CN114115184B - High-speed train suspension system fault diagnosis method based on data drive ToMFIR - Google Patents
High-speed train suspension system fault diagnosis method based on data drive ToMFIR Download PDFInfo
- Publication number
- CN114115184B CN114115184B CN202111347424.1A CN202111347424A CN114115184B CN 114115184 B CN114115184 B CN 114115184B CN 202111347424 A CN202111347424 A CN 202111347424A CN 114115184 B CN114115184 B CN 114115184B
- Authority
- CN
- China
- Prior art keywords
- fault
- matrix
- suspension system
- input
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000000725 suspension Substances 0.000 title claims abstract description 84
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000003745 diagnosis Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 68
- 238000005259 measurement Methods 0.000 claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 238000013499 data model Methods 0.000 claims abstract description 11
- 238000011156 evaluation Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 12
- 238000006073 displacement reaction Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 238000004088 simulation Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 3
- 238000002347 injection Methods 0.000 description 3
- 239000007924 injection Substances 0.000 description 3
- 239000006096 absorbing agent Substances 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 230000035939 shock Effects 0.000 description 2
- 102000002274 Matrix Metalloproteinases Human genes 0.000 description 1
- 108010000684 Matrix Metalloproteinases Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000013016 damping Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
- G05B23/0254—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
Abstract
The invention discloses a fault diagnosis method for a high-speed train suspension system based on a data drive ToMFIR, which comprises the following steps: carrying out data modeling on an actual model and a nominal model of the high-speed train suspension system to obtain an input-output data model of the actual model and an input-output data model of the nominal model, which are hidden in the input-output relation of the suspension system; designing a fault information total measurement residual error based on data driving by utilizing the input and output data matrix of the obtained actual model and the input and output data matrix of the nominal model; constructing an evaluation function by adopting fault information full-metric residual errors, and alarming when a detection index reaches an alarm threshold by combining with a fault detection alarm threshold; and estimating the amplitude value of the micro fault of the suspension system sensor by using the fault information total measurement residual error based on data driving. The invention has higher sensitivity to the micro faults of the high-speed train suspension system, and can detect and accurately estimate the dynamic characteristics of the faults in real time.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of high-speed train suspension systems, and relates to a fault diagnosis method of a high-speed train suspension system based on data drive ToMFIR (Total Measurable Fault Information Residual, fault information total measurement residual).
Background
China has built a high-speed railway network with the maximum worldwide rule and the highest operation speed after more than 10 years of development since China opens CRH (CHINA RAILWAY HIGH-speed) motor train unit trains with the speed per hour of more than 200 km on a main trunk line in 2008.
The high-speed train suspension system supports the train body and the bogie, and simultaneously isolates the wheel-rail acting force caused by unsmooth track, so that the stability and the safety of the train during high-speed running are ensured, and the high-speed train suspension system has high reliability requirements. The high-speed train suspension system is divided into active suspension and semi-active suspension, and adopts a closed-loop control structure. The suspension system adopts a two-system suspension device, wherein the first-system suspension device is arranged between the axle box and the bogie frame, and the second-system suspension device is arranged between the bogie frame and the vehicle body; including a number of components including coil springs, lateral/vertical dampers, air springs, active actuators, and sensors. The active actuator is used as an important actuator component, and is important for safe running and riding comfort of a high-speed train. The active actuator calculates the active control force required by the active suspension system according to the vehicle output signal measured by the sensor.
Suspension system faults may be categorized as actuator faults, sensor faults, and mechanical component faults. With the increase of the on-track running time of a train, some components in a suspension system, such as a coil spring, a shock absorber, an air spring, an active/semi-active actuator and a sensor, can generate a certain performance degradation, and tiny faults such as tiny cracking of the coil spring, slight oil leakage/liquid leakage of the shock absorber, slight air leakage of the air spring, small-amplitude actuation efficiency loss of the actuator, deviation drift of the sensor and the like are induced, so that potential danger is brought to the running safety of the train.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a fault diagnosis method for a high-speed train suspension system based on a data drive ToMFIR. The method aims at a suspension system of a closed-loop control structure of a high-speed train, takes an active actuator, a sensor and a mechanical component as fault diagnosis objects, and can detect and accurately estimate the dynamic characteristics of faults in real time when the suspension system breaks down or even breaks down.
In order to solve the technical problems, the invention adopts the following technical scheme.
The invention discloses a fault diagnosis method for a high-speed train suspension system based on a data drive ToMFIR, which comprises the following steps of:
Step 1, carrying out data modeling on an actual model and a nominal model of a high-speed train suspension system to obtain an input-output data model of the actual model and an input-output data model of the nominal model, which are hidden in the input-output relation of the suspension system;
step2, utilizing a displacement sensor and a gyroscope to obtain vertical displacement and angular velocity signals of the mass center of a carriage of the high-speed train and vertical displacement signals of the mass center of a bogie frame during running of the high-speed train, wherein the signals of the sensor and the gyroscope are system output; the obtained input and output signals of the actual model are used for constructing an input and output data matrix of the actual model, and the obtained input and output signals of the nominal model are used for constructing an input and output data matrix of the nominal model;
Step 3, designing fault information total measurement residual errors based on data driving by utilizing the input and output data matrix of the actual model and the input and output data matrix of the nominal model obtained in the step 2;
Step 4, constructing an evaluation function J by utilizing the full-metric residual error of the fault information based on the data driving designed in the step 3, and alarming when the detection index reaches an alarm threshold by combining with the fault detection alarm threshold;
And 5, estimating the amplitude of the micro fault of the suspension system sensor by using the full-metric residual error of the fault information based on the data driving designed in the step 3.
Further, in the step 1, the specific process of modeling the actual model and the nominal model includes:
Step 1.1. Discrete actual and nominal models G 0 of the vertical suspension system of a high-speed train can be expressed as:
Wherein A, B, C, E d,Ef,Ff are coefficient matrixes corresponding to the space state equation; x (k), y (k), u (k), d (k) are respectively state variables, output variables, control input variables and track disturbance excitation of an actual system; f (k) characterizes all possible faults; ζ (k) is process noise and measurement noise, respectively; x 0(k),y0(k),u0 (k) are the state variables, output variables and control input variables of the nominal model, respectively;
Step 1.2. From the state space equation (1) of the suspension system actual model G in step 1.1, there are:
Wherein y l(k),ul(k),dl(k),fl (k), Ζ l (k) is a stack matrix Γ l,Hu,l,Hd,l,Hf,l,/>Is a corresponding coefficient matrix, and is specifically formed as follows
Step 1.3 the expression (3) obtained in step 1.2 contains the state variable x (k), which is obtained from the discrete state space equation (1) in step 1.1 to eliminate the state variable x (k)
Wherein the method comprises the steps of
Step 1.4. Substituting the expression (5) obtained in step 1.3 into the expression (2) obtained in step 1.2, there are
Simplifying and obtaining
yl(k)≈Hu,p,lup,l(k)+Hd,p,ldp,l(k)+Hf,p,lfp,l(k)+He,p,lep,l(k) (8)
Wherein the method comprises the steps of
Considering the time interval N, from the reduced expression (8) obtained in step 1.4, a data model of a high-speed train suspension system can be obtained that implies the input-output relationship of the suspension system as follows:
Yk,l＝Hu,p,lUk,p,l+Hd,p,lDk,p,l+Hf,p,lFk,p,l+He,p,lEk,p,l (10)
Wherein Y k,l,Uk,p,l,Dk,p,l,Fk,p,l,Ek,p,l is a data matrix; the input and output data matrixes Y k,l and U k,p,l of the actual system can be obtained by measuring a sensor, a gyroscope and the like when the train runs; the data matrix forms are as follows:
step 1.5 is similar, and from the state space equation (2) of the suspension system nominal model G 0 in step 1.1, there is:
Wherein the method comprises the steps of Input-output data matrix of nominal system respectively,/>Respectively have the same form as U k,p,l,Yk,l; and/>Which can be measured by sensors, gyroscopes, etc. while the train is running.
Further, in the step 3, the specific process of designing the full-metric residual error based on the fault information driven by the data includes:
step 3.1 defining an output residual R Y for characterizing the output differences of the actual model and the nominal model
Step 3.2. define a controller residual R U for characterizing the output differences of the controller in the actual model and the nominal model
Step 3.3 defining fault information full metric residual error under closed loop control structure
Wherein,Characterizing the system output of the nominal model driven by the real-time input signal;
Step 3.4. Introduction of an orthogonal projection matrix In step 3.3, defining a fault information full-metric residual error under a closed-loop control structure, namely, simultaneously right multiplying the orthogonal projection matrix/>, on both sides of an equal sign of a formula (15)The following fault information full metric residuals based on data driving are available:
Wherein ToMFIR k is characterized in Thus, the task of micro fault diagnosis may be focused on monitoring the changes of the residual signal matrix ToMFIR k that depend only on the input and output data of the system.
Further, in the step 4, the detection of the suspension system micro fault is performed by using the fault information total metric residual error based on the data driving, and the specific steps include:
Step 4.1. Residual Signal matrix ToMFIR k∈Rm×N may be written as Where τ i＝[τ1i,…,τ1N ], i=1, …, m; let/>Is a standardized form of ToMFIR k and satisfies/>Wherein/>
Step 4.2. decomposing the matrix ToMFIR k into a scoring matrix T and a loading matrix P by Singular Value Decomposition (SVD), assuming ToMFIR k is a Gaussian or near-Gaussian distribution, there areWherein Λ = diag (λ 1,…,λm); the score matrix T, the load matrix P and the feature matrix Λ may be derived from the following unbiased estimates:
Step 4.3. Introduction of a fault-free residual Signal The corresponding scoring matrices with the same dimensions as ToMFIR k are T and T rf, respectively; assuming that the average parameters of the distribution after the occurrence of the micro-fault are unchanged, there are
Step 4.4. Let T have a probability density of F, T rf have a probability density of F rf, based on the multidimensional KL divergence, the following evaluation function can be defined:
Step 4.5. The designed fault information total measurement residual error based on data driving and evaluation function are utilized, and the fault detection mechanism of the system is as follows
Further, in the step 5, the amplitude estimation of the micro fault of the suspension system sensor is performed by using the fault information total metric residual error based on the data driving, and the specific steps include:
Step 5.1. Covariance matrix S is different from S rf due to the action of the micro-fault; the amplitude a of the micro fault is smaller and is an unknown constant value and is close to 0 but not equal to 0, so that the influence on signals is extremely weak; the effects caused by micro-faults can be converted into changes in the feature matrix, i.e
Step 5.2. For ToMFIR k, each score vector t i corresponds to a probability density functionPair/>Each score vector/>Corresponds to a probability density function/>Differences between t rf and t were measured using KL divergence:
wherein Δλ i is the characteristic value change caused by a minute failure;
Step 5.3. As a.fwdarw.0, deltalambda i also tends to 0, lambda i Taylor is developed
From s=pΛp T, the n-order derivative of the eigenvalue λ i with respect to the fault amplitude a isWherein p i isCorresponding feature vectors;
Step 5.4. There is/>Wherein R Y,i＝[ri,1,ri,2,…,ri,N is row i of R Y; for minor failure faults of the sensor, the method can be used forA representation; and then obtain/>And its standardized form/>
Step 5.5. RecordIs a fault item,/>For non-faulty items, there are
All the constant parameters are irrelevant to fault amplitude and can be calculated by fault-free reference data: δ rq =0,The first order derivative of the covariance matrix S to the fault amplitude a is
Second derivative of
The higher order derivative (n > 2) is 0;
step 5.6. Record vector p i＝[p1i…,pmi ], there is
The change in the characteristic value caused by the minute trouble can be expressed as
The KL divergence in step 5.2 can be expressed as
Wherein the method comprises the steps of
Step 5.7, utilizing the designed fault information total measurement residual error based on data driving, aiming at the tiny failure fault of the sensor of the suspension system of the high-speed train, the estimated value of the fault amplitude based on the divergence value is as follows:
compared with the prior art, the invention has the following advantages and beneficial effects:
1. The method aims at the suspension system of the closed-loop control structure of the high-speed train, takes the active actuator, the sensor and the mechanical component as fault diagnosis objects, and can detect and accurately estimate the dynamic characteristics of faults in real time when the suspension system breaks down or even breaks down.
2. The fault diagnosis method of the invention only depends on the input and output data of the actual model and the nominal model of the train suspension system, does not need to know an accurate system model, does not need to identify parameters, and is based on a data driving fault diagnosis method.
3. The fault diagnosis method can collect more comprehensive fault information, so that the fault diagnosis method is more sensitive to the micro faults of the high-speed railway suspension system actuator, the sensor and the mechanical component, and can effectively and timely detect the generated micro faults.
4. The fault diagnosis method can effectively estimate the fault evolution process and amplitude.
5. The fault diagnosis method has higher sensitivity to faults of the high-speed train suspension system and even micro faults, effectively solves the diagnosis and engineering practical problems of the micro faults of the high-speed train suspension system under a closed-loop control structure, and has important significance for micro early warning and real-time monitoring of the high-speed train suspension faults.
Drawings
Fig. 1 is a schematic diagram of a fault diagnosis system for a suspension system of a high-speed train according to an embodiment of the present invention.
FIG. 2 is a schematic diagram of the sensor and gyroscope mounting locations of a high speed train suspension system in accordance with one embodiment of the present invention.
FIG. 3 is a flow chart of a fault diagnosis method of an embodiment of the present invention.
FIG. 4 is a simulation graph of fault detection for a suspension system in accordance with one embodiment of the present invention.
FIG. 5 is a simulation graph of the detection of a micro-fault in a suspended machine component in accordance with one embodiment of the present invention.
FIG. 6 is a simulation graph of suspended actuator micro-fault detection for one embodiment of the present invention.
FIG. 7 is a simulation graph of suspension sensor micro-fault detection for one embodiment of the present invention.
FIG. 8 is a simulation plot of a suspension sensor micro-failure fault magnitude estimation for one embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings.
Fig. 1 is a schematic diagram of a fault diagnosis system for a suspension system of a high-speed train according to an embodiment of the present invention. As shown in fig. 1, the fault diagnosis system includes: the sensor is used for acquiring the vertical displacement and pitch angle of the gravity center of the train and the vertical displacement of the gravity center of the bogie frame; the acquisition system is used for acquiring and processing the measurement data of the sensor and then sending the measurement data to the fault diagnosis host; and the fault diagnosis host judges whether the high-speed train suspension system is faulty or not according to the collected information and estimates the fault amplitude. The installation positions of the sensors and the gyroscopes in each carriage are shown in figure 2.
As shown in fig. 3, a method for diagnosing a minor fault of a high-speed train suspension system based on a data drive ToMFIR according to an embodiment of the present invention includes the steps of:
Step 1, carrying out data modeling on an actual model and a nominal model of a high-speed train suspension system to obtain an input-output data model of the actual model and an input-output data model of the nominal model, which are hidden in the input-output relation of the suspension system;
and 2, acquiring vertical displacement and angular velocity signals of the mass center of the carriage of the high-speed train and vertical displacement signals of the mass center of the bogie frame by using a displacement sensor and a gyroscope, wherein the signals of the sensor and the gyroscope are output by a system. The obtained input and output signals of the actual model are used for constructing an input and output data matrix of the actual model, and the obtained input and output signals of the nominal model are used for constructing an input and output data matrix of the nominal model;
Step 3, designing fault information total measurement residual errors based on data driving by utilizing the input and output data matrix of the actual model and the input and output data matrix of the nominal model obtained in the step 2;
Step 4, constructing an evaluation function J by utilizing the full-metric residual error of the fault information based on the data driving designed in the step 3, and alarming when the detection index reaches an alarm threshold by combining with the fault detection alarm threshold;
And 5, estimating the amplitude of the micro fault of the suspension system sensor by using the full-metric residual error of the fault information based on the data driving designed in the step 3.
In the step 1, the specific process of data modeling of the actual model and the nominal model includes:
Step 1.1. According to the vertical suspension system of a high speed train shown in fig. 2, the discrete actual model G and nominal model G 0 can be expressed as:
Wherein A, B, C, E d,Ef,Ff are coefficient matrixes corresponding to the space state equation; x (k), y (k), u (k), d (k) are respectively state variables, output variables, control input variables and track disturbance excitation of an actual system; f (k) characterizes all possible faults; ζ (k) is process noise and measurement noise, respectively. x 0(k),y0(k),u0 (k) are the state variables, output variables and control input variables, respectively, of the nominal model.
Step 1.2. From the state space equation (1) of the suspension system actual model G in step 1.1, there are:
Wherein y l(k),ul(k),dl(k),fl (k), Ζ l (k) is a stack matrix Γ l,Hu,l,Hd,l,Hf,l,/>Is a corresponding coefficient matrix, and is specifically formed as follows
Step 1.3 the expression (3) obtained in step 1.2 contains the state variable x (k), which is obtained from the discrete state space equation (1) in step 1.1 to eliminate the state variable x (k)
Wherein the method comprises the steps of
Step 1.4. Substituting the expression (5) obtained in step 1.3 into the expression (2) obtained in step 1.2, there are
Simplifying and obtaining
yl(k)≈Hu,p,lup,l(k)+Hd,p,ldp,l(k)+Hf,p,lfp,l(k)+He,p,lep,l(k) (37)
Wherein the method comprises the steps of
Considering the time interval N, from the reduced expression (8) obtained in step 1.4, a data model of a high-speed train suspension system can be obtained that implies the input-output relationship of the suspension system as follows:
Yk,l＝Hu,p,lUk,p,l+Hd,p,lDk,p,l+Hf,p,lFk,p,l+He,p,lEk,p,l (39)
Wherein Y k,l,Uk,p,l,Dk,p,l,Fk,p,l,Ek,p,l is a data matrix; the input and output data matrix Y k,l and U k,p,l of the actual system can be measured by sensors, gyroscopes and the like when the train runs. The data matrix forms are as follows:
step 1.5 is similar, and from the state space equation (2) of the suspension system nominal model G 0 in step 1.1, there is:
Wherein the method comprises the steps of Input-output data matrix of nominal system respectively,/>Respectively have the same form as U k,p,l,Yk,l; and/>Which can be measured by sensors, gyroscopes, etc. while the train is running.
In the step 3, the specific process of designing the fault information total metric residual error based on data driving includes:
step 3.1 defining an output residual R Y for characterizing the output differences of the actual model and the nominal model
Step 3.2. define a controller residual R U for characterizing the output differences of the controller in the actual model and the nominal model
Step 3.3 defining fault information full metric residual error under closed loop control structure
Wherein,The system output of the nominal model driven by the real-time input signal is characterized.
Step 3.4. Introduction of an orthogonal projection matrixIn step 3.3, defining a fault information full-metric residual error under a closed-loop control structure, namely, simultaneously right multiplying the orthogonal projection matrix/>, on both sides of an equal sign of a formula (15)The following fault information full metric residuals based on data driving are available:
Wherein ToMFIR k is characterized in Thus, the task of micro fault diagnosis may be focused on monitoring the changes of the residual signal matrix ToMFIR k that depend only on the input and output data of the system.
In step4, the suspension system micro fault is detected by using the fault information total metric residual error based on the data driving. The method comprises the following specific steps:
Step 4.1. Residual Signal matrix ToMFIR k∈Rm×N may be written as Where τ i＝[τ1i,…,τ1N ], i=1, …, m. Let/>Is ToMFIR k in standardized form and meetsWherein/>
Step 4.2. decomposing the matrix ToMFIR k into a scoring matrix T and a loading matrix P by Singular Value Decomposition (SVD), assuming ToMFIR k is a Gaussian or near-Gaussian distribution, there areWhere Λ=diag (λ 1,…,λm). The score matrix T, the load matrix P and the feature matrix Λ may be derived from the following unbiased estimates:
Step 4.3. Introduction of a fault-free residual Signal And ToMFIR k have the same dimensions and their corresponding score matrices are T and T rf, respectively. Assuming that the average parameters of the distribution after the occurrence of the micro-fault are unchanged, there are
Step 4.4. Let T have a probability density of F, T rf have a probability density of F rf, based on the multidimensional KL divergence, the following evaluation function can be defined:
Step 4.5. The designed fault information total measurement residual error based on data driving and evaluation function are utilized, and the fault detection mechanism of the system is as follows
In step 5, the amplitude estimation of the suspension sensor micro fault is performed using the fault information total metric residual error based on the data driving. The method comprises the following specific steps:
Step 5.1. Covariance matrix S is different from S rf due to the effect of the micro-faults. The amplitude a of the micro fault is small and is an unknown constant value, is close to 0 but not equal to 0, and thus has very weak influence on the signal. The effects caused by micro-faults can be converted into changes in the feature matrix, i.e
Step 5.2. For ToMFIR k, each score vector t i corresponds to a probability density functionPair/>Each score vector/>Corresponds to a probability density function/>Differences between t rf and t were measured using KL divergence:
wherein Δλ i is the characteristic value change caused by a minute failure.
Step 5.3. As a.fwdarw.0, deltalambda i also tends to 0, lambda i Taylor is developed
From s=pΛp T, the n-order derivative of the eigenvalue λ i with respect to the fault amplitude a isWherein p i isCorresponding feature vectors.
Step 5.4.There is/>Where R Y,i＝[ri,1,ri,2,…,ri,N is row i of R Y. For minor failure faults of the sensor, the method can be used forAnd (3) representing. And then obtain/>And its standardized form/>
Step 5.5. RecordIs a fault item,/>For non-faulty items, there are
All the constant parameters are irrelevant to fault amplitude and can be calculated by fault-free reference data: δ rq =0,The first order derivative of the covariance matrix S to the fault amplitude a is
Second derivative of
The higher order derivative (n > 2) is 0.
Step 5.6. Record vector p i＝[p1i…,pmi ], there is
The change in the characteristic value caused by the minute trouble can be expressed as
The KL divergence in step 5.2 can be expressed as
/>
Wherein the method comprises the steps of
Step 5.7, utilizing the designed fault information total measurement residual error based on data driving, aiming at the tiny failure fault of the sensor of the suspension system of the high-speed train, the estimated value of the fault amplitude based on the divergence value is as follows:
The method of the present invention is simulated and verified as follows,
Step 1, setting the characteristic information of the mechanical component faults, which comprises the following steps: fault1, 10% of front bogie secondary springs, 10% of Fault2, 10% of front bogie secondary damping, 10% of Fault3, 10% of front bogie primary springs, 10% of Fault4, 30 seconds of Fault starting time and simulation ending time of Fault ending time, and the Fault is subjected to software injection through a Fault injection module to establish a Fault model;
Step 2, setting characteristic information of the actuator fault, including: fault5, wherein the front bogie actuator fails by 10%, fault6, the rear bogie actuator fails by 10%, the Fault starting time is 30 seconds, the Fault ending time is simulation ending time, and the Fault is injected by software through a Fault injection module to establish a Fault model;
step 3, setting the characteristic information of the sensor fault, which comprises the following steps: fault7, wherein the carriage sensor has a gradual drift micro Fault, the drift amount is theta (t) =0.01× (0.25+0.02sin (t) +0.01sin (0.2 t)), fault8, the carriage sensor fails 10%, the Fault starting time is 30 seconds, the Fault ending time is simulation ending time, the Fault is subjected to software injection through a Fault injection module, and a Fault model is established;
And 4, adopting SIMPACK and Matlab/Simulink joint simulation, introducing a model of rail-train coupling established in the SIMPACK into the Matlab/Simulink, and establishing a train vertical suspension system control simulation model in the Simulink. The running speed of the vehicle was set to 250Km/h, and the simulation was often 50 seconds.
As shown in fig. 4, when the high-speed train suspension system has no fault, the fault false alarm rate of the method provided by the invention is extremely low.
As shown in fig. 5, when the mechanical parts of different types at each position of the high-speed train suspension system have micro faults, the method provided by the invention can well detect the occurrence of the micro faults of the mechanical parts.
As shown in FIG. 6, when the actuators at each position of the suspension system of the high-speed train have micro faults, the method provided by the invention can well detect the occurrence of the micro faults of the actuators.
As shown in FIG. 7, when different types of micro faults occur to the sensors of the suspension system of the high-speed train, the method provided by the invention can well detect the occurrence of the micro faults of the sensors of the suspension system of the train.
As shown in fig. 8, when a sensor of a suspension system of a high-speed train has a tiny failure fault, the method provided by the invention has excellent performance in the aspects of fault evolution process and amplitude estimation.
As described above, the fault diagnosis method has higher sensitivity to faults of the high-speed train suspension system and even micro faults, can effectively realize the diagnosis of the micro faults of the high-speed train suspension system, effectively solves the diagnosis of the micro faults and the engineering practical problems of the micro faults under a closed-loop control structure, and has important significance to the micro early warning and real-time monitoring of the high-speed train suspension faults.
Claims (1)
1. A high-speed train suspension system fault diagnosis method based on data drive ToMFIR, comprising the steps of:
Step 1, carrying out data modeling on an actual model and a nominal model of a high-speed train suspension system to obtain an input-output data model of the actual model and an input-output data model of the nominal model, which are hidden in the input-output relation of the suspension system;
Step 2, utilizing a displacement sensor and a gyroscope to obtain vertical displacement and angular velocity signals of the mass center of a carriage of the high-speed train and vertical displacement signals of the mass center of a bogie frame during running of the high-speed train, wherein the signals of the sensor and the gyroscope are system output;
Step 3, designing fault information total measurement residual errors based on data driving by utilizing the input and output data matrix of the actual model and the input and output data matrix of the nominal model obtained in the step 2;
Step 4, constructing an evaluation function J by utilizing the full-metric residual error of the fault information based on the data driving designed in the step 3, and alarming when the detection index reaches an alarm threshold by combining with the fault detection alarm threshold;
Step 5, estimating the amplitude of the micro fault of the suspension system sensor by utilizing the fault information total measurement residual error based on the data driving designed in the step 3;
The process of the step 1 comprises the following steps:
Step 1.1. Discrete actual and nominal models G 0 of the vertical suspension system of a high-speed train can be expressed as:
Wherein A, B, C, E d,Ef,Ff are coefficient matrixes corresponding to the space state equation; x (k), y (k), u (k), d (k) are respectively state variables, output variables, control input variables and track disturbance excitation of an actual system; f (k) characterizes all possible faults; ζ (k) is process noise and measurement noise, respectively; x 0(k),y0(k),u0 (k) are the state variables, output variables and control input variables of the nominal model, respectively;
Step 1.2. From the state space equation (1) of the suspension system actual model G in step 1.1, there are:
Wherein y l(k),ul(k),dl(k),fl (k), Ζ l (k) is a stack matrix Γ l,Hu,l,Hd,l,Hf,l,/>Is a corresponding coefficient matrix, and is specifically formed as follows
Step 1.3 the expression (3) obtained in step 1.2 contains the state variable x (k), which is obtained from the discrete state space equation (1) in step 1.1 to eliminate the state variable x (k)
Wherein the method comprises the steps of
Step 1.4. Substituting the expression (5) obtained in step 1.3 into the expression (2) obtained in step 1.2, there are
Simplifying and obtaining
yl(k)≈Hu,p,lup,l(k)+Hd,p,ldp,l(k)+Hf,p,lfp,l(k)+He,p,lep,l(k) (8)
Wherein the method comprises the steps of
Considering the time interval N, from the reduced expression (8) obtained in step 1.4, a data model of a high-speed train suspension system can be obtained that implies the input-output relationship of the suspension system as follows:
Yk,l＝Hu,p,lUk,p,l+Hd,p,lDk,p,l+Hf,p,lFk,p,l+He,p,lEk,p,l (10)
Wherein Y k,l,Uk,p,l,Dk,p,l,Fk,p,l,Ek,p,l is a data matrix; the input and output data matrixes Y k,l and U k,p,l of the actual system can be obtained by measuring a sensor and a gyroscope when the train runs; the data matrix forms are as follows:
step 1.5 is similar, and from the state space equation (2) of the suspension system nominal model G 0 in step 1.1, there is:
Wherein the method comprises the steps of Input-output data matrix of nominal system respectively,/>Respectively have the same form as U k,p,l,Yk,l; and/>The sensor and the gyroscope can be used for measuring the running time of the train;
In the step 2, the obtained input/output signals of the actual model are used for constructing an input/output data matrix of the actual model, and the obtained input/output signals of the nominal model are used for constructing an input/output data matrix of the nominal model;
In the step 3, the specific process of designing the fault information total metric residual error based on data driving includes:
step 3.1 defining an output residual R Y for characterizing the output differences of the actual model and the nominal model
Step 3.2. define a controller residual R U for characterizing the output differences of the controller in the actual model and the nominal model
Step 3.3 defining fault information full metric residual error under closed loop control structure
Wherein,Characterizing the system output of the nominal model driven by the real-time input signal;
Step 3.4. Introduction of an orthogonal projection matrix In step 3.3, defining a fault information full-metric residual error under a closed-loop control structure, namely, simultaneously right multiplying the orthogonal projection matrix/>, on both sides of an equal sign of a formula (15)The following fault information full metric residuals based on data driving are available:
Wherein ToMFIR k is characterized in Thus, the task of micro fault diagnosis focuses on monitoring the variation of the residual signal matrix ToMFIR k that depends only on the input and output data of the system;
in the step4, the detection of the suspension system micro fault is performed by using the fault information total measurement residual error based on data driving, and the specific steps include:
Step 4.1. Residual Signal matrix ToMFIR k∈Rm×N may be written as Where τ i＝[τ1i,…,τ1N ], i=1, …, m; let/>Is a standardized form of ToMFIR k and satisfies/>Wherein/>
Step 4.2. decomposing the matrix ToMFIR k into a scoring matrix T and a loading matrix P by Singular Value Decomposition (SVD), assuming ToMFIR k is a Gaussian or near-Gaussian distribution, there areWherein Λ = diag (λ 1,…,λm); the score matrix T, the load matrix P and the feature matrix Λ may be derived from the following unbiased estimates:
Step 4.3. Introduction of a fault-free residual Signal The corresponding scoring matrices with the same dimensions as ToMFIR k are T and T rf, respectively; assuming that the average parameters of the distribution after the occurrence of the micro-fault are unchanged, there are
Step 4.4. Let T have a probability density of F, T rf have a probability density of F rf, based on the multidimensional KL divergence, the following evaluation function can be defined:
Step 4.5. The designed fault information total measurement residual error based on data driving and evaluation function are utilized, and the fault detection mechanism of the system is as follows
In the step 5, the amplitude estimation of the micro fault of the suspension system sensor is performed by using the fault information total measurement residual error based on data driving, and the specific steps include:
Step 5.1. Covariance matrix S is different from S rf due to the action of the micro-fault; the amplitude a of the micro fault is smaller and is an unknown constant value and is close to 0 but not equal to 0, so that the influence on signals is extremely weak; the effects caused by micro-faults can be converted into changes in the feature matrix, i.e
Step 5.2. For ToMFIR k, each score vector t i corresponds to a probability density functionPair/>Each score vector/>Corresponds to a probability density function/>Differences between t rf and t were measured using KL divergence:
wherein Δλ i is the characteristic value change caused by a minute failure;
Step 5.3. As a.fwdarw.0, Δλ i also tends to be 0, λ i Taylor is developed
From s=pΛp T, the n-order derivative of the eigenvalue λ i with respect to the fault amplitude a isWherein p i is/>Corresponding feature vectors;
Step 5.4. There is/>Wherein R Y,i＝[ri,1,ri,2,…,ri,N is row i of R Y; for minor failure faults of the sensor, the method can be used forA representation; and then obtain/>And its standardized form/>
Step 5.5. RecordIs a fault item,/>For non-faulty items, there are
All the constant parameters are irrelevant to fault amplitude and are calculated by fault-free reference data: δ rq =0,The first order derivative of the covariance matrix S to the fault amplitude a is
Second derivative of
The higher order derivative (n > 2) is 0;
step 5.6. Record vector p i＝[p1i…,pmi ], there is
The change in the characteristic value caused by the minute trouble can be expressed as
The KL divergence in step 5.2 can be expressed as
Wherein the method comprises the steps of
Step 5.7, utilizing the designed fault information total measurement residual error based on data driving, aiming at the tiny failure fault of the sensor of the suspension system of the high-speed train, the estimated value of the fault amplitude based on the divergence value is as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111347424.1A CN114115184B (en) | 2021-11-15 | 2021-11-15 | High-speed train suspension system fault diagnosis method based on data drive ToMFIR |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111347424.1A CN114115184B (en) | 2021-11-15 | 2021-11-15 | High-speed train suspension system fault diagnosis method based on data drive ToMFIR |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114115184A CN114115184A (en) | 2022-03-01 |
| CN114115184B true CN114115184B (en) | 2024-04-23 |
Family
ID=80395324
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111347424.1A Active CN114115184B (en) | 2021-11-15 | 2021-11-15 | High-speed train suspension system fault diagnosis method based on data drive ToMFIR |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114115184B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102607867A (en) * | 2012-02-22 | 2012-07-25 | 北京交通大学 | On-passage fault detection system based on GLRT (generalized likelihood ratio test) train suspension system and detection method of on-passage fault detection system |
| CN102768121A (en) * | 2012-07-31 | 2012-11-07 | 北京交通大学 | Method for detecting fault of train suspension system on basis of robust observer |
| CN102998130A (en) * | 2012-12-17 | 2013-03-27 | 北京交通大学 | Fault detecting method for train suspension system of data driving based on acceleration measuring |
| CN104155968A (en) * | 2014-07-17 | 2014-11-19 | 南京航空航天大学 | Tiny fault diagnosis method for final controlling element of high-speed train suspension system |
| CN104697804A (en) * | 2015-03-24 | 2015-06-10 | 清华大学 | Method and system for detecting and separating intermittent faults of train active suspension system |
| KR20170038348A (en) * | 2015-09-30 | 2017-04-07 | 한국전력공사 | Fault diagnosis system and fault diagnosis method in dynamic equipment |
| DE202021100535U1 (en) * | 2020-11-13 | 2021-02-16 | Beihang University | Fault diagnosis and maintenance device for a braking system of a high-speed train |
| CN113064401A (en) * | 2021-03-24 | 2021-07-02 | 重庆大学 | Closed loop system micro fault detection and estimation method based on data driving |
| CN113607443A (en) * | 2021-07-02 | 2021-11-05 | 江苏科技大学 | Early fault detection method for high-speed train suspension system based on data driving |
| CN113640607A (en) * | 2021-08-18 | 2021-11-12 | 江苏科技大学 | Early fault diagnosis method for inverter circuit and motor of high-speed train |
-
2021
- 2021-11-15 CN CN202111347424.1A patent/CN114115184B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102607867A (en) * | 2012-02-22 | 2012-07-25 | 北京交通大学 | On-passage fault detection system based on GLRT (generalized likelihood ratio test) train suspension system and detection method of on-passage fault detection system |
| CN102768121A (en) * | 2012-07-31 | 2012-11-07 | 北京交通大学 | Method for detecting fault of train suspension system on basis of robust observer |
| CN102998130A (en) * | 2012-12-17 | 2013-03-27 | 北京交通大学 | Fault detecting method for train suspension system of data driving based on acceleration measuring |
| CN104155968A (en) * | 2014-07-17 | 2014-11-19 | 南京航空航天大学 | Tiny fault diagnosis method for final controlling element of high-speed train suspension system |
| CN104697804A (en) * | 2015-03-24 | 2015-06-10 | 清华大学 | Method and system for detecting and separating intermittent faults of train active suspension system |
| KR20170038348A (en) * | 2015-09-30 | 2017-04-07 | 한국전력공사 | Fault diagnosis system and fault diagnosis method in dynamic equipment |
| DE202021100535U1 (en) * | 2020-11-13 | 2021-02-16 | Beihang University | Fault diagnosis and maintenance device for a braking system of a high-speed train |
| CN113064401A (en) * | 2021-03-24 | 2021-07-02 | 重庆大学 | Closed loop system micro fault detection and estimation method based on data driving |
| CN113607443A (en) * | 2021-07-02 | 2021-11-05 | 江苏科技大学 | Early fault detection method for high-speed train suspension system based on data driving |
| CN113640607A (en) * | 2021-08-18 | 2021-11-12 | 江苏科技大学 | Early fault diagnosis method for inverter circuit and motor of high-speed train |
Non-Patent Citations (1)
| Title |
|---|
| 数据驱动高速列车动态牵引系统的故障诊断;姜斌;陈宏田;易辉;陆宁云;中国科学(信息科学);20201231(第004期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114115184A (en) | 2022-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104155968B (en) | A kind of small fault diagnostic method for bullet train suspension system executor | |
| CN110308002B (en) | Urban rail train suspension system fault diagnosis method based on ground detection | |
| US10753827B2 (en) | Method and system for measuring vertical wheel impact force in real-time based on tire pressure monitoring | |
| CN110704801B (en) | Bridge cluster structure operation safety intelligent monitoring and rapid detection complete method | |
| CN112765778B (en) | Bogie lateral stability identification method and device and computer equipment | |
| CN102607867B (en) | On-passage fault detection system based on GLRT (generalized likelihood ratio test) train suspension system and detection method of on-passage fault detection system | |
| Rajamani et al. | A complete fault diagnostic system for automated vehicles operating in a platoon | |
| JP4298433B2 (en) | Railway vehicle abnormality detection device | |
| CN104458297B (en) | Train suspension system Transducer-fault Detecting Method based on non-linear stochastic model | |
| CN105403420B (en) | A kind of bogie method for diagnosing faults for merging multisensor | |
| CN106671724A (en) | Passive fault-tolerant control method of automotive active suspension | |
| CN104697804B (en) | For the method and system for detecting with separate train active suspension system intermittent fault | |
| US6687585B1 (en) | Fault detection and isolation system and method | |
| Capriglione et al. | ANN-based IFD in motorcycle rear suspension | |
| CN114115184B (en) | High-speed train suspension system fault diagnosis method based on data drive ToMFIR | |
| CN115032894A (en) | High-speed train suspension fault diagnosis method based on T-S fuzzy data driving ToMFIR | |
| CN104949848A (en) | Hunting instability detecting algorithm based on frequency domain kurtosis | |
| CN113607443A (en) | Early fault detection method for high-speed train suspension system based on data driving | |
| EP4008603B1 (en) | System for monitoring and predictive maintening the state of wear of mechanical components and operation method thereof | |
| CN114084199B (en) | Train stability evaluation method and system based on recursive graph analysis | |
| CN116481747A (en) | Bridge modal shape driving identification method considering bridge damping influence | |
| KR20190088933A (en) | A System for Diagnosing a Rail Train Based on a Plural of Parameters and a Method for Diagnosing the Same | |
| Imine et al. | Triangular observers for road profiles inputs estimation and vehicle dynamics analysis | |
| Nugraha et al. | Detection of railroad anomalies using machine learning approach | |
| CN103310044A (en) | Railway vehicle suspension system parameter estimation method based on improved particle filtering algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |