CN114607667A - Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation - Google Patents
Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation Download PDFInfo
- Publication number
- CN114607667A CN114607667A CN202210223251.0A CN202210223251A CN114607667A CN 114607667 A CN114607667 A CN 114607667A CN 202210223251 A CN202210223251 A CN 202210223251A CN 114607667 A CN114607667 A CN 114607667A
- Authority
- CN
- China
- Prior art keywords
- electro
- hydraulic servo
- servo valve
- port
- valve
- 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.)
- Withdrawn
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F15—FLUID-PRESSURE ACTUATORS; HYDRAULICS OR PNEUMATICS IN GENERAL
- F15B—SYSTEMS ACTING BY MEANS OF FLUIDS IN GENERAL; FLUID-PRESSURE ACTUATORS, e.g. SERVOMOTORS; DETAILS OF FLUID-PRESSURE SYSTEMS, NOT OTHERWISE PROVIDED FOR
- F15B13/00—Details of servomotor systems ; Valves for servomotor systems
- F15B13/02—Fluid distribution or supply devices characterised by their adaptation to the control of servomotors
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F15—FLUID-PRESSURE ACTUATORS; HYDRAULICS OR PNEUMATICS IN GENERAL
- F15B—SYSTEMS ACTING BY MEANS OF FLUIDS IN GENERAL; FLUID-PRESSURE ACTUATORS, e.g. SERVOMOTORS; DETAILS OF FLUID-PRESSURE SYSTEMS, NOT OTHERWISE PROVIDED FOR
- F15B13/00—Details of servomotor systems ; Valves for servomotor systems
- F15B13/02—Fluid distribution or supply devices characterised by their adaptation to the control of servomotors
- F15B13/023—Excess flow valves, e.g. for locking cylinders in case of hose burst
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F15—FLUID-PRESSURE ACTUATORS; HYDRAULICS OR PNEUMATICS IN GENERAL
- F15B—SYSTEMS ACTING BY MEANS OF FLUIDS IN GENERAL; FLUID-PRESSURE ACTUATORS, e.g. SERVOMOTORS; DETAILS OF FLUID-PRESSURE SYSTEMS, NOT OTHERWISE PROVIDED FOR
- F15B19/00—Testing; Calibrating; Fault detection or monitoring; Simulation or modelling of fluid-pressure systems or apparatus not otherwise provided for
- F15B19/005—Fault detection or monitoring
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F15—FLUID-PRESSURE ACTUATORS; HYDRAULICS OR PNEUMATICS IN GENERAL
- F15B—SYSTEMS ACTING BY MEANS OF FLUIDS IN GENERAL; FLUID-PRESSURE ACTUATORS, e.g. SERVOMOTORS; DETAILS OF FLUID-PRESSURE SYSTEMS, NOT OTHERWISE PROVIDED FOR
- F15B21/00—Common features of fluid actuator systems; Fluid-pressure actuator systems or details thereof, not covered by any other group of this subclass
- F15B21/04—Special measures taken in connection with the properties of the fluid
- F15B21/041—Removal or measurement of solid or liquid contamination, e.g. filtering
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F15—FLUID-PRESSURE ACTUATORS; HYDRAULICS OR PNEUMATICS IN GENERAL
- F15B—SYSTEMS ACTING BY MEANS OF FLUIDS IN GENERAL; FLUID-PRESSURE ACTUATORS, e.g. SERVOMOTORS; DETAILS OF FLUID-PRESSURE SYSTEMS, NOT OTHERWISE PROVIDED FOR
- F15B21/00—Common features of fluid actuator systems; Fluid-pressure actuator systems or details thereof, not covered by any other group of this subclass
- F15B21/04—Special measures taken in connection with the properties of the fluid
- F15B21/042—Controlling the temperature of the fluid
- F15B21/0423—Cooling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Fluid Mechanics (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Fluid-Pressure Circuits (AREA)
Abstract
The invention provides a system and a method for diagnosing faults of an electro-hydraulic servo valve based on characteristic distillation, wherein the system for diagnosing faults of the electro-hydraulic servo valve comprises an upper computer device, a data acquisition device and a hydraulic device which are sequentially connected; the upper computer device comprises a test software platform and a fault diagnosis module; the data acquisition device comprises a data acquisition card and a signal generator; the data acquisition card is in signal connection with the signal generator; the hydraulic device comprises an oil tank, a constant delivery pump, a filter, an overflow valve, a one-way valve, a first pressure sensor, a second pressure sensor, an electro-hydraulic servo valve, a third pressure sensor, a proportional throttle valve, a fourth pressure sensor, a flow sensor and a cooler; the device is adopted by the invention, so that the number of monitoring sensors needed by key components such as an electro-hydraulic servo valve and the like is correspondingly reduced, the weight of the airplane is reduced, the performance of the airplane is improved, and the oil consumption is reduced.
Description
Technical Field
The invention belongs to the technical field of electro-hydraulic servo valve fault diagnosis, and particularly relates to a system and a method for diagnosing faults of an electro-hydraulic servo valve based on characteristic distillation.
Background
The electro-hydraulic servo valve is a key hydraulic component for industrial application and aerospace, and is also the part with the highest failure rate in a hydraulic system, and the working performance of the system is determined by the use condition of the electro-hydraulic servo valve. The faults of the electro-hydraulic servo valve are usually represented by mechanical faults, electrical faults and hydraulic faults which are interwoven together, so that the fault phenomenon and the fault reason are not simple linear corresponding relations but are represented by serious nonlinear mapping relations, when the faults occur, the control precision and the stability of a system can be generally deteriorated, and the system can be failed in serious conditions.
Particularly, for an airborne electro-hydraulic servo valve of an airplane, in the working process, the electro-hydraulic servo valve is in extreme environments of high temperature, high pressure, strong vibration, high dynamic state and the like, so that the acquired signals can be greatly interfered, effective information is easily submerged in noise, and the signal acquisition and analysis processing of the electro-hydraulic servo valve are extremely difficult.
With the increasing complexity of the aircraft hydraulic system, in order to ensure the safety and reliability of the aircraft hydraulic system, the number of monitoring sensors required by key components such as an electro-hydraulic servo valve and the like is correspondingly increased, so that the weight of the aircraft is increased, and a series of problems such as performance reduction and oil consumption increase of the aircraft are caused.
Disclosure of Invention
The invention provides a system and a method for diagnosing faults of an electro-hydraulic servo valve based on characteristic distillation, aiming at the defects in the prior art.
In a first aspect, the invention provides a feature distillation-based electro-hydraulic servo valve fault diagnosis system, which comprises an upper computer device, a data acquisition device and a hydraulic device which are sequentially connected;
the upper computer device comprises a test software platform and a fault diagnosis module;
the data acquisition device comprises a data acquisition card and a signal generator; the data acquisition card is in signal connection with the signal generator;
the hydraulic device comprises an oil tank, a constant delivery pump, a filter, an overflow valve, a one-way valve, a first pressure sensor, a second pressure sensor, an electro-hydraulic servo valve, a third pressure sensor, a proportional throttle valve, a fourth pressure sensor, a flow sensor and a cooler;
an oil suction port of the fixed displacement pump is connected with the oil tank, and an oil pressing port of the fixed displacement pump is connected with an oil inlet of the filter; the oil outlet of the filter is respectively connected with the oil inlet of the one-way valve and the oil outlet of the overflow valve; an oil inlet of the overflow valve is connected with the oil tank; an oil outlet of the one-way valve is respectively connected with one end of the first pressure sensor and a P port of the electro-hydraulic servo valve; a T port of the electro-hydraulic servo valve is respectively connected with one end of the second pressure sensor and one end of the flow sensor; one end of the flow sensor, which is far away from the electro-hydraulic servo valve, is connected with one end of the cooler; one end of the cooler, which is far away from the flow sensor, is connected with the oil tank; the port A of the electro-hydraulic servo valve is respectively connected with one end of the third pressure sensor and one end of the proportional throttle valve, and the port B of the electro-hydraulic servo valve is respectively connected with one end of the fourth pressure sensor and one end of the proportional throttle valve, which is far away from the port A of the electro-hydraulic servo valve; the first pressure sensor, the second pressure sensor, the third pressure sensor, the fourth pressure sensor and the flow sensor are all in signal connection with the data acquisition card; the signal generator is connected with a control port of the electro-hydraulic servo valve;
the quantitative pump sucks oil from the oil tank and reaches the port P of the electro-hydraulic servo valve through the filter and the one-way valve in sequence; the oil liquid at the T port of the electro-hydraulic servo valve flows back to the oil tank through the flow sensor and the cooler; the first pressure sensor, the second pressure sensor, the third pressure sensor and the fourth pressure sensor respectively collect the pressure of a P port, a T port, an A port and a B port of the electro-hydraulic servo valve; the flow sensor acquires the outlet flow of the T port of the electro-hydraulic servo valve; the data acquisition card outputs control signals to control the action of the electro-hydraulic servo valve through the signal generator by acquiring signals of the flow sensor, the first pressure sensor, the second pressure sensor, the third pressure sensor and the fourth pressure sensor; when the output signal of the data acquisition card is '+', the electro-hydraulic servo valve works at the left position, and the oil liquid at the port P of the electro-hydraulic servo valve returns to the oil tank through the port A of the electro-hydraulic servo valve, the proportional throttle valve, the port B of the electro-hydraulic servo valve, the port T of the electro-hydraulic servo valve, the flow sensor and the cooler in sequence; when the output signal of the data acquisition card is negative, the electro-hydraulic servo valve works at the right position, and the oil liquid at the port P of the electro-hydraulic servo valve returns to the oil tank through the port B of the electro-hydraulic servo valve, the proportional throttle valve, the port A of the electro-hydraulic servo valve, the port T of the electro-hydraulic servo valve, the flow sensor and the cooler in sequence; and the data acquisition card transmits signals obtained from the first pressure sensor, the second pressure sensor, the third pressure sensor, the fourth pressure sensor and the flow sensor to the test software platform to display the state of the electro-hydraulic servo valve, and the fault diagnosis module is embedded into the test software platform.
In a second aspect, the present invention provides a method for diagnosing a fault of an electro-hydraulic servo valve based on characteristic distillation, which is applied to the system for diagnosing a fault of an electro-hydraulic servo valve according to the first aspect, and the method for diagnosing a fault includes:
adjusting the pressure of the electro-hydraulic servo valve fault diagnosis system to the rated pressure drop of the electro-hydraulic servo valve;
acquiring fault data of the electro-hydraulic servo valve;
constructing data nodes, wherein the data nodes comprise displacement feedback of an electro-hydraulic servo valve core at a target moment, displacement feedback of an electro-hydraulic servo valve at the target moment, flow at the target moment, valve inlet pressure of the electro-hydraulic servo valve at the target moment, pressure of a load A port of the electro-hydraulic servo valve at the target moment, pressure of a load B port of the electro-hydraulic servo valve at the target moment and pressure of an outlet of the electro-hydraulic servo valve at the target moment;
connecting all the data nodes pairwise, constructing full connection, and obtaining a state vector set representing the characteristics of the electro-hydraulic servo valve, wherein elements of the state vector set collapse and change to different characteristic dimensions to obtain different static characteristics of the electro-hydraulic servo valve;
a multi-head attention mechanism is introduced into a graph convolution network by utilizing a graph convolution operation of full self attention, and a mathematical model of the graph convolution network can be expressed as follows according to a paradigm of a message propagation network:
a message function M:
αijcan be expressed as:
wherein M is data processed by using weights through a neural network, and comprises the processed flow, the pressure of an electro-hydraulic servo valve P, an electro-hydraulic servo valve A, an electro-hydraulic servo valve B and the displacement of a valve core; h isjInputting the j neural network, including flow, electro-hydraulic servo valve P, electro-hydraulic servo valve A, electro-hydraulic servo valve B port pressure and valve core displacement; w is a parameter matrix for training; alpha is alphaijIs a correlation coefficient, i.e. weight; d is electro-hydraulic servo valve displacement feedback;
applying the weight obtained by teacher network pre-training to student network self-training, wherein the student network comprises a convolution layer, a batch standardization layer and a modified linear unit activation function;
the convolution layer is as follows:
in the formula, KlA one-dimensional convolution kernel for the l-th layer;the ith convolved local region of the ith layer;
the batch standardization layer is as follows:
in the formula (I), the compound is shown in the specification,is the average of the inputs of layer l;is the standard deviation of the input for layer l; epsilon, gamma and beta are constant values;
the modified linear cell activation function is:
in the formula, yiIs output data; h isiInputting data; alpha is alphaiTaking 100 as the hyperparameter;
performing regression of the global feature vector of the electro-hydraulic servo valve and fault classification of the electro-hydraulic servo valve according to the feature distillation model:
in the formula, LossFIs the regression loss to the global feature vector; lossPClassifying the loss for the fault; α is the distillation loss weight; beta is the predicted loss weight;
regression loss of global feature vectorsThe method minimizes the difference between the global feature vector output by the backbone network in the student network and the global feature vector output by the backbone network in the teacher network, and the fault classification lossThe difference between the class prediction of the student network on fault data and a real label is minimized;
the student network fits the global feature vector of the electro-hydraulic servo valve, and the minimum mean square error is adopted as a loss function:
in the classification of electrohydraulic servo valve faults, the probability distribution of the faults needs to be predicted, and cross entropy is adopted as a loss function:
further, the adjusting the pressure of the electro-hydraulic servo valve fault diagnosis system to the rated pressure drop of the electro-hydraulic servo valve comprises:
normalizing the pressure data according to the following formula:
wherein, Δ pnIs the rated pressure drop of the electro-hydraulic servo valve; p is the output pressure of the electro-hydraulic servo valve; p' is the pressure normalization of the electro-hydraulic servo valve;
the flow data is normalized according to the following formula:
wherein q isnIs the load flow under the rated pressure drop of the electro-hydraulic servo valve; q is the output flow of the electro-hydraulic servo valve; q' is the flow standardization of the electro-hydraulic servo valve;
normalizing the electro-hydraulic servo valve command signal according to the following formula:
wherein s isnThe signal amplitude is the maximum input electrical signal of the electro-hydraulic servo valve when the maximum input electrical signal is a rated test signal; s is the displacement feedback of the valve core of the electro-hydraulic servo valve; s' normalization of command signals of the electro-hydraulic servo valve;
normalizing electro-hydraulic servo valve displacement feedback according to the following formula:
wherein, dmaxThe maximum positive displacement feedback of the output maximum value electric signal of the electro-hydraulic servo valve displacement feedback sensor is realized; d' is the standardization of the displacement feedback of the electro-hydraulic servo valve; d is electro-hydraulic servo valve displacement feedback; dminAnd feeding back the maximum negative displacement of the output minimum electric signal of the electro-hydraulic servo valve displacement feedback sensor.
The invention provides a system and a method for diagnosing faults of an electro-hydraulic servo valve based on characteristic distillation, wherein the system for diagnosing faults of the electro-hydraulic servo valve comprises an upper computer device, a data acquisition device and a hydraulic device which are sequentially connected; the upper computer device comprises a test software platform and a fault diagnosis module; the data acquisition device comprises a data acquisition card and a signal generator; the data acquisition card is in signal connection with the signal generator; the hydraulic device comprises an oil tank, a constant delivery pump, a filter, an overflow valve, a one-way valve, a first pressure sensor, a second pressure sensor, an electro-hydraulic servo valve, a third pressure sensor, a proportional throttle valve, a fourth pressure sensor, a flow sensor and a cooler; an oil suction port of the fixed displacement pump is connected with the oil tank, and an oil pressing port of the fixed displacement pump is connected with an oil inlet of the filter; the oil outlet of the filter is respectively connected with the oil inlet of the one-way valve and the oil outlet of the overflow valve; an oil inlet of the overflow valve is connected with the oil tank; an oil outlet of the one-way valve is respectively connected with one end of the first pressure sensor and a P port of the electro-hydraulic servo valve; a T port of the electro-hydraulic servo valve is respectively connected with one end of the second pressure sensor and one end of the flow sensor; one end of the flow sensor, which is far away from the electro-hydraulic servo valve, is connected with one end of the cooler; one end of the cooler, which is far away from the flow sensor, is connected with the oil tank; the port A of the electro-hydraulic servo valve is respectively connected with one end of the third pressure sensor and one end of the proportional throttle valve, and the port B of the electro-hydraulic servo valve is respectively connected with one end of the fourth pressure sensor and one end of the proportional throttle valve, which is far away from the port A of the electro-hydraulic servo valve; the first pressure sensor, the second pressure sensor, the third pressure sensor, the fourth pressure sensor and the flow sensor are all in signal connection with the data acquisition card; and the signal generator is connected with a control port of the electro-hydraulic servo valve. By adopting the device, the number of monitoring sensors needed by key components such as an electro-hydraulic servo valve and the like is correspondingly reduced, so that the weight of the airplane is reduced, the performance of the airplane is improved, and the oil consumption is reduced.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a schematic structural diagram of a system for diagnosing a fault of an electro-hydraulic servo valve based on characteristic distillation according to an embodiment of the present invention;
FIG. 2 is a block diagram of a teacher network provided by an embodiment of the present invention;
fig. 3 is a structure diagram of a lightweight student network according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a method for diagnosing a fault of an electro-hydraulic servo valve based on characteristic distillation according to an embodiment of the present invention;
fig. 5 is a schematic diagram of the effect of fault diagnosis.
Wherein, 1-testing a software platform; 2-a fault diagnosis module; 3-a data acquisition card; 4-a signal generator; 5-an oil tank; 6-a fixed displacement pump; 7-a filter; 8-relief valves; 9-a one-way valve; 10-a first pressure sensor; 11-a second pressure sensor; 12-an electro-hydraulic servo valve; 13-a third pressure sensor; 14-proportional throttle valve; 15-a fourth pressure sensor; 16-a flow sensor; 18-cooler.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
As shown in fig. 1, an embodiment of the present invention provides, in part, an electro-hydraulic servo valve fault diagnosis system based on characteristic distillation, including an upper computer device, a data acquisition device, and a hydraulic device, which are connected in sequence.
The upper computer device comprises a test software platform 1 and a fault diagnosis module 2.
The data acquisition device comprises a data acquisition card 3 and a signal generator 4; the data acquisition card 3 is in signal connection with the signal generator 4.
The hydraulic device comprises an oil tank 5, a fixed displacement pump 6, a filter 7, an overflow valve 8, a check valve 9, a first pressure sensor 10, a second pressure sensor 11, an electro-hydraulic servo valve 12, a third pressure sensor 13, a proportional throttle valve 14, a fourth pressure sensor 15, a flow sensor 16 and a cooler 18.
An oil suction port of the constant delivery pump 6 is connected with the oil tank 5, and an oil pressing port is connected with an oil inlet of the filter 7; an oil outlet of the filter 7 is respectively connected with an oil inlet of the one-way valve 9 and an oil outlet of the overflow valve 8; an oil inlet of the overflow valve 8 is connected with the oil tank 5; an oil outlet of the one-way valve 9 is respectively connected with one end of a first pressure sensor 10 and a P port of an electro-hydraulic servo valve 12; a T port of the electro-hydraulic servo valve 12 is connected with one end of the second pressure sensor 11 and one end of the flow sensor 16 respectively; the end of the flow sensor 16 far away from the electro-hydraulic servo valve 12 is connected with one end of a cooler 18; the end of the cooler 18 far away from the flow sensor 16 is connected with the oil tank 5; the port A of the electro-hydraulic servo valve 12 is respectively connected with one end of a third pressure sensor 13 and one end of a proportional throttle valve 14, and the port B is respectively connected with one end of a fourth pressure sensor 15 and one end of the proportional throttle valve 14 far away from the port A of the electro-hydraulic servo valve 12; the first pressure sensor 10, the second pressure sensor 11, the third pressure sensor 13, the fourth pressure sensor 15 and the flow sensor 16 are in signal connection with the data acquisition card 3; the signal generator 4 is connected with a control port of the electro-hydraulic servo valve 12.
The constant delivery pump 6 sucks oil from the oil tank 5 and reaches the port 12P of the electro-hydraulic servo valve through the filter 7 and the one-way valve 9 in sequence; the oil liquid at the port of the electro-hydraulic servo valve 12T flows back to the oil tank 5 through the flow sensor 16 and the cooler 18; the first pressure sensor 10, the second pressure sensor 11, the third pressure sensor 13 and the fourth pressure sensor 15 respectively collect the pressure of the port P, the port T, the port A and the port B of the electro-hydraulic servo valve 12; the flow sensor 16 acquires the outlet flow of the T port of the electro-hydraulic servo valve 12; the data acquisition card 3 outputs control signals to control the action of the electro-hydraulic servo valve 12 through the signal generator 4 by acquiring signals of the flow sensor 16, the first pressure sensor 10, the second pressure sensor 11, the third pressure sensor 13 and the fourth pressure sensor 15; when the output signal of the data acquisition card 3 is "+", the electro-hydraulic servo valve 12 works at the left position, and the oil liquid at the port 12P of the electro-hydraulic servo valve returns to the oil tank 5 through the port 12A of the electro-hydraulic servo valve, the proportional throttle valve 14, the port 12B of the electro-hydraulic servo valve, the port 12T of the electro-hydraulic servo valve, the flow sensor 16 and the cooler 18 in sequence; when the output signal of the data acquisition card 3 is negative, the electro-hydraulic servo valve 12 works at the right position, and the oil liquid at the port 12P of the electro-hydraulic servo valve returns to the oil tank 5 through the port 12B of the electro-hydraulic servo valve, the proportional throttle valve 14, the port 12A of the electro-hydraulic servo valve, the port 12T of the electro-hydraulic servo valve, the flow sensor 16 and the cooler 18 in sequence; signals obtained by the data acquisition card 3 from the first pressure sensor 10, the second pressure sensor 11, the third pressure sensor 13, the fourth pressure sensor 15 and the flow sensor 16 are transmitted to the test software platform 1 to display the state of the electro-hydraulic servo valve 12, and the fault diagnosis module 2 is embedded into the LabView test software platform 1.
As shown in fig. 2, an embodiment of the present invention provides, in part, an electro-hydraulic servo valve fault diagnosis method based on characteristic distillation, which is applied to the electro-hydraulic servo valve fault diagnosis system, and the fault diagnosis method includes:
and step S101, adjusting the pressure of the electro-hydraulic servo valve fault diagnosis system to the rated pressure drop of the electro-hydraulic servo valve.
In this step, in order to standardize the collected fault data of the electrohydraulic servo valve, the pressure of the fault diagnosis system of the electrohydraulic servo valve is usually adjusted to the rated pressure drop of the electrohydraulic servo valve during testing, so that the load flow under the rated pressure drop of the electrohydraulic servo valve can be found according to the model of the electrohydraulic servo valve, and the pressure data and the flow data are standardized:
normalizing the pressure data according to the following formula:
wherein, Δ pnIs the rated pressure drop of the electro-hydraulic servo valve; p is the output pressure of the electro-hydraulic servo valve; p' is the pressure normalization of the electro-hydraulic servo valve.
The flow data is normalized according to the following formula:
wherein q isnIs the load flow under the rated pressure drop of the electro-hydraulic servo valve; q is the output flow of the electro-hydraulic servo valve; q' is the flow normalization of the electro-hydraulic servo valve.
Normalizing the electro-hydraulic servo valve command signal according to the following formula:
wherein the content of the first and second substances,snthe signal amplitude value is the maximum input electric signal of the electro-hydraulic servo valve and is a rated test signal; s is the displacement feedback of the valve core of the electro-hydraulic servo valve; s' standardization of command signals of an electro-hydraulic servo valve;
normalizing electro-hydraulic servo valve displacement feedback according to the following formula:
wherein, dmaxThe maximum positive displacement feedback of the output maximum value electric signal of the electro-hydraulic servo valve displacement feedback sensor is realized; d' is the standardization of the displacement feedback of the electro-hydraulic servo valve; d is electro-hydraulic servo valve displacement feedback; dminAnd feeding back the maximum negative displacement of the output minimum electric signal of the electro-hydraulic servo valve displacement feedback sensor.
The standardization of displacement feedback is to mark the maximum positive displacement of the electro-hydraulic servo valve as 1 and the maximum negative displacement as 0, and the displacement of the valve core can return to (0,1) after standardization, so that the same data in different batches can be prevented from being standardized to (0,1) to obtain different characteristics after standardization when no full-scale valve core displacement exists in the whole batch.
Step S102, collecting fault data of the electro-hydraulic servo valve.
In the step, for the fault data of the electro-hydraulic servo valve, data more than one period is extracted by combining downsampling and a sliding window for data enhancement. The analysis is carried out according to the data in the operation process of the electro-hydraulic servo valve, and the obvious characteristics of most data do not exist in a high frequency band, so that the down-sampling processing can be carried out on the data. Under the condition of ensuring that the total length of the data is not changed, one group of data is decomposed into a plurality of groups of data through equidistant extraction and splitting so as to increase the data sample size. After down-sampling is completed, sliding window extraction can be performed on the operation data in a certain time period, data can be generalized, and the influence of the phase of periodic data on the application of the network model after deployment is avoided.
Step S103, constructing data nodes, wherein the data nodes comprise displacement feedback of an electro-hydraulic servo valve core at a target moment, displacement feedback of an electro-hydraulic servo valve at the target moment, flow at the target moment, valve inlet pressure of the electro-hydraulic servo valve at the target moment, pressure of a load A port of the electro-hydraulic servo valve at the target moment, pressure of a load B port of the electro-hydraulic servo valve at the target moment and pressure of an outlet of the electro-hydraulic servo valve at the target moment.
In the step, data nodes are required to be constructed after data standardization, each data node represents the characteristics of an electro-hydraulic servo valve fault diagnosis system at a certain moment, and the data nodes of the electro-hydraulic servo valve can be represented ass(i)Feeding back data for the displacement of the valve core of the electro-hydraulic servo valve at the moment i; d(i)Feedback is carried out on the displacement of the electro-hydraulic servo valve at the moment i; f. of(i)The flow at the moment i;valve inlet pressure at time i for the electro-hydraulic servo valve;pressure data of a load A port of the electro-hydraulic servo valve at the moment i;pressure data of a load port B of the electro-hydraulic servo valve at the moment i;the pressure data at the outlet of the electro-hydraulic servo valve at the moment i. In this node vector, some characteristics of the i-time electro-hydraulic servo valve are already included.
And S104, connecting all the data nodes pairwise to construct full connection to obtain a state vector set representing the characteristics of the electro-hydraulic servo valve, wherein elements of the state vector set collapse and change to different characteristic dimensions to obtain different static characteristics of the electro-hydraulic servo valve.
In the step, test data are constructed and fully connected according to a method of connecting all nodes in pairs to form a graph, a state vector set describing the characteristics of the electro-hydraulic servo valve is obtained, different static characteristics of the servo valve can be obtained by collapsing and transforming the set elements to different characteristic dimensions, the data set can contain more characteristics compared with a single sensor data set, and high-order characteristics can be learned according to guidance of data set labels through propagation and learning among the nodes.
Step S105, using a graph convolution operation with full attention to introduce a multi-point attention mechanism into a graph convolution network, and according to a paradigm of a message propagation network, a mathematical model thereof can be expressed as:
a message function M:
αijcan be expressed as:
wherein M is data processed by using weights through a neural network, and comprises the processed flow, the pressure of an electro-hydraulic servo valve P, an electro-hydraulic servo valve A, an electro-hydraulic servo valve B and the displacement of a valve core; h isjInputting the j neural network, including flow, electro-hydraulic servo valve P, electro-hydraulic servo valve A, electro-hydraulic servo valve B port pressure and valve core displacement; w is a parameter matrix for training; alpha (alpha) ("alpha")ijIs a correlation coefficient, i.e. weight; d is the displacement feedback of the electro-hydraulic servo valve.
As shown in fig. 2, the teacher network is used to pre-train the neural network input to obtain the correlation coefficient, and to transfer the correlation coefficient to the student network to guide the student network to converge quickly and accurately.
The backbone network of the teacher network is derived based on a graph neural network model. The Transformer-GCN model implicitly learns the implicit relationship between system state vectors when the electro-hydraulic servo valve is in different health states through the interaction of an adjacency matrix and an attention mechanism, and finally obtains global feature vectors describing the electro-hydraulic servo valve in different health states through multilayer graph convolution operation and pooling operation based on a residual error model. The fault diagnosis of the electro-hydraulic servo valve is realized through a multilayer full-connection network.
The classification network of the teacher network adopts a multilayer perceptron Model (MLP) of a graph neural network model, wherein the MLP layer comprises three Layers: the MLP neural network comprises an input layer, a hidden layer and an output layer, wherein different layers of the MLP neural network are fully connected. The input of the MLP layer is a global feature vector, and the output is a classification result. The MLP layer is responsible for classifying global feature vectors in feature distillation.
And step S106, applying the weight obtained by the teacher network pre-training to student network self-training, wherein the student network comprises a convolution layer, a batch standardization layer and a modified linear unit activation function.
As shown in fig. 3, a light-weighted student backbone network model adopts a one-dimensional basic unit of Darknet19, and a light-weighted student network applies weights obtained by teacher network pre-training to self-training, so that convergence can be faster and better.
The convolution layer is as follows:
in the formula, KlA one-dimensional convolution kernel for the l-th layer;the ith convolved local region of the ith layer;
the batch standardization layer is as follows:
in the formula (I), the compound is shown in the specification,is the average of the inputs of the l-th layer;is the standard deviation of the input for layer l; epsilon, gamma and beta are constant values;
the modified linear cell activation function is:
in the formula, yiIs output data; h isiInputting data; alpha is alphaiFor the hyper-parameter, take 100.
On the basis of the basic units, four convolution sets are constructed, wherein the first layer of convolution set is composed of a layer of basic unit and a layer of maximum pooling layer, the other three convolution sets are composed of three layers of basic units and a layer of maximum pooling layer, and the four convolution sets and the layer of maximum pooling layer are all connected to form a backbone network of the student network.
The backbone network parameters of the student network are shown in table 1. The backbone network of the student network performs 4-time 2-time downsampling through 4-time pooling effect of convolution concentration, 16-time downsampling is achieved in total, and finally the function of extracting the global feature vector from the fault data of the electro-hydraulic servo valve is achieved through the full connection layer.
TABLE 1 student backbone network parameters
The classification layer of the student network adopts an MLP network with the same structure as the teacher network.
Step S107, carrying out regression of the global feature vector of the electro-hydraulic servo valve and fault classification of the electro-hydraulic servo valve according to the feature distillation model:
in the formula, LossFIs the regression loss to the global feature vector; lossPClassifying the loss for the fault; α is the distillation loss weight; beta is the predicted loss weight;
regression loss of global feature vectorsThe method minimizes the difference between the global feature vector output by the backbone network in the student network and the global feature vector output by the backbone network in the teacher network, and the fault classification lossThe difference between the class prediction of the student network on fault data and a real label is minimized;
the student network fits the global feature vector of the electro-hydraulic servo valve, and the minimum mean square error is adopted as a loss function:
in the classification of electrohydraulic servo valve faults, the probability distribution of the faults needs to be predicted, and cross entropy is adopted as a loss function:
as shown in fig. 4, the method for diagnosing the fault of the electro-hydraulic servo valve based on the characteristic distillation further solves the problem that an industrial personal computer deployed on site does not have huge computing power, selects a light-weight convolutional neural network, and optimizes a fault diagnosis algorithm of the neural network by the characteristic distillation method. In the training of a teacher network, the graph convolution network performs feature reading on the electro-hydraulic servo valve fault data after convolution and pooling to obtain a global feature vector for expressing the electro-hydraulic servo valve, and the global feature vector realizes the classification of the fault data through a multi-layer fully-connected network. When the student network is trained, the global feature vector read out by the teacher network is used as a global feature vector label, then the lightweight convolution network of the student network is fully connected with the single layer to learn the global feature vector, and feature distillation is carried out on the global feature vector. And simultaneously, transferring the weight of the multilayer perceptron of the teacher network to the multilayer full connection of the student network, and carrying out network fine adjustment on the full connection network.
As shown in FIG. 5, it can be seen that the minimum error converges on about 10-3, the accuracy at the minimum error is 1.000, and the training result of the model achieves the expected effect.
The invention has been described in detail with reference to specific embodiments and illustrative examples, but the description is not intended to be construed in a limiting sense. Those skilled in the art will appreciate that various equivalent substitutions, modifications or improvements may be made to the technical solution of the present invention and its embodiments without departing from the spirit and scope of the present invention, which fall within the scope of the present invention. The scope of the invention is defined by the appended claims.
Claims (3)
1. A fault diagnosis system of an electro-hydraulic servo valve based on characteristic distillation is characterized by comprising an upper computer device, a data acquisition device and a hydraulic device which are sequentially connected;
the upper computer device comprises a test software platform (1) and a fault diagnosis module (2);
the data acquisition device comprises a data acquisition card (3) and a signal generator (4); the data acquisition card (3) is in signal connection with the signal generator (4);
the hydraulic device comprises an oil tank (5), a fixed displacement pump (6), a filter (7), an overflow valve (8), a one-way valve (9), a first pressure sensor (10), a second pressure sensor (11), an electro-hydraulic servo valve (12), a third pressure sensor (13), a proportional throttle valve (14), a fourth pressure sensor (15), a flow sensor (16) and a cooler (18);
an oil suction port of the fixed displacement pump (6) is connected with the oil tank (5), and an oil pressing port is connected with an oil inlet of the filter (7); an oil outlet of the filter (7) is respectively connected with an oil inlet of the one-way valve (9) and an oil outlet of the overflow valve (8); an oil inlet of the overflow valve (8) is connected with the oil tank (5); an oil outlet of the one-way valve (9) is respectively connected with one end of the first pressure sensor (10) and a P port of the electro-hydraulic servo valve (12); a T port of the electro-hydraulic servo valve (12) is respectively connected with one end of the second pressure sensor (11) and one end of the flow sensor (16); the end of the flow sensor (16) far away from the electro-hydraulic servo valve (12) is connected with one end of the cooler (18); the end of the cooler (18) far away from the flow sensor (16) is connected with the oil tank (5); an A port of the electro-hydraulic servo valve (12) is respectively connected with one end of the third pressure sensor (13) and one end of the proportional throttle valve (14), and a B port of the electro-hydraulic servo valve (12) is respectively connected with one end of the fourth pressure sensor (15) and one end of the proportional throttle valve (14) far away from the A port of the electro-hydraulic servo valve (12); the first pressure sensor (10), the second pressure sensor (11), the third pressure sensor (13), the fourth pressure sensor (15) and the flow sensor (16) are in signal connection with the data acquisition card (3); the signal generator (4) is connected with a control port of the electro-hydraulic servo valve (12);
the fixed displacement pump (6) sucks oil from the oil tank (5) and reaches a port P of the electro-hydraulic servo valve (12) through the filter (7) and the one-way valve (9) in sequence; the oil liquid at the T port of the electro-hydraulic servo valve (12) flows back to the oil tank (5) through the flow sensor (16) and the cooler (18); the first pressure sensor (10), the second pressure sensor (11), the third pressure sensor (13) and the fourth pressure sensor (15) respectively collect the pressure of a P port, a T port, an A port and a B port of the electro-hydraulic servo valve (12); the flow sensor (16) collects the outlet flow of a T port of the electro-hydraulic servo valve (12); the data acquisition card (3) outputs control signals to control the action of the electro-hydraulic servo valve (12) through the signal generator (4) by acquiring signals of the flow sensor (16), the first pressure sensor (10), the second pressure sensor (11), the third pressure sensor (13) and the fourth pressure sensor (15); when the output signal of the data acquisition card (3) is '+', the electro-hydraulic servo valve (12) works at the left position, and the oil liquid at the P port of the electro-hydraulic servo valve (12) returns to the oil tank (5) through the A port of the electro-hydraulic servo valve (12), the proportional throttle valve (14), the B port of the electro-hydraulic servo valve (12), the T port of the electro-hydraulic servo valve (12), the flow sensor (16) and the cooler (18) in sequence; when the output signal of the data acquisition card (3) is negative, the electro-hydraulic servo valve (12) works at the right position, and the oil liquid at the P port of the electro-hydraulic servo valve (12) returns to the oil tank (5) through an electro-hydraulic servo valve port B, a proportional throttle valve (14), an electro-hydraulic servo valve port A (12), an electro-hydraulic servo valve port T (12), a flow sensor (16) and a cooler (18) in sequence; the data acquisition card (3) transmits signals obtained from the first pressure sensor (10), the second pressure sensor (11), the third pressure sensor (13), the fourth pressure sensor (15) and the flow sensor (16) to the test software platform (1) to display the state of the electro-hydraulic servo valve (12), and the fault diagnosis module (2) is embedded into the test software platform (1).
2. An electro-hydraulic servo valve fault diagnosis method based on characteristic distillation, which is applied to the electro-hydraulic servo valve fault diagnosis system of claim 1, and is characterized by comprising the following steps:
adjusting the pressure of the electro-hydraulic servo valve fault diagnosis system to the rated pressure drop of the electro-hydraulic servo valve;
acquiring fault data of the electro-hydraulic servo valve;
constructing data nodes, wherein the data nodes comprise displacement feedback of an electro-hydraulic servo valve core at a target moment, displacement feedback of an electro-hydraulic servo valve at the target moment, flow at the target moment, valve inlet pressure of the electro-hydraulic servo valve at the target moment, pressure of a load A port of the electro-hydraulic servo valve at the target moment, pressure of a load B port of the electro-hydraulic servo valve at the target moment and pressure of an outlet of the electro-hydraulic servo valve at the target moment;
connecting all the data nodes pairwise, constructing full connection, and obtaining a state vector set representing the characteristics of the electro-hydraulic servo valve, wherein elements of the state vector set collapse and change to different characteristic dimensions to obtain different static characteristics of the electro-hydraulic servo valve;
a multi-head attention mechanism is introduced into a graph convolution network by utilizing a graph convolution operation of full self attention, and a mathematical model of the graph convolution network can be expressed as follows according to a paradigm of a message propagation network:
a message function M:
αijcan be expressed as:
wherein M is data processed by using weight through a neural network, and comprises the processed flow, the pressure of an electro-hydraulic servo valve P, an electro-hydraulic servo valve A, an electro-hydraulic servo valve B port and valve core displacement; h isjInputting the j-th neural network, including flow, electro-hydraulic servo valve P, electro-hydraulic servo valve A, electro-hydraulic servo valve B port pressure and valve core displacement; w is a parameter matrix for training; alpha is alphaijIs a correlation coefficient, i.e. weight; d is electro-hydraulic servo valve displacement feedback;
applying the weight obtained by teacher network pre-training to student network self-training, wherein the student network comprises a convolution layer, a batch standardization layer and a modified linear unit activation function;
the convolution layer is as follows:
in the formula, KlA one-dimensional convolution kernel of the l-th layer;the ith convolved local region of the ith layer;
the batch standardization layer is as follows:
in the formula (I), the compound is shown in the specification,is the average of the inputs of layer l;is the standard deviation of the input for layer l; epsilon, gamma and beta are constant values;
the modified linear cell activation function is:
in the formula, yiIs output data; h isiInputting data; alpha is alphaiTaking 100 as the hyperparameter;
performing regression of the global feature vector of the electro-hydraulic servo valve and fault classification of the electro-hydraulic servo valve according to the feature distillation model:
in the formula, LossFIs the regression loss to the global feature vector; lossPClassifying the loss for the fault; α is the distillation loss weight; beta is the predicted loss weight;
global featureRegression loss of eigenvectorsThe method minimizes the difference between the global feature vector output by the backbone network in the student network and the global feature vector output by the backbone network in the teacher network, and the fault classification lossThe difference between the class prediction of the student network on fault data and a real label is minimized;
the student network fits the global feature vector of the electro-hydraulic servo valve, and the minimum mean square error is adopted as a loss function:
in the classification of electrohydraulic servo valve faults, the probability distribution of the faults needs to be predicted, and cross entropy is adopted as a loss function:
3. the method for diagnosing the fault of the electro-hydraulic servo valve according to claim 2, wherein the adjusting the pressure of the electro-hydraulic servo valve fault diagnosis system to a rated pressure drop of the electro-hydraulic servo valve comprises:
normalizing the pressure data according to the following formula:
wherein, Δ pnIs the rated pressure drop of the electro-hydraulic servo valve; p is the output pressure of the electro-hydraulic servo valve; p' is the pressure normalization of the electro-hydraulic servo valve;
the flow data is normalized according to the following formula:
wherein q isnIs the load flow under the rated pressure drop of the electro-hydraulic servo valve; q is the output flow of the electro-hydraulic servo valve; q' is the flow standardization of the electro-hydraulic servo valve;
normalizing the electro-hydraulic servo valve command signal according to the following formula:
wherein s isnThe signal amplitude value is the maximum input electric signal of the electro-hydraulic servo valve and is a rated test signal; s is the displacement feedback of the valve core of the electro-hydraulic servo valve; s' normalization of command signals of the electro-hydraulic servo valve;
normalizing electro-hydraulic servo valve displacement feedback according to the following formula:
wherein, dmaxThe maximum positive displacement feedback of the output maximum value electric signal of the electro-hydraulic servo valve displacement feedback sensor is realized; d' is the standardization of the displacement feedback of the electro-hydraulic servo valve; d is electro-hydraulic servo valve displacement feedback; dminAnd feeding back the maximum negative displacement of the output minimum electric signal of the electro-hydraulic servo valve displacement feedback sensor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210223251.0A CN114607667A (en) | 2022-03-07 | 2022-03-07 | Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210223251.0A CN114607667A (en) | 2022-03-07 | 2022-03-07 | Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114607667A true CN114607667A (en) | 2022-06-10 |
Family
ID=81861635
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210223251.0A Withdrawn CN114607667A (en) | 2022-03-07 | 2022-03-07 | Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114607667A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117688504A (en) * | 2024-02-04 | 2024-03-12 | 西华大学 | Internet of things abnormality detection method and device based on graph structure learning |
-
2022
- 2022-03-07 CN CN202210223251.0A patent/CN114607667A/en not_active Withdrawn
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117688504A (en) * | 2024-02-04 | 2024-03-12 | 西华大学 | Internet of things abnormality detection method and device based on graph structure learning |
CN117688504B (en) * | 2024-02-04 | 2024-04-16 | 西华大学 | Internet of things abnormality detection method and device based on graph structure learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111898095B (en) | Deep migration learning intelligent fault diagnosis method, device, storage medium and equipment | |
Shao et al. | A deep learning approach for fault diagnosis of induction motors in manufacturing | |
Jain et al. | A review of online learning in supervised neural networks | |
CN109918752A (en) | Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks | |
CN111830408A (en) | Motor fault diagnosis system and method based on edge calculation and deep learning | |
Fu et al. | A hybrid CNN-LSTM model based actuator fault diagnosis for six-rotor UAVs | |
Jin et al. | New domain adaptation method in shallow and deep layers of the CNN for bearing fault diagnosis under different working conditions | |
Tamilselvan et al. | Deep belief network based state classification for structural health diagnosis | |
Zhang et al. | Fault detection and recognition of multivariate process based on feature learning of one-dimensional convolutional neural network and stacked denoised autoencoder | |
CN113486578A (en) | Method for predicting residual life of equipment in industrial process | |
CN114607667A (en) | Electro-hydraulic servo valve fault diagnosis system and method based on characteristic distillation | |
Li et al. | Remaining useful life prognostics of bearings based on a novel spatial graph-temporal convolution network | |
Yang et al. | An end-to-end denoising autoencoder-based deep neural network approach for fault diagnosis of analog circuit | |
An et al. | Rolling bearing fault diagnosis algorithm using overlapping group sparse-deep complex convolutional neural network | |
Zhu et al. | A novel feature-fusion-based end-to-end approach for remaining useful life prediction | |
Tao et al. | Real-Time Fault Diagnosis for Hydraulic System Based on Multi-Sensor Convolutional Neural Network | |
Chen et al. | Aircraft Landing Gear Retraction/Extension System Fault Diagnosis with 1-D Dilated Convolutional Neural Network | |
Tamilselvan et al. | Multi-sensor health diagnosis using deep belief network based state classification | |
CN111985158A (en) | Electro-hydraulic servo valve fault diagnosis method and system based on transfer learning | |
Xu et al. | A composite quantile regression long short-term memory network with group lasso for wind turbine anomaly detection | |
CN111967507A (en) | Discrete cosine transform and U-Net based time sequence anomaly detection method | |
Stastny et al. | Analysis of algorithms for radial basis function neural network | |
CN116610940A (en) | Equipment fault diagnosis system based on wavelet transformation and deep neural network | |
CN113469263B (en) | Prediction model training method and device suitable for small samples and related equipment | |
CN114607679A (en) | Electro-hydraulic servo valve fault diagnosis system and method based on message propagation mechanism |
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 | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220610 |
|
WW01 | Invention patent application withdrawn after publication |