A SINS / GPS integrated navigation fault diagnosis method based on RVM / chi-square-sequential probability ratio

Abstract

The present invention proposes a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, which uses the fault diagnosis method combining residual chi-square and sequential probability ratio to judge GPS information. When GPS works normally, the information of the accelerometer and gyroscope of SINS is used as the training input of the prediction model, and the training target output is the position increment of GPS at the same moment. The prediction model of GPS output is established by using RVM regression algorithm. When there is an abnormal source of GPS, this virtual GPS can predict the position increment of GPS based on the information of accelerometer and gyroscope collected by SINS, and establish the position information of virtual GPS in combination with the position result of the previous cycle, so as to realize SINS / GPS continuous navigation and avoid the accumulation of navigation system errors when GPS errors are large.

Description

Technical Field

[0001] The invention relates to a SINS / GPS combined navigation fault diagnosis technology, in particular to a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio. Background Art

[0002] SINS / GPS is one of the most widely used combined navigation systems. Although GPS has all-weather, high-precision positioning characteristics, in cities, canyons, forests and other areas, due to the obstruction of buildings, mountains, trees and the like, the GPS signal will be attenuated or lost, so that the GPS positioning result will usually contain wild value noise and slowly varying noise, resulting in a decrease in positioning accuracy or even failure to work properly. In order to accurately identify abnormal wild values ​​and slowly varying noise in the combined navigation system and reduce the influence of measurement noise on positioning accuracy, the present invention proposes a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio. The fault diagnosis method combining residual chi-square and sequential probability ratio is used to judge GPS information. When GPS works normally, the information of the accelerometer and gyroscope of SINS is used as the training input of the prediction model. The training target output is the position increment of GPS at the same time. The prediction model of GPS output is established by using RVM regression algorithm. When there is a GPS anomaly source, this virtual GPS can predict the GPS position increment based on the information of the accelerometer and gyroscope collected by SINS, and establish the position information of the virtual GPS in combination with the position result of the previous cycle, thereby realizing SINS / GPS continuous navigation and avoiding the accumulation of navigation system errors when the GPS error is large. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, and uses the fault diagnosis method combining residual chi-square and sequential probability ratio to judge GPS information. When GPS works normally, the information of the accelerometer and gyroscope of SINS is used as the training input of the prediction model, and the training target output is the position increment of GPS at the same moment. The prediction model of GPS output is established by using RVM regression algorithm. When there is an abnormal source of GPS, this virtual GPS can predict the position increment of GPS based on the information of accelerometer and gyroscope collected by SINS, and establish the position information of virtual GPS in combination with the position result of the previous cycle, so as to realize SINS / GPS continuous navigation and avoid the accumulation of navigation system errors when GPS errors are large.

[0004] A SINS / GPS integrated navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, comprising:

[0005] Step 1: Record the gyroscope and accelerometer information of SINS navigation;

[0006] The acceleration measurement information output by SINS is f b =[f E f U f N ] T , the gyroscope measurement information is

[0007] Step 2: Record the GPS navigation location information and calculate the GPS location increment information;

[0008] Position information output from GPS Compared with the position information output by the integrated navigation at the previous moment, the position increment is ΔP = [ΔL Δλ Δh] T ;

[0009] Step 3: Calculate the system sequential probability ratio detection value;

[0010] The system equation is as follows:

[0011]

[0012] Wherein, F(k) is the state transfer matrix, W(k) is the system noise, X(k) represents the n-dimensional state vector, G(k) represents the system noise distribution matrix, W(k) represents the system noise vector, Z(k) represents the m-dimensional measurement vector, H(k) represents the m×n-order measurement matrix, and V(k) represents the m-dimensional measurement noise vector;

[0013]

[0014] The innovation value of the navigation system at time k is expressed as:

[0015] v(k)=z(k)-H(k)X(k)

[0016] Its k consecutive independent samples are {v(1), v(2), ..., v(k)}, subject to is the sample mean, (σ(k)) 2 is the sample variance, which is calculated as follows:

[0017] Introducing the gradually disappearing memory factor, the recursive calculation formula of v(k) is:

[0018]

[0019] Among them, s represents the forgetting factor;

[0020] The sample variance is adaptively updated by the actual residual time series:

[0021]

[0022] Under the fault-free state H0 and the faulty state H1, the likelihood functions are:

[0023]

[0024] Taking the residuals of k independent moments {v(1), v(2), ..., v(k)}, according to the above binary hypothesis, the likelihood ratio is:

[0025]

[0026] Taking the logarithm of L(k):

[0027]

[0028] Then the sequential probability ratio increment is:

[0029]

[0030] Step 4: Using the information of the SINS gyroscope and accelerometer as input and the GPS position increment as output, a learning model based on the RVM algorithm is established;

[0031] Step 5: Determine whether the system fails based on the chi-square / fading sequential probability ratio and the probability ratio increment in step 3;

[0032] Step 6: If the system fails as determined in step 5, the current GPS position increment is calculated using the output information of the gyroscope and accelerometer of the current SINS and the learning model of step 4;

[0033] When a system failure is detected in step 5, the measurement information of the accelerometer and gyroscope collected by SINS in step 6 is used to calculate the virtual GPS position increment ΔP according to the previously established RVM neural network model. GPS , according to the position result P at the previous moment t-1 Calculate the current GPS virtual location information as P = P t-1 +ΔP GPS ;

[0034] Step 7: If the system fails in step 5, the current GPS position information is calculated based on the position information of the previous moment of the SINS / GPS combined navigation and the position increment information of step 6 to form new measurement information. If the system is not faulty, the current actual output of GPS is used to form system measurement information.

[0035] Step 8: Use Kalman filtering to perform SINS / GPS multi-source data fusion.

[0036] As a further improvement of the present invention, the step 4 establishes a learning model based on the RVM algorithm, and the steps are as follows:

[0037] 1) At time k, the acceleration measurement information output by SINS is f b (k), the gyroscope measurement information is The position information output by GPS is P GPS (k), compared with the previous moment, the position increment is ΔP GPS (k), take the system output at time k, f b ={f b (1), f b (2), ..., f b (k)}, As the training input of the prediction model, ΔP GPS = {ΔP GPS (1), ΔP GPS (2), ..., ΔP GPS (k)} as the training output of the prediction model;

[0038] 2) Use the RVM algorithm to establish a data model between the acceleration and gyroscope measurement information of SINS and the GPS position increment. The training process is as follows:

[0039] Given a dataset where x i is the input of the training sample, t i is the output of the training sample, which is assumed to be independent and accompanied by an expectation of 0 and a variance of σ 2 Gaussian noise ξ, establish the model:

[0040] t i =y(x i ;ω)+ξ i

[0041] Among them, ω is the weight vector, y(x i ;ω) is expressed as:

[0042]

[0043] Among them, K(x,x i ) is the model kernel function, and the Gaussian radial basis kernel function is selected as the model kernel function, and its form is:

[0044]

[0045] Among them, y c is the center of the kernel function, σ is the width of the kernel function, and the likelihood function of the RVM model is obtained by the selected kernel function:

[0046]

[0047] Among them, Φ is an N×(N+1) matrix based on the kernel function:

[0048] Φ=[φ(x1),φ(x2),...,φ(x N )] T

[0049] φ(x i )=[1,K(x i , x1), K(x i , x2), ..., K(x i , x N )] T

[0050] The ultimate goal is to give a new input x * , make a prediction * ,Right now:

[0051]

[0052] As a further improvement of the present invention, the step 5 performs fault diagnosis on the SINS / GPS integrated navigation system, and the steps are as follows:

[0053] 1) Fault detection based on residual chi-square:

[0054] At time k, if the system is working normally, the measured true value Z k and the measured predicted value Z k,k-1 The difference Should obey the Gaussian white noise distribution with zero mean:

[0055]

[0056] Its variance is:

[0057] S(k)=H(k)P k，k-1 H T (k)+R k

[0058] Constructing the statistic λ z (k):

[0059]

[0060] Under normal system conditions, z (k) Subject to χ with m degrees of freedom 2 Distribution, from which the following judgment criteria can be obtained:

[0061] when When , the system is considered to be working normally;

[0062] when When , the system is considered to have failed;

[0063] Among them, λ z (k) is the detection function, is the detection threshold, σ is the false alarm rate;

[0064] 2) Fault detection based on sequential probability ratio:

[0065] Set the detection threshold based on the false alarm rate and missed detection rate of fault diagnosis:

[0066]

[0067] Among them, P M represents the missed detection rate, P F represents the false alarm rate;

[0068] The following judgment criteria can be obtained:

[0069] When λ(k)≥T(H1), H1 is judged to be true;

[0070] When λ(k)≤T(H0), H0 is judged to be true;

[0071] When T(H0)≤λ(k)≤T(H1), add data and continue the test;

[0072] 3) Joint fault judgment based on chi-square-sequential probability ratio:

[0073] A fault diagnosis mechanism is established with the increment of the sequential probability ratio test value as the judgment condition. When the increment of the sequential probability ratio test value is positive, it indicates that the system may have a fault. At this time, as long as any result of the residual card method or the sequential probability ratio test method detects a fault, it is judged that the system has a fault. Otherwise, it indicates that the system has no fault. When the increment of the sequential probability ratio test value is negative, it indicates that the fault is in the disappearance stage. The diagnosis result of the residual card method should be used as the final diagnostic output of the system. At this time, if the residual card detects that the system has a fault, it is judged that the system has a fault. Otherwise, it indicates that the system has no fault.

[0074] The invention proposes a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, comprising the following steps: step 1: recording information of a gyroscope and an accelerometer of SINS navigation; step 2: recording position information of GPS navigation and calculating position increment information of GPS; step 3: calculating a system sequential probability ratio detection value; step 4: using information of a gyroscope and an accelerometer of SINS as input and a GPS position increment as output to establish a learning model based on an RVM algorithm; step 5: judging whether a system fault occurs according to a chi-square / fading sequential probability ratio and a probability ratio increment of step 3; step 6: if step 5 judges that a system fault occurs, using output information of a gyroscope and an accelerometer of the current SINS and the learning model of step 4 to calculate a current position increment of GPS; step 7: if step 5 judges that a system fault occurs, calculating GPS position information at a current moment according to the position information of the SINS / GPS combined navigation at a previous moment and the position increment information of step 6 to form new measurement information, and if the system has no fault, using the current actual output of GPS to form system measurement information. Step 8: Use Kalman filtering to perform SINS / GPS multi-source data fusion. Repeat the above steps until the navigation is completed, and use the fault diagnosis method combining residual chi-square and sequential probability ratio to judge the GPS information. When the GPS is working normally, the information of the accelerometer and gyroscope of SINS is used as the training input of the prediction model, and the training target output is the position increment of GPS at the same time. The RVM regression algorithm is used to establish the prediction model of GPS output. In the case of GPS anomaly source, this virtual GPS can predict the position increment of GPS based on the information of the accelerometer and gyroscope collected by SINS, and establish the position information of the virtual GPS based on the position result of the previous cycle, so as to realize SINS / GPS continuous navigation and avoid the accumulation of navigation system errors when the GPS error is large. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a schematic diagram of chi-square-sequential probability ratio fault detection;

[0076] Figure 2 Schematic diagram of virtual GPS information calculation based on RVM. DETAILED DESCRIPTION

[0077] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0078] The present invention proposes a SINS / GPS combined navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, and uses the fault diagnosis method combining residual chi-square and sequential probability ratio to judge GPS information. When GPS works normally, the information of the accelerometer and gyroscope of SINS is used as the training input of the prediction model, and the training target output is the position increment of GPS at the same moment. The prediction model of GPS output is established by using RVM regression algorithm. When there is an abnormal source of GPS, this virtual GPS can predict the position increment of GPS based on the information of the accelerometer and gyroscope collected by SINS, and establish the position information of virtual GPS in combination with the position result of the previous cycle, so as to realize SINS / GPS continuous navigation and avoid the accumulation of navigation system errors when GPS errors are large.

[0079] A SINS / GPS integrated navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, where the chi-square-sequential probability ratio fault detection diagram is shown in Figure 1 As shown, the schematic diagram of virtual GPS information calculation based on RVM is as follows Figure 2 As shown:

[0080] 1. The acceleration measurement information output by SINS is f b =[f E f U f N ] T , the gyroscope measurement information is Position information output from GPS Compared with the position information output by the integrated navigation at the previous moment, the position increment is ΔP = [ΔL Δλ Δh] T .

[0081] 2. Calculate the system sequential probability ratio detection value. The specific steps are as follows:

[0082] The system equation is as follows:

[0083]

[0084] Among them, F(k) is the state transfer matrix, W(k) is the system noise, X(k) represents the n-dimensional state vector, G(k) represents the system noise distribution matrix, W(k) represents the system noise vector, Z(k) represents the m-dimensional measurement vector, H(k) represents the m×n-order measurement matrix, and V(k) represents the m-dimensional measurement noise vector.

[0085] The innovation value of the navigation system at time k can be expressed as:

[0086] v(k)=Z(k)-H(k)X(k)

[0087] Its k consecutive independent samples are {v(1), v(2), ...., v(k)}, subject to is the sample mean, (σ(k)) 2 is the sample variance, which is calculated as follows:

[0088] Introducing the gradually disappearing memory factor, the recursive calculation formula of v(k) is:

[0089]

[0090] Where s represents the forgetting factor.

[0091] The sample variance is adaptively updated by the actual residual time series:

[0092]

[0093] Under the fault-free state H0 and the faulty state H1, the likelihood functions are:

[0094]

[0095] Taking the residuals of k independent moments {v(1), v(2), ..., v(k)}, according to the above binary hypothesis, the likelihood ratio is:

[0096]

[0097] Taking the logarithm of L(k):

[0098]

[0099] Then the sequential probability ratio increment is:

[0100]

[0101] 3. Establish a learning model based on the RVM algorithm. The steps are as follows:

[0102] (1) At time k, the acceleration measurement information output by SINS is f b (k), the gyroscope measurement information is The position information output by GPS is P GPS (k), compared with the previous moment, the position increment is ΔP GPS (k). Take the system output at k moments, f b ={f b (1), f b (2), ..., f b (k)}, As the training input of the prediction model, ΔP GPS = {ΔP GPS (1), ΔPGPS (2), ..., ΔP GPS (k)} as the training output of the prediction model.

[0103] (2) The RVM algorithm is used to establish a data model between the acceleration and gyroscope measurement information of SINS and the GPS position increment. The training process is as follows:

[0104] Given a dataset where x i is the input of the training sample, t i is the output of the training sample, which is assumed to be independent and accompanied by an expectation of 0 and a variance of σ 2 Gaussian noise ξ, establish the model:

[0105] t i =y(x i ;ω)+ξ i

[0106] Where ω is the weight vector. i ;ω) can be expressed as:

[0107]

[0108] Among them, K(x,x i ) is the model kernel function. The Gaussian radial basis kernel function is selected as the model kernel function, and its form is:

[0109]

[0110] Among them, y c is the center of the kernel function, and σ is the width of the kernel function. The likelihood function of the RVM model is obtained by the selected kernel function:

[0111]

[0112] Among them, Φ is an N×(N+1) matrix based on the kernel function:

[0113] Φ=[φ(x1),φ(x2),...,φ(x N )] T

[0114] φ(x i )=[1,K(x i , x1), K(x i , x2), ..., K(x i , x N )] T

[0115] The ultimate goal is to give a new input x * , make a prediction* ,Right now:

[0116]

[0117] 4. To diagnose the fault of SINS / GPS integrated navigation system, the steps are as follows:

[0118] Fault detection based on residual chi-square:

[0119] At time k, if the system is working normally, the measured true value Z k and the measured predicted value Z k,k-1 The difference Should obey the Gaussian white noise distribution with zero mean:

[0120]

[0121] Its variance is:

[0122] S(k)=H(k)P k，k-1 H T (k)+R k

[0123] Constructing the statistic λ z (k):

[0124]

[0125] Under normal system conditions, z (k) Subject to x with m degrees of freedom 2 Distribution, from which the following judgment criteria can be obtained:

[0126] when When , the system is considered to be working normally;

[0127] when , the system is considered to have failed.

[0128] Among them, λ z (k) is the detection function, is the detection threshold, and α is the false alarm rate.

[0129] Fault detection based on sequential probability ratio:

[0130] Set the detection threshold based on the false alarm rate and missed detection rate of fault diagnosis:

[0131]

[0132] Among them, P M represents the missed detection rate, P F Indicates the false alarm rate.

[0133] The following judgment criteria can be obtained:

[0134] When λ(k)≥T(H1), H1 is judged to be true;

[0135] When λ(k)≤T(H0), H0 is judged to be true;

[0136] When T(H0)≤λ(k)≤T(H1), add data and continue the test.

[0137] Joint fault judgment based on chi-square-sequential probability ratio:

[0138] A fault diagnosis mechanism is established with the increment of the sequential probability ratio detection value as the judgment condition. When the increment of the sequential probability ratio detection value is positive, it indicates that the system may have a fault. At this time, as long as any result of the residual card method or the sequential probability ratio detection method detects a fault, it is judged that the system has a fault, otherwise it indicates that the system has no fault. When the increment of the sequential probability ratio detection value is negative, it indicates that the fault is in the disappearance stage. The diagnosis result of the residual card method should be used as the final diagnostic output of the system. At this time, if the residual card detects that the system has a fault, it is judged that the system has a fault, otherwise it indicates that the system has no fault.

[0139] 5. Create virtual GPS information, the steps are as follows:

[0140] When a system failure is detected, the virtual GPS position increment ΔP is calculated based on the previously established RVM neural network model using the measurement information of the accelerometer and gyroscope collected by SINS. GPS , according to the position result P at the previous moment t-1 Calculate the current GPS virtual location information as P = P t-1 +ΔP GPS .

[0141] 6. When a system failure is detected, the calculated virtual GPS position information is used to participate in data fusion. If no system failure is detected, the current GPS position information is directly used for data fusion.

[0142] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A SINS / GPS integrated navigation fault diagnosis method combining RVM / chi-square-sequential probability ratio, comprising: Step 1: Record the gyroscope and accelerometer information of SINS navigation; The acceleration measurement information output by SINS is f b =[f E f U f N ] T , the gyroscope measurement information is Step 2: Record the GPS navigation location information and calculate the GPS location increment information; Position information output from GPS Compared with the position information output by the integrated navigation at the previous moment, the position increment is ΔP = [ΔL Δλ Δh] T ; Step 3: Calculate the system sequential probability ratio detection value; The system equation is as follows: Wherein, F(k) is the state transfer matrix, X(k) represents the n-dimensional state vector, G(k) represents the system noise allocation matrix, W(k) represents the system noise vector, Z(k) represents the m-dimensional measurement vector, H(k) represents the m×n-order measurement matrix, and V(k) represents the m-dimensional measurement noise vector; The innovation value of the navigation system at time k is expressed as: v(k)=z(k)-H(k)X(k) Its k consecutive independent samples are {v(1), v(2), ..., v(k)}, subject to is the sample mean, (σ(k)) 2 is the sample variance, which is calculated as follows: Introducing the gradually disappearing memory factor, the recursive calculation formula of v(k) is: Among them, s represents the forgetting factor; The sample variance is adaptively updated by the actual residual time series: Under the fault-free state H0 and the faulty state H1, the likelihood functions are: Taking the residuals of k independent moments {v(1), v(2), ..., v(k)}, according to the above binary hypothesis, the likelihood ratio is: Taking the logarithm of L(k): Then the sequential probability ratio increment is: Step 4: Using the information of the SINS gyroscope and accelerometer as input and the GPS position increment as output, a learning model based on the RVM algorithm is established; Step 5: Determine whether the system fails based on the chi-square / fading sequential probability ratio and the probability ratio increment in step 3; Step 6: If the system fails as determined in step 5, the current GPS position increment is calculated using the output information of the gyroscope and accelerometer of the current SINS and the learning model of step 4; When a system failure is detected in step 5, the measurement information of the accelerometer and gyroscope collected by SINS in step 6 is used to calculate the virtual GPS position increment ΔP according to the previously established RVM neural network model. GPS , according to the position result P at the previous moment t-1 Calculate the current GPS virtual location information as P = P t-1 +ΔP GPS ; Step 7: If the system fails in step 5, the current GPS position information is calculated based on the position information of the previous moment of the SINS / GPS combined navigation and the position increment information of step 6 to form new measurement information. If the system is not faulty, the current actual output of GPS is used to form system measurement information. Step 8: Use Kalman filtering to perform SINS / GPS multi-source data fusion.

2. The SINS / GPS integrated navigation fault diagnosis method of claim 1 is characterized by: The step 4 establishes a learning model based on the RVM algorithm, and the steps are as follows: 1) At time k, the acceleration measurement information output by SINS is f b (k), the gyroscope measurement information is The position information output by GPS is P GPS (k), compared with the previous moment, the position increment is ΔP GPS (k), take the system output at time k, f b ={f b (1), f b (2), ..., f b (k)}, As the training input of the prediction model, ΔP GPS = {ΔP GPS (1), ΔP GPS (2), ..., ΔP GPS (k)} as the training output of the prediction model; 2) Use the RVM algorithm to establish a data model between the acceleration and gyroscope measurement information of SINS and the GPS position increment. The training process is as follows: Given a dataset where x i is the input of the training sample, t i is the output of the training sample, which is assumed to be independent and accompanied by an expectation of 0 and a variance of σ 2 Gaussian noise ξ, establish the model: t i =y(x i ;ω)+ξ i Among them, ω is the weight vector, y(x i ;ω) is expressed as: Among them, K(x,x i ) is the model kernel function, and the Gaussian radial basis kernel function is selected as the model kernel function, and its form is: Among them, y c is the center of the kernel function, σ is the width of the kernel function, and the likelihood function of the RVM model is obtained by the selected kernel function: Among them, Φ is an N×(N+1) matrix based on the kernel function: Φ=[φ(x1),φ(x2),...,φ(x N )] T φ(x i )=[1,K(x i ,x1),K(x i ,x2),...,K(x i x N )] T The ultimate goal is to give a new input x * , make a prediction * ,Right now:

3. The SINS / GPS integrated navigation fault diagnosis method combined with RVM / chi-square-sequential probability ratio according to claim 1 is characterized in that: Step 5 performs fault diagnosis on the SINS / GPS integrated navigation system, and the steps are as follows: 1) Fault detection based on residual chi-square: At time k, if the system is working normally, the measured true value Z k and the measured predicted value Z k，k-1 The difference Should obey the Gaussian white noise distribution with zero mean: Its variance is: S(k)=H(k)P k，k-1 H T (k)+R k Constructing the statistic λ z (k): Under normal system conditions, z (k) Subject to χ with m degrees of freedom 2 Distribution, from which the following judgment criteria can be obtained: when When , the system is considered to be working normally; when When , the system is considered to have failed; Among them, λ z (k) is the detection function, is the detection threshold, α is the false alarm rate; 2) Fault detection based on sequential probability ratio: Set the detection threshold based on the false alarm rate and missed detection rate of fault diagnosis: Among them, P M represents the missed detection rate, P F represents the false alarm rate; This leads to the following judgment criteria: When λ(k)≥T(H1), H1 is judged to be true; When λ(k)≤T(H0), H0 is judged to be true; When T(H0)≤λ(k)≤T(H1), add data and continue the test; 3) Joint fault judgment based on chi-square-sequential probability ratio: A fault diagnosis mechanism is established with the increment of the sequential probability ratio test value as the judgment condition. When the increment of the sequential probability ratio test value is positive, it indicates that the system may have a fault. At this time, as long as any result of the residual card method or the sequential probability ratio test method detects a fault, it is judged that the system has a fault. Otherwise, it indicates that the system has no fault. When the increment of the sequential probability ratio test value is negative, it indicates that the fault is in the disappearance stage. The diagnosis result of the residual card method should be used as the final diagnostic output of the system. At this time, if the residual card detects that the system has a fault, it is judged that the system has a fault. Otherwise, it indicates that the system has no fault.

Images