CN114135580B

CN114135580B - Position evaluation method and device for magnetic bearing rotor

Info

Publication number: CN114135580B
Application number: CN202111301200.7A
Authority: CN
Inventors: 霍玥潼; 李雪; 赵子静
Original assignee: Gree Electric Appliances Inc of Zhuhai
Current assignee: Gree Electric Appliances Inc of Zhuhai
Priority date: 2021-11-04
Filing date: 2021-11-04
Publication date: 2022-07-26
Anticipated expiration: 2041-11-04
Also published as: CN114135580A

Abstract

The invention discloses a position evaluation method and a position evaluation device of a magnetic bearing rotor, wherein the method comprises the following steps: constructing a position evaluation network model; acquiring the current position of a magnetic bearing rotor; evaluating whether the current position of the magnetic bearing rotor is in a set area by using a position evaluation network model; if the current position of the magnetic bearing rotor is in the standard normal working state area, determining that the magnetic bearing rotor normally operates according to a set standard operation mode; if the current position of the magnetic bearing rotor is not in the standard normal working state area, determining that the magnetic bearing rotor does not normally operate according to a set standard operation mode, and estimating whether the magnetic bearing rotor has a fault according to the current position of the magnetic bearing rotor. According to the scheme, the position of the magnetic suspension bearing rotor is automatically monitored, and the safety of a magnetic suspension bearing system can be improved.

Description

Position evaluation method and device for magnetic bearing rotor

Technical Field

The invention belongs to the technical field of motors, particularly relates to a method and a device for estimating the position of a magnetic bearing rotor, and particularly relates to a method and a device for realizing the real-time estimation and monitoring of the position of the magnetic bearing rotor, the rotor fault and the network estimation.

Background

Compared with the traditional bearing (such as a contact bearing), the electromagnetic bearing has the excellent characteristics of no contact, no abrasion, no need of lubrication and the like, and is widely applied to the fields of vacuum systems, high-speed turbo machinery, air-conditioning compressors and the like. In a typical magnetic suspension magnetic bearing system, a controller acquires position information of a rotor in real time through a displacement sensor, calculates electromagnetic force required for controlling the rotor, and drives an electromagnet through a power amplifier to enable the rotor to be suspended in a controlled manner. The rotor system is nonlinear in nature and has uncertainty, and when a displacement sensor of the magnetic bearing system fails, the magnetic suspension motor system cannot operate, and even the compressor is seriously damaged.

The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.

Disclosure of Invention

The invention aims to provide a position evaluation method and a position evaluation device for a magnetic bearing rotor, which are used for solving the problems that in a magnetic suspension magnetic bearing system, a controller acquires position information of the magnetic suspension bearing rotor through a displacement sensor of the magnetic bearing system, but when the displacement sensor of the magnetic bearing system fails, the magnetic suspension motor system cannot operate, even a compressor is damaged, and the magnetic suspension bearing system is poor in safety, so that the position of the magnetic suspension bearing rotor is automatically monitored, the phenomenon that the magnetic suspension motor system cannot operate or even the compressor is damaged due to inaccurate position information of the magnetic suspension bearing rotor can be avoided, and the safety of the magnetic suspension bearing system is improved.

The invention provides a position evaluation method of a magnetic bearing rotor, which comprises the following steps: constructing a position evaluation network model; the position estimation network model is a network model for estimating whether the position of the magnetic bearing rotor is in a set region; obtaining a current position of the magnetic bearing rotor; evaluating whether the current position of the magnetic bearing rotor is in a set area by using the position evaluation network model to obtain an evaluation result; the set region is a standard normal working state region of the magnetic bearing rotor; in the evaluation result, if the current position of the magnetic bearing rotor is in the standard normal working state area, determining that the magnetic bearing rotor normally operates according to a set standard operation mode; and in the evaluation result, if the current position of the magnetic bearing rotor is not in the standard normal working state area, determining that the magnetic bearing rotor does not normally operate according to a set standard operation mode, and estimating whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor.

In some embodiments, constructing a location assessment network model comprises: building a network model by utilizing a full-connection network and a forward propagation algorithm and adding a Sigmoid function; and collecting a position sample of the magnetic bearing rotor, and training the network model to obtain a position evaluation network model of the magnetic bearing rotor.

In some embodiments, in the case where the depth of the network model is two layers, and three evaluation elements are given in the second layer and one evaluation element is given in the third layer, the model of the position evaluation network model of the magnetic bearing rotor is defined as follows:

wherein, A ₁ For the evaluation results, x, y are two eigenvectors of the position coordination of the magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ An offset for the evaluation element; w represents the parameters in the evaluation element, the superscript of W indicates the number of network layers in the area where the evaluation rotor is in the normal operating state, and the subscript of W indicates the number of the connected evaluation element.

In some embodiments, evaluating whether the current position of the magnetic bearing rotor is in a set region using the position evaluation network model comprises: extracting features based on the current position of the magnetic bearing rotor to obtain the motion track of the magnetic bearing rotor; judging whether each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state or not so as to: under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered to be in the standard normal working state area; and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is not in the standard normal working state area.

In some embodiments, estimating whether the magnetic bearing rotor is malfunctioning based on the current position of the magnetic bearing rotor comprises: determining whether the current position of the magnetic bearing rotor is within a set allowable range if the current position of the magnetic bearing rotor is not within the standard normal operating state region; predicting a next position of the magnetic bearing rotor if the current position of the magnetic bearing rotor is within the allowable range under a condition that the current position of the magnetic bearing rotor is not in the standard normal operating state region; and according to the next position of the magnetic bearing rotor, the position evaluation network model is reused to evaluate whether the current position of the magnetic bearing rotor is in a set area so as to evaluate whether the magnetic bearing rotor fails; the next position of the magnetic bearing rotor is the position of the magnetic bearing rotor at the next set moment; predicting the magnetic bearing rotor fault if the current position of the magnetic bearing rotor is not within the allowable range if the current position of the magnetic bearing rotor is not within the standard normal operating condition zone.

In accordance with the above method, another aspect of the present invention provides a position estimating apparatus for a magnetic bearing rotor, comprising: a modeling unit configured to construct a location evaluation network model; the position estimation network model is a network model for estimating whether the position of the magnetic bearing rotor is in a set region; an acquisition unit configured to acquire a current position of the magnetic bearing rotor; an evaluation unit configured to evaluate whether a current position of the magnetic bearing rotor is in a set region using the position evaluation network model, resulting in an evaluation result; the set region is a standard normal working state region of the magnetic bearing rotor; the evaluation unit is further configured to determine that the magnetic bearing rotor normally operates according to a set standard operation mode if the current position of the magnetic bearing rotor is in the standard normal operating state region in the evaluation result; the evaluation unit is further configured to determine that the magnetic bearing rotor does not normally operate according to a set standard operation mode if the current position of the magnetic bearing rotor is not in the standard normal operating state region in the evaluation result, and estimate whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor.

In some embodiments, the modeling unit, constructing a location estimation network model, comprises: building a network model by utilizing a fully connected network and a forward propagation algorithm and adding a Sigmoid function; and collecting a position sample of the magnetic bearing rotor, and training the network model to obtain a position evaluation network model of the magnetic bearing rotor.

wherein, A ₁ For the evaluation, x, y are two eigenvectors of the position-coordinated magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ To evaluate the offset of the element; w represents the parameters in the evaluation element, the superscript of W indicates the number of network layers in the area where the evaluation rotor is in the normal operating state, and the subscript of W indicates the number of the connected evaluation element.

In some embodiments, the evaluating unit, using the position evaluation network model, to evaluate whether the current position of the magnetic bearing rotor is in a set region, includes: performing characteristic extraction based on the current position of the magnetic bearing rotor to obtain the motion track of the magnetic bearing rotor; judging whether each axis position track point in the motion track of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state or not so as to: under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered to be in the standard normal working state area; and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered not to be in the standard normal working state area.

In some embodiments, the estimating unit estimating whether the magnetic bearing rotor is faulty or not according to the current position of the magnetic bearing rotor, includes: determining whether the current position of the magnetic bearing rotor is within a set allowable range in case that the current position of the magnetic bearing rotor is not in the standard normal operation state region; predicting a next position of the magnetic bearing rotor if the current position of the magnetic bearing rotor is within the allowable range, if the current position of the magnetic bearing rotor is not within the standard normal operating state region; and according to the next position of the magnetic bearing rotor, the position evaluation network model is utilized again to evaluate whether the current position of the magnetic bearing rotor is in a set area so as to evaluate whether the magnetic bearing rotor fails; the next position of the magnetic bearing rotor is the position of the magnetic bearing rotor at the next set moment; predicting the magnetic bearing rotor fault if the current position of the magnetic bearing rotor is not within the allowable range under the condition that the current position of the magnetic bearing rotor is not in the standard normal operating state region.

In accordance with another aspect of the present invention, there is provided a magnetic levitation system, comprising: the position estimating apparatus of a magnetic bearing rotor described above.

In line with the above method, a further aspect of the present invention provides a storage medium comprising a stored program, wherein the program when executed controls a device in which the storage medium is located to perform the above-described method of estimating the position of a magnetic bearing rotor.

In line with the above method, a further aspect of the invention provides a processor for running a program, wherein the program is run to perform the above-described method of position estimation of a magnetic bearing rotor.

Therefore, according to the scheme of the invention, the position evaluation network model is constructed, whether the real-time position of the magnetic suspension bearing rotor is in a standard normal working state area or not is evaluated by using the position evaluation network model, the position of the rotor in a non-standard operation working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of fault occurrence is further evaluated, and the automatic monitoring of the position of the magnetic suspension bearing rotor is realized.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.

Drawings

FIG. 1 is a schematic flow chart diagram of one embodiment of a method of position estimation of a magnetic bearing rotor of the present invention;

FIG. 2 is a schematic flow chart diagram illustrating one embodiment of constructing a location estimation network model in the method of the present invention;

FIG. 3 is a schematic flow chart illustrating one embodiment of using the position estimation network model to estimate whether the current position of the magnetic bearing rotor is within a set zone in the method of the present invention;

FIG. 4 is a schematic flow chart illustrating one embodiment of the method of the present invention for estimating whether the magnetic bearing rotor is malfunctioning based on the current position of the magnetic bearing rotor;

FIG. 5 is a schematic structural view of an embodiment of the position estimation device for a magnetic bearing rotor of the present invention;

FIG. 6 is a schematic diagram of a rotor axial position trajectory;

FIG. 7 is a schematic diagram of the model after 88 rounds of rotor position visualization training;

FIG. 8 is a graphical illustration of a Sigmoid function;

FIG. 9 is a schematic diagram illustrating an execution flow of an evaluation meta structure added to a sigmoid function;

FIG. 10 is a schematic diagram of a network model for enhanced rotor position estimation;

fig. 11 is a flowchart illustrating a process of determining the rotor position.

The reference numbers in the embodiments of the present invention are as follows, in combination with the accompanying drawings:

102-a modeling unit; 104-an obtaining unit; 106-evaluation unit.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the specific embodiments of the present invention and the accompanying drawings. It is to be understood that the disclosed embodiments are merely exemplary of the invention, and are not intended to be exhaustive or exhaustive. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.

The monitoring system of the working process of the magnetic bearing is perfected through the key information hidden in the rotor position data, the fault diagnosis and the predictive maintenance of the rotor are broken through, and the method is an urgent requirement for realizing the normal operation of a green and safe magnetic suspension bearing system.

In some schemes, the soft measurement method of the magnetic linkage of the bearingless permanent magnet synchronous motor based on the weighted least square support vector machine realizes the online real-time prediction control of the magnetic linkage value of the bearingless permanent magnet synchronous motor. The algorithm is complex, the parameter configuration lacks certain standards, and the calculation amount is large. Once a problem occurs in the implementation process, the system is directly damaged.

In other schemes, the rotor displacement is processed in real time by utilizing composite Fourier transform and a self-adaptive generalized second-order integrator, and then fundamental wave and harmonic components are obtained. The rotor state is monitored on line through real-time calculation, and when a fault occurs, the state of the molecular pump during deceleration is utilized for fault diagnosis, so that mechanical faults, global demagnetization faults and local demagnetization faults are distinguished. And only the fault when the molecular pump rotor is decelerated can be alarmed, and three fault conditions are judged. The position of the rotor cannot be monitored in real time and early-stage fault estimation is carried out, the method is large and complex in calculation amount, and system faults cannot be visually monitored.

According to an embodiment of the present invention, a method of position estimation of a magnetic bearing rotor is provided, as shown in the flow diagram of an embodiment of the method of the present invention in fig. 1. The position estimation method of a magnetic bearing rotor may include: step S110 to step S150.

At step S110, a location estimation network model is constructed. The position estimation network model is a network model for estimating whether or not the position of the magnetic bearing rotor is in a set region.

In some embodiments, a specific process of constructing the location estimation network model in step S110 is further described with reference to a flowchart of an embodiment of constructing the location estimation network model in the method of the present invention shown in fig. 2, where the specific process includes: step S210 and step S220.

And step S210, building a network model by utilizing the full-connection network and the forward propagation algorithm and adding a Sigmoid function.

Step S220, collecting position samples of the magnetic bearing rotor, and training the network model to obtain a position evaluation network model of the magnetic bearing rotor.

Specifically, the scheme of the invention provides a network for forward propagation reinforced evaluation of whether the rotor position is in a standard normal working state area, and the network is used for evaluating whether the real-time position of the magnetic suspension bearing rotor is in the standard normal working state area, further judging whether the rotor position in a non-standard operation working state is in a normal working area or a fault area, predicting the position of the next state of the rotor in the normal working area, further evaluating the possibility of fault occurrence, realizing automatic monitoring of the position of the magnetic suspension bearing rotor, and evaluating and finding the fault of the magnetic suspension bearing rotor system in real time.

In the scheme of the invention, a mode of adding a Sigmoid function into a forward propagation reinforced position evaluation network model is adopted, namely a monitoring algorithm is a forward propagation method, so that the problems of high nonlinearity and uncertainty degree of a magnetic suspension bearing system are solved, and the automatic detection of the rotor fault of the magnetic suspension bearing is realized. Wherein, the problem that the position point of the axis of the rotor is inseparable can be solved by using the activation function.

In the scheme of the invention, the full-connection network is used for strengthening and evaluating the faults of the magnetic suspension bearing rotor, the judgment of the working state type of the rotor is realized through the identification result, and the faults of the magnetic suspension bearing rotor system can be found as early as possible. The fully-connected network is essentially a single switch that connects all inputs and outputs, and has the characteristics of high throughput, high reliability and low delay.

wherein, A ₁ For the evaluation results, x, y are two eigenvectors of the position coordination of the magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ To evaluate the offset of the element. W represents the parameters in the evaluation element, the superscript of W indicates the number of network layers in the area where the evaluation rotor is in the normal operating state, and the subscript of W indicates the number of the connected evaluation element.

In the scheme of the invention, only a simple three-layer full-connection network using a Sigmoid function is taken as an example to strengthen and evaluate the position of the rotor of the magnetic suspension bearing, and the method is also suitable for strengthening and evaluating the position of the rotor of the magnetic suspension bearing by other deep-layer connection network structures under other functions capable of solving the indistinguishable problem.

In the model shown in fig. 7, only two layers of depths are given, the second layer is given 3 evaluation elements, and the third layer is given 1 evaluation element, so as to make a brief simulation building description. In actual processing, the depth and the breadth can be changed, and the more complex the model is built, the denser the network is, and the better the obtained effect is.

In the solution of the invention, the two inputs can be obtained by extracting the eigenvector in the axial position of the magnetic bearing rotor in the actual displacement sensor.

FIG. 6 is a schematic diagram of a rotor axial position trajectory. Fig. 6 is a side view of a magnetic suspension bearing system rotor, in which the axis locus is extracted, each point represents a position sample of the rotor, blue represents a position of the rotor in a standard normal operation state, and orange represents a position of the rotor in a non-standard operation state. As shown in fig. 6: each point represents a sample position of the rotor, blue for the rotor position in standard normal operation and orange for the rotor position in non-standard operation.

Fig. 7 is a schematic diagram of the model after 88 rounds of rotor position visualization training. Fig. 7 is a specific example of a Tensflow (i.e. symbolic mathematical system based on dataflow programming) visualization interface, on-line fairground simulation (which can be understood as software with a simulation library), according to the solution of the invention, it is necessary to construct a model as shown in fig. 7.

In the example shown in fig. 7, x1 and x2 in FEATURES indicate the input of two feature values. 2HIDDEN LAYERS shows that the model has two hidden layers, i.e. depth of 2. neurons are neurons, 3neurons indicates 3neurons in the second layer, and 1neurons indicates 1neuron in the third layer. OUTPUT represents the OUTPUT value, Test loss finger loss value is 0.006, Training loss finger loss value is 0.009. Lower right corner colors windows data, neuron and weight vlaues: the colors in fig. 7 show the data values, neurons, and weight weights.

One is x and the other is y, which correspond to the two inputs shown in fig. 7, the axes positions of which are coordinated in the side view of the magnetic bearing.

The following takes a three-layer fully-connected network using an activation function as an example to illustrate a specific implementation process of the solution of the present invention.

In the scheme of the invention, whether the real-time position of the magnetic bearing rotor is in a standard normal working state area or not is intensively evaluated, the position of the rotor at the next moment is predicted when the rotor is in a non-standard running working state, and the possibility of fault occurrence is further estimated. The magnetic bearing rotor can evaluate and judge the position state of the magnetic bearing rotor through real-time information; in addition, the position state of the current moment can be predicted through the previous moment in a prediction mode. Two ways enhance the determination of the rotor position, called enhanced evaluation. The output of the result of the enhanced estimation of the real-time position area of the rotor is the weighted sum of two input x and y of position coordination, and the different input weights are the parameters of the area where the enhanced estimation of the rotor position is located. As shown in fig. 6, after the origin is set in the figure, the position (x, y) coordinates of the rotor at each time are easily found, corresponding to x1, x2 in fig. 7. The network optimization process for the strengthened evaluation of the rotor real-time position area is the process of parameter value taking in the result of the area where the rotor position is optimized.

FIG. 10 is a schematic diagram of a network model for enhanced rotor position estimation. The structure of the evaluation element after adding the Sigmoid function is shown in fig. 9, and the bias term is expressed as an evaluation element whose output is always 1, so as to obtain a network model for forward propagation enhanced evaluation of whether the rotor position is in a standard normal working state area or not, which is shown in fig. 10. A model of a network model for forward propagation enhanced assessment of whether the rotor position is in a normal operating state region is defined as:

in the network model for forward propagation reinforced evaluation of whether the rotor position is in the standard normal working state area, the derivation formula of the hidden layer is as follows:

in the network model for evaluating whether the rotor position is in the standard normal working state area or not by forward propagation reinforcement, the derivation formula of the output layer is as follows:

wherein, a ₁₁ 、a ₁₂ 、a ₁₃ And F is an evaluation element. With a ₁₁ The evaluator is an example, with two inputs: are the outputs of x and y, respectively, and a ₁₁ The output of (1) is then the input of the evaluation element F.

W denotes a parameter in the evaluation bin. The superscript of W indicates: and evaluating the network layer number of the rotor in the standard normal working state area. I.e. W ⁽¹⁾ Parameter, W, representing first-level evaluation element ⁽²⁾ Parameters representing the second level evaluator. The subscript of W indicates: numbering of connection evaluation elements, i.e.

Denotes the connection x and a ₁₂ Weights are evaluated on edges of the element.

In the solution of the present invention, the two classification problems involved, that is, whether the rotor position is in the standard normal operating region condition or not, are determined, and the output of the given evaluation rotor network only includes one evaluation element, and this evaluation element outputs a real value, which is F in fig. 10.

In summary, according to the scheme of the present invention, by giving the input of the estimated rotor network, estimating the network structure of the rotor, the weight on the side, and solving the indistinguishable problem function Sigmoid, it can be calculated by the forward propagation algorithm whether the estimated rotor position is in the output of the standard normal working area network.

At step S120, a current position of the magnetic bearing rotor is acquired.

At step S130, it is evaluated whether the current position of the magnetic bearing rotor is in a set region using the position evaluation network model, resulting in an evaluation result. The set region is a standard normal operating state region of the magnetic bearing rotor. The evaluation result comprises: the current position of the magnetic bearing rotor is in the standard normal operating state region, or the current position of the magnetic bearing rotor is not in the standard normal operating state region.

In some embodiments, in combination with the flowchart of an embodiment of using the position estimation network model to estimate whether the current position of the magnetic bearing rotor is in the set region in the method of the present invention shown in fig. 3, a specific process of using the position estimation network model to estimate whether the current position of the magnetic bearing rotor is in the set region in step S130 is further described, which includes: step S310 and step S320.

Step S310, based on the current position of the magnetic bearing rotor, feature extraction is carried out to obtain the motion track of the magnetic bearing rotor.

Step S320, determining whether each axis position track point in the motion trajectory of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, so as to: and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, considering that the current position of the magnetic bearing rotor is in the standard normal working state area. And under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is not in the standard normal working state area.

Four parts of information are needed for calculating a network for forward propagation reinforced evaluation of whether the rotor position is in a standard normal working state area, and the method comprises the following specific steps:

part 1, the first part information is two inputs x, y, which are position coordinated.

And (3) correspondingly classifying an actual problem (such as judging whether each axis position track point in the side view of the rotor is a position point of the rotor in a standard normal working state) into two categories of different colors on a plane. In the scheme of the invention, the position of the axial center of the side view of the rotor in the practical problem is changed into one point on the coordinate plane, namely the characteristic extraction. Entities in practical problems can be converted into points in space through feature extraction, and the feature extraction methods are numerous, for example: a method of feature extraction, comprising: and sampling and storing the position signal into a database by using a data card through the position signal of the axis of the magnetic bearing rotor obtained from the actual displacement sensor, and performing wavelet transformation denoising treatment to obtain the motion track of the magnetic bearing rotor in the side view.

And part 2 and the second part of information are network connection structures for evaluating whether the rotor is in a standard normal working area, and a network for intensively evaluating whether the rotor is in a standard normal working state area, which can solve the indivisible problem, is composed of a plurality of evaluation elements. The connection structure gives the input and output connection relationship between different evaluation elements, as shown in fig. 10: a is a ₁₁ The evaluator has two inputs, which are the outputs of x and y, respectively, and a ₁₁ The output of (1) is then the input of the evaluation element F.

Fig. 7 and fig. 10 correspond to each other, i.e. fig. 7 is a visual representation of fig. 10: the small blocks in fig. 7 in the solution according to the invention correspond to the evaluation elements in fig. 10, i.e.: x1, x2 of the first layer in fig. 7 correspond to x, y in fig. 10; the two layers correspond to a11, a12 and a 13; the third level corresponds to F. The color depth of the connecting lines in fig. 7 corresponds to the weight W value in fig. 10. The darker the color of the evaluation element connecting line is, the larger the absolute value of the weight value is, and when the color is changed into white, the weight value is 0; the color of the evaluation element represents an output value at the position coordinate, and the color is darker as the output value is larger.

Part 3, the third part of information is the parameters (such as weight) in each evaluation element.

In FIG. 10, the parameters in the evaluation element are denoted by W, and the superscript of W indicates the number of network layers in the area where the rotor is evaluated in the normal operating state, i.e., W ⁽¹⁾ Parameter, W, representing first-level evaluation element ⁽²⁾ Parameters representing the second level evaluator. Subscript table of WKeep track of the number of connection evaluation elements, i.e.

Denotes the connection x and a ₁₂ Weights on the edges of the element are evaluated.

In FIG. 7, the small blocks are network evaluation units for evaluating whether the rotor is in the normal working area, such as evaluation unit a ₁₁ 、a ₁₂ 、a ₁₃ . The virtual connecting line is the connection between the evaluation elements, the darker the color of the connecting line of the evaluation elements is, the larger the absolute value of the weight value is, and when the color is changed to white, the weight value is 0. The color of the evaluation element represents an output value at the position coordinate, and the color is darker as the output value is larger.

As shown in FIG. 7, x ₁ The evaluation element is x in position coordinate, and the distinguishing plane is y axis. Because the output of the evaluation element is x ₁ Value of itself, so when x ₁ Below 0, the evaluator output is negative, so the left side is orange. When x is ₁ Above 0, the evaluator output is positive, so the right side is blue. x is a radical of a fluorine atom ₂ The evaluation element is y in the position coordination.

And part 4 and the fourth part of information are added Sigmoid functions.

Fig. 8 is a graph illustrating Sigmoid function. And adding a Sigmoid function to solve the problem that the position point of the axis of the rotor is not distinguishable. Considering that the problem solved by the linear model is limited, the problem of judging whether the rotor position is in a standard normal working area in practice cannot be divided by a straight line or a plane of a high-dimensional space, that is, cannot be clearly divided linearly. Therefore, the scheme of the invention adds a Sigmoid function (as shown in fig. 8) to non-linearize the model in order to evaluate whether the rotor position is in the normal working area. The Sigmoid function added is as follows:

fig. 9 is a flowchart illustrating an execution of an evaluation meta structure added to the sigmoid function. Evaluation of the incorporation of sigmoid function shown in FIG. 9The meta structure is an evaluation meta general structure added to the sigmoid function, and is only schematically illustrated here. In FIG. 9, b is the offset, x ₀ 、x ₁ 、x ₂ For input, w ₀ 、w ₁ 、w ₂ A weight for each evaluation element.

In step S140, in the evaluation result, if the current position of the magnetic bearing rotor is in the standard normal operating state region, it is determined that the magnetic bearing rotor normally operates according to a set standard operating mode, and the magnetic bearing rotor is controlled to continue to operate according to the set standard operating mode.

In step S150, in the evaluation result, if the current position of the magnetic bearing rotor is not in the standard normal operating state region, it is determined that the magnetic bearing rotor does not normally operate according to a set standard operating mode, and it is estimated whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor.

The scheme of the invention provides a network for strengthening the evaluation of the position of a magnetic bearing rotor, the real-time rotor fault and the pre-estimation monitoring, and the invention further judges the position of the rotor in a non-standard operation working state to be in a normal working area or a fault area by evaluating whether the real-time position of the magnetic bearing rotor is in a standard normal working state area or not, and predicts the position of the rotor in the normal working area in the next state, thereby evaluating the possibility of fault occurrence, realizing the automatic monitoring of the position of the magnetic bearing rotor, and being capable of visually carrying out early diagnosis and predictive maintenance on the fault of the magnetic bearing rotor. Therefore, the problems of possible overshoot and compressor damage when the displacement sensor of the magnetic bearing system fails are solved. The problems that the magnetic suspension bearing rotor fault diagnosis and identification accuracy is not high and cannot be estimated are solved. And the problem that the displacement sensor cannot accurately evaluate the fault type of the rotor is solved.

In some embodiments, referring to the flow chart of fig. 4 illustrating an embodiment of estimating whether the magnetic bearing rotor is faulty according to the current position of the magnetic bearing rotor, the specific process of estimating whether the magnetic bearing rotor is faulty according to the current position of the magnetic bearing rotor in step S150 is further described, which comprises: step S410 and step S420.

Step S410, determining whether the current position of the magnetic bearing rotor is within a set allowable range, if the current position of the magnetic bearing rotor is not within the standard normal operating state region.

Step S420, under the condition that the current position of the magnetic bearing rotor is not in the standard normal working state area, if the current position of the magnetic bearing rotor is within the allowable range, predicting the next position of the magnetic bearing rotor; and evaluating whether the current position of the magnetic bearing rotor is in a set region according to the next position of the magnetic bearing rotor by using the position evaluation network model again to evaluate whether the magnetic bearing rotor fails. The next position of the magnetic bearing rotor is the position of the magnetic bearing rotor at the next set time. Wherein the current position of the magnetic bearing rotor within the permitted range is determined by a number of magnetic bearing systems, which is a certain magnetic bearing system when evaluated.

Step S430, under the condition that the current position of the magnetic bearing rotor is not in the standard normal operating state region, if the current position of the magnetic bearing rotor is not within the allowable range, predicting the magnetic bearing rotor fault.

The setting area is a standard normal working state area and is used for judging whether the rotor is in a standard working state or not. And an allowable range for determining whether the rotor is in a normal (i.e., non-standard) operating state or a fault state when the rotor is not in the standard operating state. And in a set area, namely a standard normal working state area, determining that the magnetic bearing rotor is in the standard working area, and dividing the magnetic bearing rotor into a normal working area and a fault area when the magnetic bearing rotor is in the non-standard area. The possibility of a failure in the next state of the rotor is predicted even further for the case of normal (non-standard) operating conditions, since the operating state is not the best state even though the rotor is not at all, and therefore the possibility of a failure in the rotor under evaluation is predicted.

Fig. 11 is a flowchart illustrating a process of determining the rotor position. As shown in fig. 11, in the solution of the present invention, the process of determining the rotor position by using the network of intensively estimating the rotor position capable of solving the indivisible problem includes:

step 1, obtaining rotor displacement through a magnetic suspension bearing rotor displacement sensor.

In a related solution, 10 displacement sensors are required: 2 sensor probes are needed in the axial direction, and 8 sensor probes are needed in the radial direction. The method mentioned in the solution of the present invention is based on the setting of the origin as shown in fig. 6, and the position (x, y) coordinates of the rotor at each moment can be easily obtained, which corresponds to x1, x2 in fig. 7. When a certain sensor fails, because other sensors can acquire a certain information quantity value of x1 or x2, the other position information quantity can be determined through the position information quantity acquired at the previous moment and the enhanced evaluation, and the functions of the failed sensor can be compensated.

And 2, obtaining a motion track of the side view of the rotor by processing the displacement.

And 3, extracting the characteristic vectors of the entities in the rotor motion trail diagram as the input of the reinforced evaluation rotor position network, wherein different entities can extract different characteristic vectors.

And 4, adjusting the value of the weight (as a parameter) in the reinforced evaluation rotor position network through the training data.

And 5, judging whether the rotor position is in the standard working area or the non-standard area through the area where the rotor position is in the visual trained model.

And 6, further judging whether the rotor position in the non-standard area is in a normal working area or a fault area.

In the scheme of the invention, the specific description is given on how to judge whether the rotor is in a standard normal working area or a non-standard working area (the non-standard working area comprises a normal working area and a fault area). Accordingly, as shown in fig. 10, the determination process is: giving a threshold value 0, wherein the more reliable the real value F is away from the threshold value, and if the output value F is greater than the threshold value 0, the position is the position in the standard normal working state; otherwise, the position is a non-standard working state position. Accordingly, as shown in fig. 7, the determination process is: as shown on the right side of fig. 7: the position points of the axis of the rotor are obviously classified, and whether the position of the rotor is in a standard normal working state area or not can be judged through the area map: the upper right blue is a standard normal working area of the rotor, and the lower left orange is a non-standard operation working area of the rotor.

The method for distinguishing the normal (non-standard) working state from the fault state when the rotor is in the non-standard working region is the same as the following steps: and (4) setting position information and establishing a model. As shown in fig. 7, the position information at this time is input, and after the discrimination:

the orange zone represents a fault condition; the blue region represents normal (non-standard) operation.

And 7, predicting the position of the next state of the rotor in the normal working area, and further evaluating the possibility of failure.

As shown in fig. 7, after obtaining certain rotor position information (not only a certain magnetic bearing rotor system, but also a model is established according to a plurality of magnetic bearing systems), the position of the rotor of a certain magnetic bearing rotor system (which is single at this time, and refers to a certain system in particular) on the right side of fig. 7 is determined, and the operating condition of the system is predicted (in a short period, all in a single area).

As shown in fig. 7, the orange zone represents a fault condition; the blue region represents normal (non-standard) operation; when the blue region is close to the boundary line, the rotor is in a normal working state, but the probability of failure in the future is high (the closer the distance is, the higher the probability of failure is). Further, the magnitude of the probability of failure is specifically measured by the distance of the rotor position at that time from the boundary line. The foregoing process is determined by a plurality of magnetic bearing rotor systems. At this time, the possibility of a failure of a magnetic bearing rotor system is determined, and the distance from the operating region of the system (the center of gravity is found by the region and is converted into one point) to the boundary line is determined.

Accordingly, the determination process of fig. 10 is: given a threshold value of 0, the more reliable the real value F is from the threshold value, and if the output value F is greater than the threshold value 0, this position is the position in the normal operating state. Otherwise, the position is a non-standard working state position. And when the rotor is judged to be in the non-standard working state, judging whether the rotor is in the fault state according to the region of the rotor position point in the non-standard working process. If the current position of the rotor is only a non-standard position and is not enough to cause the fault, the possibility of the fault of the rotor can be evaluated according to the predicted position of the rotor at the next moment.

Accordingly, the determination process of fig. 7 is: the evaluation element at Output in FIG. 7 is the Output, showing the network discrimination plane for the evaluation rotor position Output and the discrimination data points for the desired evaluation rotor position network. After 88 rounds (custom) of the visual training evaluation rotor position network, the results are shown on the right side of fig. 7: the position points of the axis of the rotor are obviously classified, and whether the position of the rotor is in a standard normal working state area or not can be judged through the area map: the upper right blue is a standard normal working area of the rotor, and the lower left orange is a non-standard operating working area of the rotor. Through actual conditions testing, the nonstandard operation working area of the rotor can be further divided into: a normal (non-standard but tolerable) rotor operation zone and a rotor fault zone may be allowed. If the rotor is in a normal operation but non-standard region at the moment, the possibility of the rotor being in a fault can be predicted by evaluating the region where the position point of the rotor is located at the next moment.

In the solution of the present invention, fig. 7 is the visualization model of fig. 10, and the two are the same.

By adopting the technical scheme of the embodiment, the position evaluation network model is constructed, whether the real-time position of the magnetic suspension bearing rotor is in a standard normal working state area or not is evaluated by utilizing the position evaluation network model, the position of the rotor in a non-standard operation working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of fault occurrence is further evaluated, and the automatic monitoring of the position of the magnetic suspension bearing rotor is realized.

According to an embodiment of the present invention, there is also provided a position estimation apparatus of a magnetic bearing rotor corresponding to the position estimation method of a magnetic bearing rotor. Referring to fig. 5, a schematic diagram of an embodiment of the apparatus of the present invention is shown. The position estimation apparatus of a magnetic bearing rotor may include: a modeling unit 102, an acquisition unit 104, and an evaluation unit 106.

Wherein the modeling unit 102 is configured to build a location estimation network model. The position estimation network model is a network model for estimating whether the position of the magnetic bearing rotor is in a set region. The specific function and processing of the modeling unit 102 are referred to in step S110.

In some embodiments, the modeling unit 102, constructing a location estimation network model, includes:

the modeling unit 102 is further specifically configured to build a network model by using a fully connected network and a forward propagation algorithm and adding a Sigmoid function. The detailed function and processing of the modeling unit 102 are also referred to in step S210.

The modeling unit 102 is further configured to collect position samples of the magnetic bearing rotor, train the network model, and obtain a position estimation network model of the magnetic bearing rotor. The detailed function and processing of the modeling unit 102 are also referred to in step S220.

Specifically, the scheme of the invention provides a network for forward propagation reinforced evaluation of whether the rotor position is in a standard normal working state area, and the network is used for evaluating whether the real-time position of the magnetic suspension bearing rotor is in the standard normal working state area, further judging whether the rotor position in a non-standard operation working state is in a normal working area or a fault area, predicting the position of the next state of the rotor in the normal working area, and further evaluating the possibility of fault, thereby realizing automatic monitoring of the position of the magnetic suspension bearing rotor, and being capable of evaluating and finding the fault of the magnetic suspension bearing rotor system in real time and as early as possible.

In the scheme of the invention, a mode of adding a Sigmoid function into a forward propagation reinforced position evaluation network model is adopted, namely, a monitoring algorithm is a forward propagation method, so that the problems of nonlinearity and high uncertainty degree of a magnetic suspension bearing system are solved, and the automatic detection of the rotor fault of the magnetic suspension bearing is realized. Wherein, the problem that the position point of the axis of the rotor is inseparable can be solved by using the activation function.

In some embodiments, where the depth of the network model is two layers, and three evaluation elements are given in the second layer, and one evaluation element is given in the third layer, the model of the position evaluation network model of the magnetic bearing rotor is defined as follows:

wherein, A ₁ For the evaluation, x, y are two eigenvectors of the position-coordinated magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ To evaluate the offset of the element. W denotes the parameters in the evaluation element, the superscript of W indicates the number of network layers in the area where the rotor is evaluated in the normal operating state, and the subscript of W indicates the number of the connected evaluation element.

In the scheme of the invention, only a simple three-layer full-connection network using a Sigmoid function is taken as an example to enhance and evaluate the position of the magnetic suspension bearing rotor, and the method is also suitable for enhancing and evaluating the position of the magnetic suspension bearing rotor by other deep-layer connection network structures under the function capable of solving the indistinguishable problem.

In the scheme of the invention, the two inputs can be obtained by extracting the characteristic vector in the axial center position of the magnetic bearing rotor in the actual displacement sensor.

FIG. 6 is a schematic diagram of a rotor axial position trajectory. Fig. 6 is a side view of a magnetic bearing system rotor, showing axial locus extraction points, each point representing a position example of the rotor, blue representing the rotor position in a standard normal operation state, and orange representing the rotor position in a non-standard operation state. As shown in fig. 6: each point represents a sample position of the rotor, blue for the rotor position in standard normal operation and orange for the rotor position in non-standard operation.

In the example shown in fig. 7, x1 and x2 in FEATURES indicate the input of two feature values. 2HIDDEN LAYERS shows that the model has two hidden layers, i.e. depth of 2. neurons are neurons, 3neurons indicates 3neurons in the second layer, and 1neurons indicates 1neuron in the third layer. OUTPUT represents the OUTPUT value, Test loss refers to a loss value of 0.006, and Training loss refers to a Training loss value of 0.009. Bottom right corner colors shows data, neuron and weight vlaues: the colors in fig. 7 show the data values, neurons, and weights.

One is x and the other is y, which are coordinated in axial position in the side view of the magnetic bearing, corresponding to the two inputs shown in fig. 7.

The following is an example of a three-layer fully-connected network using an activation function, and a specific implementation process of the scheme of the present invention is described.

In the scheme of the invention, whether the real-time position of the magnetic bearing rotor is in a standard normal working state area or not is intensively evaluated, the position of the rotor at the next moment is predicted when the rotor is in a non-standard running working state, and the possibility of fault occurrence is further estimated. The output of the result of the enhanced estimation of the real-time position area of the rotor is the weighted sum of two input x and y of position coordination, and the different input weights are the parameters of the area where the enhanced estimation of the rotor position is located. The network optimization process for intensively evaluating the real-time position area of the rotor is the process of optimizing the value of the parameter in the result of the area where the position of the rotor is located.

FIG. 10 is a schematic diagram of a network model for enhanced rotor position estimation. The structure of the evaluation element after adding the Sigmoid function is shown in fig. 9, and the bias term is expressed as an evaluation element with an output of 1 forever, so that a network model for forward propagation enhanced evaluation of whether the rotor position is in a standard normal working state area is obtained and shown in fig. 10. A model of a network model for forward propagation enhanced assessment of whether the rotor position is in a normal operating state region is defined as:

wherein, a ₁₁ 、a ₁₂ 、a ₁₃ And F is an evaluation element. With a ₁₁ The evaluator is an example, with two inputs: are the outputs of x and y, respectively, and a ₁₁ Is the input of the evaluation element F.

W denotes a parameter in the evaluation bin. The superscript of W indicates: and evaluating the network layer number of the rotor in the area with the standard normal working state. I.e. W ⁽¹⁾ Parameter, W, representing first-level evaluation element ⁽²⁾ Parameters representing the second level evaluation elements. The subscript of W indicates: numbering of connection evaluation elements, i.e.

In the solution of the present invention, the two classification problems involved, that is, whether the rotor position is in the standard normal operating region condition or not, are determined, the output of the given rotor network to be evaluated only includes one evaluation element, and this evaluation element outputs a real value, which is F in fig. 10.

In summary, according to the scheme of the present invention, by giving the input of the rotor network to be evaluated, evaluating the network structure of the rotor, the weight on the edge, and the function Sigmoid for solving the indivisible problem, it can be calculated by the forward propagation algorithm to evaluate whether the rotor position is in the output of the standard normal work area network.

An obtaining unit 104 configured to obtain a current position of the magnetic bearing rotor. The specific function and processing of the acquisition unit 104 are referred to in step S120.

An evaluation unit 106 configured to evaluate whether the current position of the magnetic bearing rotor is in a set region using the position evaluation network model, resulting in an evaluation result. The set region is a standard normal operating state region of the magnetic bearing rotor. The evaluation result comprises: the current position of the magnetic bearing rotor is in the standard normal operating state region, or the current position of the magnetic bearing rotor is not in the standard normal operating state region. The detailed function and processing of the evaluation unit 106 are shown in step S130.

In some embodiments, the evaluating unit 106, using the position evaluation network model, to evaluate whether the current position of the magnetic bearing rotor is in a set region, includes:

the evaluation unit 106 is in particular further configured to perform a feature extraction based on the current position of the magnetic bearing rotor, resulting in a motion trajectory of the magnetic bearing rotor. The specific function and processing of the evaluation unit 106 are also referred to in step S310.

The evaluation unit 106 is specifically further configured to determine whether each axis position track point in the motion trajectory of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal operating state, so as to: and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered to be in the standard normal working state area. And under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered not to be in the standard normal working state area. The specific function and processing of the evaluation unit 106 are also referred to in step S320.

part 1, the first part information, is two inputs x, y, which are position coordinated.

And (3) correspondingly classifying an actual problem (such as judging whether each axis position track point in the side view of the rotor is a position point of the rotor in a standard normal working state) into two categories of different colors on a plane. In the scheme of the invention, the position of the side view axis of the rotor in the practical problem is changed into one point on the coordinate plane, namely the characteristic extraction. Entities in an actual problem can be converted into points in space through feature extraction, and the feature extraction devices are numerous, such as: an apparatus for feature extraction, comprising: and sampling and storing the position signal into a database by using a data card through the position signal of the axle center of the magnetic bearing rotor obtained from the actual displacement sensor, and performing wavelet transformation denoising treatment to further obtain the motion track of the magnetic bearing rotor in the side view.

And part 2 and the second part of information are network connection structures for evaluating whether the rotor is in a standard normal working area, and the network for intensively evaluating whether the rotor is in the standard normal working area, which can solve the indivisible problem, is composed of a plurality of evaluation elements. The connection structure gives the input and output connection relation between different evaluation elements, as shown in fig. 10: a is a ₁₁ The evaluator has two inputs, which are the outputs of x and y, respectively, and a ₁₁ Is the input of the evaluation element F.

Part 3, the third part information is the parameters (such as weight) in each evaluation element.

In fig. 10, the parameters in the evaluation unit are denoted by W, and the superscript of W indicates the number of network layers in the area where the rotor is evaluated in the normal operating state, i.e., W ⁽¹⁾ Parameter, W, representing first-level evaluation element ⁽²⁾ Parameters representing the second level evaluation elements. The subscript of W indicates the number of the connection evaluator, i.e.

In FIG. 7, the small blocks are network evaluation elements for evaluating whether the rotor is in a normal working area, such as evaluation element a ₁₁ 、a ₁₂ 、a ₁₃ . The virtual connecting line is the connection between the evaluation elements, the darker the color of the connecting line of the evaluation elements is, the larger the absolute value of the weight value is, and when the color is changed to white, the weight value is 0. The color of the evaluation element represents an output value at the position coordinate, and the color is darker as the output value is larger.

As shown in fig. 7, x ₁ The evaluation element is x in position coordinate, and the distinguishing plane is y axis. Because the output of the evaluation element is x ₁ Value of itself, so when x ₁ Below 0, the evaluator output is negative, so the left side is orange. When x is ₁ Above 0, the evaluator output is positive, so the right side is blue. x is the number of ₂ The evaluation element is y in the position coordination.

Part 4, the fourth part information is added Sigmoid function.

Fig. 8 is a graph illustrating Sigmoid function. And adding a Sigmoid function to solve the problem that the position point of the axis of the rotor is not distinguishable. Considering that the problem solved by the linear model is limited, the problem of judging whether the rotor position is in the standard normal working area in practice cannot be divided by a plane of a linear or high-dimensional space, that is, the linear model cannot be clearly divided. Therefore, the scheme of the invention adds a Sigmoid function (as shown in fig. 8) to non-linearize the model in order to evaluate whether the rotor position is in the normal working area. The Sigmoid function added is as follows:

FIG. 9 is a flow chart illustrating the execution of the evaluation meta structure incorporating the sigmoid function. The structure of the evaluation element added to the sigmoid function shown in fig. 9 is a general structure of the evaluation element added to the sigmoid function, and is only schematically illustrated here. In FIG. 9, b is the offset, x ₀ 、x ₁ 、x ₂ For input, w ₀ 、w ₁ 、w ₂ A weight for each evaluation element.

The evaluation unit 106 is further configured to determine that the magnetic bearing rotor normally operates according to a set standard operation mode if the current position of the magnetic bearing rotor is in the standard normal operation state region in the evaluation result, and control the magnetic bearing rotor to continue to operate according to the set standard operation mode. The specific function and processing of the evaluation unit 106 are also referred to in step S140.

The evaluation unit 106 is further configured to determine that the magnetic bearing rotor does not normally operate according to a set standard operation mode if the current position of the magnetic bearing rotor is not in the standard normal operating state region in the evaluation result, and estimate whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor. The specific function and processing of the evaluation unit 106 are also shown in step S150.

In some embodiments, the estimating unit 106, estimating whether the magnetic bearing rotor is faulty according to the current position of the magnetic bearing rotor, comprises:

the evaluation unit 106 is in particular further configured to determine whether the current position of the magnetic bearing rotor is within a set allowed range in case the current position of the magnetic bearing rotor is not within the standard normal operating state region. The detailed function and processing of the evaluation unit 106 are also shown in step S410.

The evaluation unit 106, in particular, is further configured to predict a next position of the magnetic bearing rotor if the current position of the magnetic bearing rotor is within the allowed range, in case the current position of the magnetic bearing rotor is not within the standard normal operation state region; and according to the next position of the magnetic bearing rotor, the position evaluation network model is reused to evaluate whether the current position of the magnetic bearing rotor is in a set area so as to evaluate whether the magnetic bearing rotor fails. The next position of the magnetic bearing rotor is the position of the magnetic bearing rotor at the next set time. The detailed function and processing of the evaluation unit 106 are also shown in step S420.

The evaluation unit 106 is in particular further configured to predict the magnetic bearing rotor failure if the current position of the magnetic bearing rotor is not within the allowable range in case the current position of the magnetic bearing rotor is not within the standard normal operating state region. The specific function and processing of the evaluation unit 106 are also referred to in step S430.

The setting area is a standard normal working state area and is used for judging whether the rotor is in a standard working state or not. And an allowable range for determining whether the rotor is in a normal (i.e., non-standard) operating state or a fault state when the rotor is not in the standard operating state. And in a set area, namely a standard normal working state area, determining that the magnetic bearing rotor is in the standard working area, and dividing the magnetic bearing rotor into a normal working area and a fault area when the magnetic bearing rotor is in the non-standard area. The possibility of a fault in the next state of the rotor is predicted even further for the case of normal (non-standard) operating conditions, since the operating state is not the best state, although the rotor is not faulty, and therefore the possibility of a fault in the rotor is estimated.

And 5, judging whether the rotor position is in a standard working area or a non-standard area through the area where the rotor position is in the model after the visual training.

Accordingly, the determination process of fig. 7 is: the evaluation element at Output in FIG. 7 is the Output, showing the network discrimination plane for the evaluation rotor position Output and the discrimination data points for the desired evaluation rotor position network. After 88 rounds (custom) of visual training to evaluate the rotor position network, the results are shown on the right side of fig. 7: the position points of the axis of the rotor are obviously classified, and whether the position of the rotor is in a standard normal working state area or not can be judged through the area map: the upper right blue is a standard normal working area of the rotor, and the lower left orange is a non-standard operating working area of the rotor. Through actual conditions testing, the nonstandard operation working area of the rotor can be further divided into: a normal (non-standard but tolerable) rotor operation zone and a rotor fault zone may be allowed. If the rotor is in a normal operation but non-standard region at the moment, the possibility of the rotor being in a fault can be predicted by evaluating the region where the position point of the rotor is located at the next moment.

Since the processes and functions implemented by the apparatus of this embodiment substantially correspond to the embodiments, principles and examples of the method, reference may be made to the related descriptions in the embodiments without being detailed in the description of this embodiment, which is not described herein again.

By adopting the technical scheme of the invention, the position evaluation network model is constructed, the position evaluation network model is utilized to evaluate whether the real-time position of the magnetic bearing rotor is in a standard normal working state area, the position of the rotor in a non-standard running working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of failure is further evaluated, the automatic monitoring of the position of the magnetic bearing rotor is realized, and the problems of possible overshoot and compressor damage when the displacement sensor of the magnetic bearing system fails are solved.

According to an embodiment of the present invention, there is also provided a magnetic levitation system corresponding to the position estimation apparatus of the magnetic bearing rotor. The magnetic levitation system may include: the position estimating apparatus of a magnetic bearing rotor described above.

Since the processing and functions of the magnetic levitation system of the present embodiment substantially correspond to the embodiments, principles, and examples of the apparatus, reference may be made to the related descriptions in the embodiments without being detailed in the description of the present embodiment, which is not described herein again.

By adopting the technical scheme of the invention, the position evaluation network model is constructed, the position evaluation network model is utilized to evaluate whether the real-time position of the rotor of the magnetic suspension bearing is in a standard normal working state area, the position of the rotor in a non-standard running working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of the fault is further evaluated, the automatic monitoring of the position of the rotor of the magnetic suspension bearing is realized, and the problem that the fault diagnosis and identification accuracy of the rotor of the magnetic suspension bearing is not high and cannot be predicted is solved.

According to an embodiment of the present invention, there is also provided a storage medium corresponding to the position estimation method of a magnetic bearing rotor, the storage medium including a stored program, wherein the apparatus on which the storage medium is located is controlled to perform the above-described position estimation method of a magnetic bearing rotor when the program is run.

Since the processing and functions implemented by the storage medium of this embodiment substantially correspond to the embodiments, principles, and examples of the foregoing method, reference may be made to the related descriptions in the foregoing embodiments without being detailed in the description of this embodiment.

By adopting the technical scheme of the invention, the position evaluation network model is constructed, whether the real-time position of the magnetic suspension bearing rotor is in a standard normal working state area or not is evaluated by utilizing the position evaluation network model, the position of the rotor in a non-standard operation working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of fault occurrence is further evaluated, the automatic monitoring of the position of the magnetic suspension bearing rotor is realized, and the problem that the fault type of the rotor cannot be accurately evaluated by a displacement sensor is solved.

According to an embodiment of the invention, there is also provided a processor corresponding to the method of position estimation of a magnetic bearing rotor, the processor being adapted to run a program, wherein the program is run to perform the method of position estimation of a magnetic bearing rotor as described above.

Since the processing and functions implemented by the processor of this embodiment substantially correspond to the embodiments, principles, and examples of the foregoing method, reference may be made to the related descriptions in the foregoing embodiments without being detailed in the description of this embodiment.

By adopting the technical scheme of the invention, the position evaluation network model is constructed, the position evaluation network model is utilized to evaluate whether the real-time position of the magnetic suspension bearing rotor is in a standard normal working state area, the position of the rotor in a non-standard operation working state is further judged to be in a normal working area or a fault area, the position of the rotor in the normal working area is predicted in the next state, the possibility of fault occurrence is further evaluated, the automatic monitoring of the position of the magnetic suspension bearing rotor is realized, the early diagnosis and predictive maintenance of the magnetic suspension bearing rotor fault are realized, and the safety is better.

In summary, it is readily understood by those skilled in the art that the advantageous modes described above can be freely combined and superimposed without conflict.

The above description is only an example of the present invention, and is not intended to limit the present invention, and various modifications and changes may be made to the present invention by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method of position estimation of a magnetic bearing rotor, comprising:

constructing a position evaluation network model; the position estimation network model is a network model for estimating whether the position of the magnetic bearing rotor is in a set region;

obtaining a current position of the magnetic bearing rotor;

evaluating whether the current position of the magnetic bearing rotor is in a set area by using the position evaluation network model to obtain an evaluation result; the set area is a standard normal working state area of the magnetic bearing rotor;

in the evaluation result, if the current position of the magnetic bearing rotor is in the standard normal working state area, determining that the magnetic bearing rotor normally operates according to a set standard operation mode;

and in the evaluation result, if the current position of the magnetic bearing rotor is not in the standard normal working state area, determining that the magnetic bearing rotor does not normally operate according to a set standard operation mode, and estimating whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor.

2. The method of position estimation of a magnetic bearing rotor of claim 1, wherein constructing a position estimation network model comprises:

building a network model by utilizing a full-connection network and a forward propagation algorithm and adding a Sigmoid function;

and collecting a position sample of the magnetic bearing rotor, and training the network model to obtain a position evaluation network model of the magnetic bearing rotor.

3. The position estimation method of a magnetic bearing rotor according to claim 2, wherein in the case where the depth of the network model is two layers, and three estimation elements are given in the second layer, and one estimation element is given in the third layer, the model of the position estimation network model of the magnetic bearing rotor is defined as follows:

wherein A is ₁ For the evaluation results, x, y are two eigenvectors of the position coordination of the magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ To evaluate the offset of the element; w denotes the parameters in the evaluation element, the superscript of W indicates the number of network layers in the area where the rotor is evaluated in the normal operating state, and the subscript of W indicates the number of the connected evaluation element.

4. The position estimation method of a magnetic bearing rotor according to any one of claims 1 to 3, wherein estimating whether the current position of the magnetic bearing rotor is in a set region using the position estimation network model includes:

performing characteristic extraction based on the current position of the magnetic bearing rotor to obtain the motion track of the magnetic bearing rotor;

judging whether each axis position track point in the motion track of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state or not so as to: under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered to be in the standard normal working state area; and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered not to be in the standard normal working state area.

5. The method for assessing the position of a magnetic bearing rotor as recited in claim 4, wherein estimating whether the magnetic bearing rotor is malfunctioning based on the current position of the magnetic bearing rotor comprises:

determining whether the current position of the magnetic bearing rotor is within a set allowable range if the current position of the magnetic bearing rotor is not within the standard normal operating state region;

predicting a next position of the magnetic bearing rotor if the current position of the magnetic bearing rotor is within the allowable range, if the current position of the magnetic bearing rotor is not within the standard normal operating state region; and according to the next position of the magnetic bearing rotor, the position evaluation network model is reused to evaluate whether the current position of the magnetic bearing rotor is in a set area so as to evaluate whether the magnetic bearing rotor fails; the next position of the magnetic bearing rotor is the position of the magnetic bearing rotor at the next set moment;

predicting the magnetic bearing rotor fault if the current position of the magnetic bearing rotor is not within the allowable range under the condition that the current position of the magnetic bearing rotor is not in the standard normal operating state region.

6. A position estimation device for a magnetic bearing rotor, comprising:

a modeling unit configured to construct a location evaluation network model; the position estimation network model is a network model for estimating whether the position of the magnetic bearing rotor is in a set region;

an acquisition unit configured to acquire a current position of the magnetic bearing rotor;

an evaluation unit configured to evaluate whether the current position of the magnetic bearing rotor is in a set region using the position evaluation network model, resulting in an evaluation result; the set area is a standard normal working state area of the magnetic bearing rotor;

the evaluation unit is further configured to determine that the magnetic bearing rotor is normally operated according to a set standard operation mode if the current position of the magnetic bearing rotor is in the standard normal operating state region in the evaluation result;

the evaluation unit is further configured to determine that the magnetic bearing rotor does not normally operate according to a set standard operation mode if the current position of the magnetic bearing rotor is not in the standard normal operating state region in the evaluation result, and estimate whether the magnetic bearing rotor fails according to the current position of the magnetic bearing rotor.

7. The position estimation device for a magnetic bearing rotor of claim 6, wherein the modeling unit, which constructs a position estimation network model, includes:

8. The position estimation device for a magnetic bearing rotor according to claim 7, wherein in the case where the depth of the network model is two layers, and three estimation elements are given in a second layer, and one estimation element is given in a third layer, the model of the position estimation network model for a magnetic bearing rotor is defined as follows:

wherein A is ₁ For the evaluation results, x, y are two eigenvectors of the position coordination of the magnetic bearing rotor, b ₁ 、b ₂ 、b ₃ To evaluate the offset of the element; w represents the parameter in the evaluation element, and the superscript of W indicates that the rotor is in standard normalityThe number of network layers in the active state area, the subscript of W indicates the number of the connection evaluation element.

9. The position estimation device for a magnetic bearing rotor according to any one of claims 6 to 8, wherein the estimation unit estimates whether the current position of the magnetic bearing rotor is in a set zone using the position estimation network model, including:

extracting features based on the current position of the magnetic bearing rotor to obtain the motion track of the magnetic bearing rotor;

judging whether each axis position track point in the motion track of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state or not so as to: under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is considered to be in the standard normal working state area; and under the condition that each axis position track point in the motion trail of the magnetic bearing rotor is not a position point when the magnetic bearing rotor is in a set standard normal working state, the current position of the magnetic bearing rotor is not in the standard normal working state area.

10. The position estimation apparatus of a magnetic bearing rotor according to any one of claims 6 to 8, wherein the estimation unit estimating whether the magnetic bearing rotor is malfunctioning or not based on the current position of the magnetic bearing rotor includes:

determining whether the current position of the magnetic bearing rotor is within a set allowable range in case that the current position of the magnetic bearing rotor is not in the standard normal operation state region;

predicting the magnetic bearing rotor fault if the current position of the magnetic bearing rotor is not within the allowable range if the current position of the magnetic bearing rotor is not within the standard normal operating condition zone.