detection processor 4, communication module 5 and action module 6;

[0060] The data acquisition module 1 collects analog quantity information in the main circuit of the power supply and distribution network in real time, and transmits the analog quantity information to the data conditioning module 2;

[0061] The data conditioning module 2 conditions the analog information and transmits the conditioned data to the arc detection processor 4;

[0062] The arc detection processor 4 detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm; when an arc fault is detected, the arc detection processor controls the action module 6 to implement arc protection, and at the same time stores the arc information through the data storage module 3 and reports the arc information through the communication module 5.

[0063] The basic principle of the technical solution of the method of the present invention is as follows Figure 2 shown.

[0064] A method for detecting arc faults in an aircraft, comprising:

[0065] S1, the data acquisition module collects analog information in the main circuit of the power supply and distribution network in real time, and transmits the analog data to the data conditioning module;

[0066] S2, the data conditioning module conditions the analog data and transmits the conditioned data to the arc detection processor;

[0067] S3, the arc detection processor detects in real time whether there is an arc fault in the power supply and distribution network through the built-in arc detection algorithm.

[0068] S4, the arc detection processor controls the action module through control instructions to achieve on-off control of the main circuit of the power supply and distribution network;

[0069] S5, the arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own status information.

[0070] The arc detection algorithm embedded in the arc detection processor collects analog information in the circuit in real time, performs data validity verification and data preprocessing on the collected analog data, performs feature extraction after data processing, extracts multiple periodic features to form a feature vector matrix, and injects the feature vector matrix into a pre-trained arc fault classifier to realize arc fault detection.

[0071] Embodiment 1:

[0072] In order to verify the feasibility of the present invention, the specific example provided by the present invention takes the arc data collected under the power supply of 270V DC as the input signal, according to Figure 3 and Figure 4 The identification process is processed

[0073] The specific implementation steps are as follows:

[0074] Step 1: The data acquisition module collects the current data in the main circuit of the power supply and distribution network in real time, and transmits the periodic current data to the data conditioning module.

[0075] Step 2: The data conditioning module conditions the current data, amplifies the current data, performs 1KHz-100KHz bandpass filtering, and transmits the conditioned data to the arc detection processor.

[0076] Step 3: The arc detection processor detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm.

[0077] The algorithm processing steps are as follows:

[0078] The arc detection processor collects the current data transmitted by the data processing module in real time at a specific sampling frequency and converts it into periodic current data. The sampling frequency here is 200KHz, which is related to the frequency range of the bandpass filter and is twice the highest frequency band of the bandpass filter.

[0079] The arc detection processor performs a data validity check on the periodic current data to check whether the data is a weak signal. The collected current signal is compared with the filter threshold 1. When the current data is greater than the filter threshold 1, the sampling filter over-threshold count accumulation Ncount is accumulated. In a sampling cycle, if the filter over-threshold count accumulation Ncount exceeds the filter threshold 2, it is considered that the collected current data is not a weak signal, the data is valid, and the data is cached. Otherwise, the periodic current data is considered to be a weak signal, which is invalid data and the periodic data is discarded, and the next periodic current data is collected again.

[0080] If the data is valid, the next step of data preprocessing is entered, and the periodic current data is filtered and reconstructed using Kalman filtering to filter the interference information in the periodic current data.

[0081] After preprocessing the periodic current data, the arc detection processor performs feature extraction on the periodic current data, extracts three features: the mean amplitude of the periodic current frequency band, the combination index of the number of periodic current amplitudes and spectrum energy, and the periodic current discrete data index, and forms a periodic current feature vector matrix.

[0082] The mean amplitude of the periodic current frequency band refers to the mean amplitude of the current in a specific frequency band. Here, the mean amplitude of 1KHz-100KHz is extracted.

[0083] Specifically: for periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0084] Select a specific frequency range [f i1 ,f i2 ] form a matrix [FM i1 ,...,FM i2 ], the frequency range can be [1KHz, 100KHz] or [1KHz, 50KHz], and can also be determined based on the filter frequency band range.

[0085] For the matrix [FM i1 ,...,FM i2 ] to obtain the analog value of the specific frequency band and Sum FM .

[0086] Matrix [FM i1 ,...,FM i2 ]The number of elements in is: Number Fm , then the characteristic point T of this cycle 1 Expressed as: T 1 =SumFM / Number Fm .

[0087] The combination of the number of periodic current amplitudes and spectrum energy refers to extracting the spectrum of a specific frequency band, dividing it into several intervals, and calculating the amplitude mean of several intervals. Further judgment is made on the number of intervals whose amplitude mean exceeds a specific threshold, which is used as the characteristic point of this cycle. The details are as follows:

[0088] Specifically: Specifically: For the periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0089] Select the amplitude within a specific frequency range [1KHz, 100kHz] to form a matrix [FM i1 ,...,FM i2 ]

[0090] Divide the specific frequency range [1KHz, 100kHz] into 10 intervals [1kHz, 10KHz], [10KHz, 20KHz], ... [90kHz, 100KHz], and calculate the amplitude and mean value in each interval, which is [FMK 1 ,FMK 10 ].

[0091] Statistics [FMK 1 ,FMK 2 ] exceeds the interval amplitude and the number of threshold FMCs, counted as T 2 .

[0092] Cycle analog characteristic T 3 It is a discrete data indicator and is calculated as follows:

[0093] T 3 = (cycle current standard deviation) / cycle current mean

[0094] The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier ELM for arc fault determination. ELM is a pre-trained extreme learning machine, and the setting method is as follows:

[0095] Set the input layer, hidden layer, and output layer nodes of the neural network, and set parameters such as the learning rate and activation function. The number of neurons in the input layer is the number of elements in the eigenvector matrix, which can be set to 3 here. The number of neurons in the output layer is the number of fault classifications, which are arc and non-arc here, so the number of neurons in the output layer is set to 2.

[0096] (2) Feature extraction is performed through the pre-collected arc current data, normal current data, load current data and interference current data, and the current feature vector matrix [T 1 ,T 2 ,T 3 ] the corresponding eigenvector matrices of the four types of data are obtained, and the corresponding current arc eigenvector matrix, current normal eigenvector matrix, load current eigenvector matrix and interference current eigenvector matrix are obtained. The array composed of the four types of eigenvector matrices is used as a training set for neural network training, and the output results of the neural network are arc fault and non-arc fault.

[0097] If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence:

[0098] Issue control instructions to the action mechanism to control the action mechanism to perform arc protection actions.

[0099] Communicate with external devices through the communication module to report arc fault information.

[0100] The arc fault information is stored in the data storage module.

[0101] Step 4: The arc detection processor controls the action module through control instructions to achieve on-off control of the main circuit of the power supply and distribution network.

[0102] Step 5: The arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own fault information and other status information.

[0103] Embodiment 2:

[0104] In order to verify the feasibility of the present invention, the specific example provided by the present invention uses the arc data collected under the power supply of high-voltage DC power supply as the input signal, according to Figure 5 The identification process is processed

[0105] The specific implementation steps are as follows:

[0106] Step 1: The data acquisition module collects the current data of the branch where the arc detection device is located in real time through the sampling resistor, and converts the periodic current data into periodic voltage data and transmits it to the data conditioning module.

[0107] Step 2: The data conditioning module conditions the periodic voltage data through its own data conditioning circuit, scales the periodic voltage data value to an appropriate size, and then transmits it to the filter circuit. The filter circuit filters the periodic voltage data and selects the periodic voltage data with a frequency range of 1KHz-100KHz to transmit to the arc detection processor.

[0108] Step 3: The arc detection processor detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm.

[0109] The algorithm processing steps are as follows:

[0110] The arc detection processor collects the periodic voltage data transmitted by the data conditioning module in real time at a specific sampling frequency, converts it into periodic ADC data, and maps the periodic ADC data into periodic current data. The sampling frequency here is 200KHz, which is related to the frequency band range of the bandpass filter, and the sampling frequency is twice the highest frequency band of the bandpass filter.

[0111] The arc detection processor performs a data validity check on the periodic current data to check whether the data is a weak signal. The collected current signal is compared with the filter threshold 1. When the current data is greater than the filter threshold 1, the sampling filter over-threshold count accumulation Ncount is accumulated. In a sampling cycle, if the filter over-threshold count accumulation Ncount exceeds the filter threshold 2, it is considered that the collected current data is not a weak signal, the data is valid, and the data is cached. Otherwise, the periodic current data is considered to be a weak signal, which is invalid data and the periodic data is discarded, and the next periodic current data is collected again.

[0112] If the data is valid, the next step of data preprocessing is entered, and the periodic current data is filtered and reconstructed using Kalman filtering to filter the interference information in the periodic current data.

[0113] After preprocessing the periodic current data, the arc detection processor performs feature extraction on the periodic current data, extracts three features: the mean amplitude of the periodic current frequency band, the combination index of the number of periodic current amplitudes and spectrum energy, and the periodic current discrete data index, and forms a periodic current feature vector matrix.

[0114] The mean amplitude of the periodic current frequency band refers to the mean amplitude of the current in a specific frequency band. Here, the mean amplitude of 1KHz-100KHz is extracted.

[0115] Specifically: for periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0116] Select a specific frequency range [f i1 ,f i2 ] form a matrix [FM i1 ,...,FM i2], the frequency range can be [1KHz, 100KHz] or [1KHz, 50KHz], and can also be determined based on the filter frequency band range.

[0117] For the matrix [FM i1 ,...,FM i2 ] to obtain the analog value of the specific frequency band and Sum FM .

[0118] Matrix [FM i1 ,...,FM i2 ]The number of elements in is: Number Fm , then the characteristic point T of this cycle 1 Expressed as: T 1 =Sum FM / Number Fm .

[0119] The combination of the number of periodic current amplitudes and spectrum energy refers to extracting the spectrum of a specific frequency band, dividing it into several intervals, and calculating the amplitude mean of several intervals. Further judgment is made on the number of intervals whose amplitude mean exceeds a specific threshold, which is used as the characteristic point of this cycle. The details are as follows:

[0120] Specifically: Specifically: For the periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0121] Select the amplitude within a specific frequency range [1KHz, 100kHz] to form a matrix [FM i1 ,...,FM i2 ]

[0122] Divide the specific frequency range [1KHz, 100kHz] into 10 intervals [1kHz, 10KHz], [10KHz, 20KHz], ... [90kHz, 100KHz], and calculate the amplitude and mean value in each interval, which is [FMK 1 ,FMK 10 ].

[0123] Statistics [FMK 1 ,FMK 2 ] exceeds the interval amplitude and the number of threshold FMCs, counted as T 2 .

[0124] Cycle analog characteristic T 3 It is a discrete data indicator and is calculated as follows:

[0125] T3 = (cycle current standard deviation) / cycle current mean

[0126] The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier ELM for arc fault determination. ELM is a pre-trained extreme learning machine, and the setting method is as follows:

[0127] Set the input layer, hidden layer, and output layer nodes of the neural network, and set parameters such as the learning rate and activation function. The number of neurons in the input layer is the number of elements in the eigenvector matrix, which can be set to 3 here. The number of neurons in the output layer is the number of fault classifications, which are arc and non-arc here, so the number of neurons in the output layer is set to 2.

[0128] (2) Feature extraction is performed using the pre-collected arc current data, normal current data, load current data, and interference current data. The corresponding feature vector matrices of the four types of data of the current feature vector matrix [T1, T2, T3] are extracted respectively to obtain the corresponding current arc feature vector matrix, current normal feature vector matrix, load current feature vector matrix, and interference current feature vector matrix. The array composed of the four types of feature vector matrices is used as a training set for neural network training. The output results of the neural network are arc fault and non-arc fault.

[0129] If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence:

[0130] (i) Issue control instructions to the action mechanism to control the action mechanism to perform arc protection action.

[0131] (ii) Communicate with external devices through the communication module to report arc fault information.

[0132] (iii) Storing arc fault information in a data storage module.

[0133] Step 4: The arc detection processor controls the drive circuit of the action module through control instructions. The drive circuit drives the MOSFET to switch on and off, thereby realizing the on-off control of the main circuit of the power supply and distribution network.

[0134] Step 5: The arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own fault information and other status information.

[0135] Embodiment three:

[0136] In order to verify the feasibility of the present invention, the specific example provided by the present invention uses the arc data collected under the power supply of high-voltage DC power supply as the input signal, according to Figure 6 The identification process is processed

[0137] The device can simultaneously detect arc faults in multi-channel branches in the power supply and distribution network and provide protection.

[0138] The specific implementation steps are as follows:

[0139] Step 1: The data acquisition module collects analog information of multiple channels in the main circuit of the power supply and distribution network in real time through a multi-channel data acquisition channel, and transmits the multi-channel periodic analog information to the data conditioning module.

[0140] Step 2: The data conditioning module conditions the received multi-channel periodic analog information through the multi-channel conditioning channel, amplifies the multi-channel analog information, performs 1KHz-100KHz bandpass filtering, and transmits the conditioned multi-channel periodic analog information to the arc detection processor.

[0141] Step 3: The arc detection processor uses a built-in arc detection algorithm to detect in real time whether there are arc faults in multiple branches in the power supply and distribution network.

[0142] The algorithm processing steps are as follows:

[0143] The arc detection processor collects the multi-channel periodic analog information transmitted by the data processing module in real time at a specific sampling frequency and converts it into multi-channel periodic discrete data. The sampling frequency here is 200KHz, which is related to the frequency band range of the bandpass filter and is twice the highest frequency band of the bandpass filter.

[0144] The arc detection processor performs the following operations on each cycle of discrete data:

[0145] (1) The arc detection processor verifies the validity of the discrete data to see if the data is a weak signal. The collected discrete signal is compared with the filter threshold 1. When the discrete data is greater than the filter threshold 1, the sampled filter over-threshold count accumulation Ncount is accumulated. Within a sampling cycle, if the filter over-threshold count accumulation Ncount exceeds the filter threshold 2, the collected discrete data is considered not to be a weak signal, the data is valid, and the data is cached. Otherwise, the discrete data of this cycle is considered to be a weak signal, which is invalid data and is discarded. The next cycle of discrete data is collected again.

[0146] (2) If the data is valid, the next step of data preprocessing is entered, and the periodic discrete data is filtered and reconstructed using Kalman filtering to filter the interference information in the periodic discrete data.

[0147] (3) After the arc detection processor preprocesses each periodic discrete data, it performs feature extraction on the periodic discrete data, extracts three features: the mean amplitude of the periodic frequency band, the index combining the number of periodic amplitudes with the spectrum energy, and the periodic discrete data index, and forms a periodic feature vector matrix.

[0148] (4) The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier ELM for arc fault determination. ELM is a pre-trained extreme learning machine, and the setting method is as follows:

[0149] Set the input layer, hidden layer, and output layer nodes of the neural network, and set parameters such as the learning rate and activation function. The number of neurons in the input layer is the number of elements in the eigenvector matrix, which can be set to 3 here. The number of neurons in the output layer is the number of fault classifications, which are arc and non-arc here, so the number of neurons in the output layer is set to 2.

[0150] (2) Feature extraction is performed through the pre-collected arc data, normal data, load data and interference data, and the feature vector matrix [T 1 ,T 2 ,T 3 ] the corresponding eigenvector matrices of the four types of data are obtained, and the corresponding arc eigenvector matrix, normal eigenvector matrix, load eigenvector matrix and interference eigenvector matrix are obtained. The array composed of the four types of eigenvector matrices is used as a training set for neural network training, and the output results of the neural network are arc fault and non-arc fault.

[0151] If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence:

[0152] Issue control instructions to the action mechanism to control the action channel corresponding to the action module to perform arc protection action.

[0153] Communicate with external devices through the CAN communication module to report arc fault information.

[0154] The arc fault information is stored in the data storage module.

[0155] Step 4: The arc detection processor controls the action module through control instructions, and controls the different action channels in the action module to achieve on-off control of multiple branch circuits of the power supply and distribution network.

[0156] Step 5: The arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own fault information and other status information.

[0157] Embodiment 4:

[0158] In order to verify the feasibility of the present invention, the specific example provided by the present invention takes the arc data collected under the power supply of 270V DC as the input signal, according to Figure 7 and Figure 8 The identification process is processed

[0159] The specific implementation steps are as follows:

[0160] Step 1: The data acquisition module collects the current data and voltage data in the main circuit of the power supply and distribution network in real time, and transmits the periodic current data and periodic voltage data to the data conditioning module.

[0161] Step 2: The data conditioning module conditions the periodic current data and the periodic voltage data, amplifies the current data and the periodic voltage data respectively, performs 1KHz-100KHz bandpass filtering, and transmits the conditioned data to the arc detection processor.

[0162] Step 3: The arc detection processor detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm.

[0163] The algorithm processing steps are as follows:

[0164] The arc detection processor collects the current data and current data transmitted by the data processing module in real time at a specific sampling frequency, and converts them into periodic current discrete data and periodic voltage discrete data. The sampling frequency here is 200KHz, which is related to the frequency band range of the bandpass filter, and the sampling frequency is twice the highest frequency band of the bandpass filter.

[0165] The arc detection processor performs a data validity check on the periodic current data to check whether the data is a weak signal. The collected current signal is compared with the filter threshold 1. When the current data is greater than the filter threshold 1, the sampling filter over-threshold count accumulation Ncount is accumulated. Within a sampling cycle, if the filter over-threshold count accumulation Ncount exceeds the filter threshold 2, it is considered that the collected current data is not a weak signal, the data is valid, and the data is cached. Otherwise, the periodic current data is considered to be a weak signal, which is invalid data and the periodic data is discarded, and the next periodic current data and periodic current data are re-collected.

[0166] If the data is valid, the next step of data preprocessing is entered, and the periodic current data and the periodic voltage data are filtered and reconstructed using Kalman filtering to filter the interference information in the periodic current data and the periodic voltage data.

[0167] After the arc detection processor preprocesses the periodic current data, it extracts features from the periodic current data and periodic voltage data, extracting three features: the mean value of the periodic current frequency band amplitude, the combination index of the periodic current amplitude number and spectrum energy, and the periodic current discrete data index [T 1 ,T 2 ,T 3 ] and the mean value of the periodic voltage frequency band amplitude, the combination index of the periodic voltage amplitude number and spectrum energy, and the periodic voltage discrete data index [T 4 ,T 5 ,T 6 ], the features are combined into a periodic eigenvector matrix [T 1 ,T 2 ,T 3 ,T 4 ,T 5 ,T 6 ].

[0168] The feature extraction methods of periodic voltage and periodic current are the same, and the feature extraction of periodic current is taken as an example for explanation.

[0169] The mean amplitude of the periodic current frequency band refers to the mean amplitude of the current in a specific frequency band. Here, the mean amplitude of 1KHz-100KHz is extracted.

[0170] Specifically: for periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0171] Select a specific frequency range [f i1 ,f i2 ] form a matrix [FM i1 ,...,FM i2 ], the frequency range can be [1KHz, 100KHz] or [1KHz, 50KHz], and can also be determined based on the filter frequency band range.

[0172] For the matrix [FM i1 ,...,FM i2 ] to obtain the analog value of the specific frequency band and Sum FM .

[0173] Matrix [FM i1 ,...,FM i2 ]The number of elements in is: Number Fm , then the characteristic point T of this cycle 1 Expressed as: T 1 =Sum FM / NumberFm .

[0174] The combination of the number of periodic current amplitudes and spectrum energy refers to extracting the spectrum of a specific frequency band, dividing it into several intervals, and calculating the amplitude mean of several intervals. Further judgment is made on the number of intervals whose amplitude mean exceeds a specific threshold, which is used as the characteristic point of this cycle. The details are as follows:

[0175] Specifically: Specifically: For the periodic current [M 1 ,M n ], and Fourier transform is performed to obtain [FM 1 ,...,FM 2 ].

[0176] Select the amplitude within a specific frequency range [1KHz, 100kHz] to form a matrix [FM i1 ,...,FM i2 ]

[0177] Divide the specific frequency range [1KHz, 100kHz] into 10 intervals [1kHz, 10KHz], [10KHz, 20KHz], ... [90kHz, 100KHz], and calculate the amplitude and mean value in each interval, which is [FMK 1 ,FMK 10 ].

[0178] Statistics [FMK 1 ,FMK 2 ] exceeds the interval amplitude and the number of threshold FMCs, counted as T 2 .

[0179] Cycle analog characteristic T 3 It is a discrete data indicator and is calculated as follows:

[0180] T 3 = (cycle current standard deviation) / cycle current mean

[0181] The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier ELM for arc fault determination. ELM is a pre-trained extreme learning machine, and the setting method is as follows:

[0182] Set the input layer, hidden layer, and output layer nodes of the neural network, and set parameters such as the learning rate and activation function. The number of neurons in the input layer is the number of elements in the eigenvector matrix, which can be set to 6 here. The number of neurons in the output layer is the number of fault classifications, which are arc and non-arc here, so the number of neurons in the output layer is set to 2.

[0183] (2) Feature extraction is performed through the pre-collected arc current data, normal current data, load current data, interference current data, arc voltage data, normal voltage data, load voltage data and interference voltage data, and the corresponding eigenvector matrices of the four types of data of the eigenvector matrix [T1, T2, T3, T4, T5, T6] are extracted respectively to obtain the corresponding arc eigenvector matrix, normal eigenvector matrix, load eigenvector matrix and interference eigenvector matrix. The array composed of the four types of eigenvector matrices is used as a training set for neural network training, and the output results of the neural network are arc fault and non-arc fault.

[0184] If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence:

[0185] Issue control instructions to the action mechanism to control the action mechanism to perform arc protection actions.

[0186] Communicate with external devices through the communication module to report arc fault information.

[0187] The arc fault information is stored in the data storage module.

[0188] Step 4: The arc detection processor controls the action module through control instructions to achieve on-off control of the main circuit of the power supply and distribution network.

[0189] Step 5: The arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own fault information and other status information.

[0190] The above is only a specific embodiment of the present invention, and the present invention is described in detail. The unexplained part is a conventional technology. However, the protection scope of the present invention is not limited to this. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An aircraft arc fault detection device, characterized in that: include: A data acquisition module (1), a data conditioning module (2), a data storage module (3), an arc detection processor (4), a communication module (5) and an action module (6); The data acquisition module (1) acquires analog quantity information in the main circuit of the power supply and distribution network in real time, and transmits the analog quantity information to the data conditioning module (2); The data conditioning module (2) conditions the analog quantity information and transmits the conditioned data to the arc detection processor (4); The arc detection processor (4) detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm; The arc detection processor (4) controls the action module (6) through control instructions, thereby realizing on-off control of the main circuit of the power supply and distribution network; The arc detection processor (4) communicates with an external device via a communication module (5), receives communication commands from the external device, and reports its own status information.

2. An aircraft arc fault detection device as claimed in claim 1, characterized in that: The status information reported by the communication module (5) includes fault information.

3. The aircraft arc fault detection device according to claim 1, characterized in that: The analog information collected by the data collection module (1) is current data, voltage data or combined data of current and voltage in the power supply and distribution network.

4. The aircraft arc fault detection device according to claim 1, characterized in that: The data acquisition module (1) acquires analog quantity information in the following manners: acquiring analog quantity information through a voltage sensor, a current sensor or a sampling resistor.

5. The aircraft arc fault detection device according to claim 1, characterized in that: The data conditioning module (2) conditions the analog information including amplification and filtering.

6. The aircraft arc fault detection device according to claim 1, characterized in that: The arc detection processor (4) collects analog data transmitted by the data processing module in a real-time periodic manner at a specific sampling frequency by means of ADC sampling.

7. The aircraft arc fault detection device according to claim 1, characterized in that: Also includes: The data storage module (3) receives arc information from the arc detection processor (4) for calling by the arc detection processor (4).

8. A method for detecting arc faults in an aircraft, characterized in that: include: S1, the data acquisition module (1) collects analog quantity information in the main circuit of the power supply and distribution network in real time, and transmits the analog data to the data conditioning module (2); S2, the data conditioning module (2) conditions the analog data and transmits the conditioned data to the arc detection processor (4); S3, the arc detection processor (4) detects in real time whether there is an arc fault in the power supply and distribution network through a built-in arc detection algorithm.

9. The method for detecting an arc fault in an aircraft according to claim 8, characterized in that: Step S3 includes: (1) The arc detection processor collects analog data transmitted by the data processing module in real time at a specific sampling frequency and converts the analog data into periodic discrete data; (2) The arc detection processor performs a data validity check on the periodic discrete data to verify the validity of the data. If the data is valid, it proceeds to step (3). If the data is invalid, it proceeds to the next period of data collection and processing; (3) After the arc detection processor verifies that the periodic discrete data is valid, it performs data preprocessing on the periodic discrete data to filter out interference information in the data; (4) After the arc detection processor preprocesses the periodic discrete data, it performs feature extraction on the periodic discrete data and extracts a periodic feature vector matrix; (5) The arc detection processor inputs the extracted periodic feature vector matrix into the arc fault classifier for arc fault determination.

10. The method for detecting an arc fault in an aircraft according to claim 9, characterized in that: Step S3 also includes: (6) If the arc fault classifier detects an arc fault, the arc detection processor performs the following operations in sequence: (i) issuing control instructions to the action mechanism to control the action mechanism to perform arc protection action; (ii) communicating with external equipment through a communication module to report arc fault information; (iii) Storing arc fault information in a data storage module.

11. The method for detecting arc faults in aircraft according to claim 8, characterized in that: Also includes: S4, the arc detection processor controls the action module through control instructions to achieve on-off control of the main circuit of the power supply and distribution network; S5, the arc detection processor communicates with the external device through the communication module, receives the communication command of the external device, and reports its own status information.

12. The method for detecting arc faults in aircraft according to claim 9, characterized in that: Step (2) The arc detection processor performs data validity verification on the periodic discrete data using a weak signal detection method, as follows: (21) Determine periodic discrete data [M1,Mn] Whether each point exceeds the set weak signal filtering threshold M C ; (22) During the statistical period, MC The number of weak signal count variables is counted as: CountM ; (23) Determine the weak signal counting variable CountM Is it greater than the weak signal counting threshold? CountMF If it exceeds, the cycle is considered not to be a weak cycle and the process goes to step (3); if it does not exceed, the cycle is considered to be a weak cycle and the process goes directly to the next cycle analog data processing.

13. The method for detecting arc faults in aircraft according to claim 9, characterized in that: Step (3) Data preprocessing uses the Kalman filter method to reconstruct and smooth the data, and then extract the periodic eigenvector matrix after data reconstruction.

14. The method for detecting arc faults in aircraft according to claim 13, characterized in that: The arc fault classifier uses a pre-trained extreme learning machine, and the training method is as follows: (1) Set the input layer, hidden layer, and output layer nodes of the neural network, and set the learning rate and activation function; (2) Feature extraction is performed through the pre-collected arc data, normal data, load data and interference data, and the feature vector matrix is ​​extracted respectively. [T 1, T 2, T 3 ] The corresponding eigenvector matrices of the four types of data are obtained, and the corresponding analog arc eigenvector matrix, analog normal eigenvector matrix, load analog eigenvector matrix and interference analog eigenvector matrix are obtained. The array composed of the four types of eigenvector matrices is used as a training set for neural network training, and the output results of the neural network are arc fault and non-arc fault.

Images