A magnetic levitation vacuum pump and its control method

Abstract

The present invention provides a magnetic levitation vacuum pump and a control method thereof, which includes a casing. A drive shaft is disposed in cooperation with the casing. An impeller is disposed in cooperation with the casing at one end of the drive shaft. A motor and a combined magnetic bearing unit are disposed outside the drive shaft behind the impeller. A combined displacement detection unit is disposed in cooperation with the combined magnetic bearing unit. A controller is disposed in cooperation with the motor, the combined magnetic bearing unit and the combined displacement detection unit. The combined displacement detection unit obtains combined displacement detection data, and the controller adjusts the outputs of the combined magnetic bearing unit and the motor based on the trained improved prediction model. The control method performs feedback control, collects sufficient historical data, and trains the improved prediction model based on this. The feed-forward control of the magnetic levitation vacuum pump is realized through the improved prediction model. The present invention realizes the feed-forward control of the magnetic levitation vacuum pump, improves the use efficiency, prolongs the service life, and realizes the reproduction of different control measures and their effects; it can be applied to quickly detect the fault point and eliminate the fault.

Description

Technical Field

[0001] The present invention relates to the technical field of non-variable displacement pumps, and particularly to a magnetic levitation vacuum pump and a control method thereof. Background Art

[0002] A magnetic levitation vacuum pump is a vacuum device that uses magnetic force to drive a rotating body. Through the action of magnetic force, the impeller (and its drive shaft) is suspended in space, forming a relatively stable gap between the impeller and the pump housing, thereby achieving a non-contact vacuum working environment. Then, liquid or gas is inhaled at the inlet of the pump and accelerated to a high-speed state under the action of the impeller and discharged from the outlet.

[0003] The working principle of the magnetic levitation vacuum pump enables it to avoid the friction and wear caused by traditional mechanical bearings during use, thereby significantly improving the efficiency and service life of the pump. Its durable characteristics are particularly prominent, greatly reducing the energy loss caused by friction. Compared with traditional vacuum pumps, it can save 30% - 70% of energy, and also greatly reduces the maintenance frequency. Moreover, since the direct contact during the working process is almost zero, the noise generated during its operation is relatively low, with less pollution, making it more environmentally friendly.

[0004] Considering ensuring the normal operation and service life of the magnetic levitation vacuum pump, during the operation process, real-time adjustment needs to be carried out based on the point displacement of the drive shaft, so that the drive shaft and the impeller always remain within a reasonable operating range, and then the control scheme is optimized.

[0005] In the prior art, generally, an axial displacement sensor is arranged at the motor position. The displacement monitoring at a single position is not perfect for monitoring the state of the drive shaft, and the radial displacement monitoring at a single position cannot optimize the control method of the magnetic levitation bearing. Although the axial displacement sensor monitors the movement of the drive shaft in the length direction, in fact, the ultimate goal is to monitor the movement of the impeller. There are certain deviations in the above monitoring methods.

[0006] Chinese Patent No. CN101243263A discloses a magnetic levitation bearing device with an improved vacuum feedthrough, which discloses a first sensor board, a second sensor board, and a third sensor board, attempting to achieve accurate monitoring of the drive shaft displacement. However, its monitoring also belongs to feedback monitoring and ultimately performs "feedback control", which can only optimize the overall control of the drive shaft and cannot fundamentally solve the problem of the drive shaft's position deviation. Summary of the Invention

[0007] To solve the above technical problems, the present invention provides a magnetic levitation vacuum pump and a control method thereof.

[0008] The technical solution adopted by the present invention is a magnetic levitation vacuum pump, which includes a housing. A drive shaft is arranged inside the housing in cooperation. An impeller is arranged inside the housing at one end of the drive shaft. An electric motor and a combined magnetic bearing unit are arranged outside the drive shaft behind the impeller. A combined displacement detection unit is arranged in cooperation with the combined magnetic bearing unit. A controller is arranged in cooperation with the electric motor, the combined magnetic bearing unit and the combined displacement detection unit.

[0009] The combined displacement detection unit obtains combined displacement detection data, and the controller adjusts the outputs of the combined magnetic bearing unit and the electric motor based on the trained improved prediction model.

[0010] Preferably, the combined magnetic bearing unit includes a radial magnetic levitation bearing arranged outside the drive shaft at both ends of the electric motor respectively, and an axial magnetic levitation bearing arranged outside the two radial magnetic levitation bearings respectively. A fixed disk is arranged on the drive shaft in cooperation with the axial magnetic levitation bearing. The radial magnetic levitation bearing and the axial magnetic levitation bearing form a balance system with the impeller.

[0011] Preferably, the combined displacement detection unit includes a radial displacement sensor arranged in cooperation with any one of the radial magnetic levitation bearings and an axial displacement sensor arranged in cooperation with any one of the fixed disks.

[0012] Preferably, any one of the axial displacement sensors includes one or more measured light points arranged outside the fixed disk. A detection window is arranged in cooperation with the measured light points, and the detection window is arranged in cooperation with the controller.

[0013] Preferably, an XOY coordinate system is established in cooperation with the detection window, and the origin of the XOY coordinate system corresponds to the position of the measured light point. When any one of the measured light points has a displacement, it is controlled and adjusted by the controller, and the displacement vectors of the two measured light points are obtained and fed back to the controller.

[0014] Preferably, the improved prediction model includes a double-branch structure;

[0015] One branch includes a number of convolutional layers connected in sequence. A temporal feature aggregation layer is arranged after each convolutional layer. A fully connected layer, a softmax layer and an output layer are arranged in cooperation after the last temporal feature aggregation layer. One or more convolutional blocks are arranged in parallel inside any one of the convolutional layers;

[0016] The other branch includes a number of LSTM units, and the output of any one LSTM is associated with the input of the corresponding temporal feature aggregation layer.

[0017] Preferably, the trend of continuous data is obtained as a training label; the data for training the improved prediction model includes the detection data of the continuous combined displacement detection unit and the corresponding control parameters, and the detection data is used to characterize the spatial position of the drive shaft in the current spatial coordinate system; the paired detection data and control parameters with time series characteristics are used as training data pairs and input into the improved prediction model for training in cooperation with the training label.

[0018] Preferably, the casing includes a main casing and a volute casing which are cooperatively arranged. The volute casing is provided with an air inlet in the axial direction facing the impeller, and an air outlet is provided at the position where the volute casing cooperates with the blades of the impeller; an elastic member is provided between the assembly positions of the main casing and the volute casing.

[0019] Preferably, a temperature sensor is provided in the main casing, and the temperature sensor is cooperatively arranged with the controller; a plurality of flow guiding holes are provided on the main casing.

[0020] A control method for the magnetic levitation vacuum pump as described above, the method comprising the following steps:

[0021] S1 Assemble the magnetic levitation vacuum pump, initialize the combined displacement detection unit, and obtain an initial signal at the controller;

[0022] S2 Obtain the feedback signal and control parameters of the radial displacement sensor; obtain the signal that the measured light spot of the axial displacement sensor is captured in the detection window;

[0023] Obtain the displacement vectors of two measured light spots and the relationship parameters between them, record the time and the corresponding control parameters, and make manual markings;

[0024] S3 Obtain sufficient non-empty detection data of the continuous combined displacement detection unit and the corresponding control parameters, obtain the manually marked training label, and input it into the improved prediction model for training;

[0025] S4 Obtain several groups of non-empty detection data of the continuous combined displacement detection unit and input them into the improved prediction model. The improved prediction model outputs classification information to the controller, and the controller adjusts the control parameters for control adjustment.

[0026] The present invention provides a magnetic levitation vacuum pump and its control method, which includes a machine housing. A drive shaft is arranged inside the machine housing in a matching manner. An impeller is arranged inside the machine housing at one end of the drive shaft in a matching manner. An electric motor and a combined magnetic bearing unit are arranged outside the drive shaft behind the impeller in a matching manner. A combined displacement detection unit is arranged in cooperation with the combined magnetic bearing unit. A controller is arranged in cooperation with the electric motor, the combined magnetic bearing unit and the combined displacement detection unit. The combined displacement detection unit obtains combined displacement detection data, and the controller adjusts the outputs of the combined magnetic bearing unit and the electric motor based on the trained improved prediction model. The control method realizes feedforward control of the magnetic levitation vacuum pump through feedback control, collects sufficient historical data, and trains the improved prediction model based on this, and finally realizes feedforward control of the magnetic levitation vacuum pump through the improved prediction model.

[0027] The beneficial effects of the present invention are as follows:

[0028] (1) Improve the control logic of the magnetic levitation vacuum pump. After collecting sufficient data during operation or testing and realizing feedback control, train the improved prediction model based on the collected data, and realize feedforward control of the magnetic levitation vacuum pump based on the improved prediction model, so as to improve the use efficiency of the magnetic levitation vacuum pump and extend its service life;

[0029] (2) Adopt a combined displacement detection unit to obtain more detailed state representations and spatial forms of the drive shaft, which are used to fit the detailed features of the drive shaft in different states, and then realize the reproduction of different control measures and their effects;

[0030] (3) It can be applied to quickly detect fault points and troubleshoot faults. Description of the Drawings

[0031] Figure 1 It is a schematic cross-sectional view structure diagram of the present invention;

[0032] Figure 2 It is a schematic cross-sectional view structure diagram of the assembly surface of the main housing and the volute of the present invention;

[0033] Figure 3 It is a block diagram of the improved prediction model of the present invention. Detailed Embodiments

[0034] The following further describes the present invention in detail in conjunction with embodiments, but the protection scope of the present invention is not limited thereto.

[0035] Embodiment 1

[0036] A magnetic levitation vacuum pump includes a casing. A drive shaft 1 is provided inside the casing in a mating manner. Inside the casing at one end of the drive shaft 1, an impeller 2 is provided in a mating manner. Outside the drive shaft 1 behind the impeller 2, a motor 3 and a combined magnetic bearing unit are provided in a mating manner. A combined displacement detection unit is provided in conjunction with the combined magnetic bearing unit; a controller is provided in conjunction with the motor 3, the combined magnetic bearing unit, and the combined displacement detection unit;

[0037] The combined displacement detection unit obtains combined displacement detection data, and the controller adjusts the outputs of the combined magnetic bearing unit and the motor 3 based on the trained improved prediction model.

[0038] In this embodiment, the rotation of the drive shaft 1 drives the impeller 2 to rotate. Here, the impeller 2 is generally a three-dimensional flow impeller 2. After being processed and optimized through processes, it can accelerate and discharge gas under high-speed rotation, realizing the transportation of gas and the generation of vacuum, and is particularly suitable for the medical field, laboratory equipment, environmental protection field, industrial field, aerospace field, etc.

[0039] In this embodiment, the casing is an assembled structure. In order to better achieve the bearing of the drive shaft 1, the impeller 2, etc., multiple supports are provided inside the casing for axially or circumferentially limiting the positions of the motor 3 and the combined magnetic bearing unit, etc. This is content that is easily understood by those skilled in the art;

[0040] For each independent magnetic bearing of the combined magnetic bearing unit, a number of electromagnets are symmetrically arranged at its center for the controller to control the independent electromagnets of each independent magnetic bearing, thereby realizing the spatial centering of the drive shaft 1 during operation;

[0041] The motor 3 is implemented as a high-speed permanent magnet synchronous motor 3, which has the characteristics of high efficiency and energy saving and can cooperate with the magnetic bearing to make the impeller 2 rotate at high speed.

[0042] In this embodiment, more importantly, a combined displacement detection unit is provided corresponding to the combined magnetic bearing unit to fully obtain the combined displacement detection data of different magnetic bearing units. After cold start and turning to normal operation, the data collected is trained and applied to the improved prediction model. Through the output of the model, classification (trend) is obtained, and then the outputs of the combined magnetic bearing unit and the motor 3 are continuously adjusted in a feedforward manner to obtain better control results; the controller realizes functions such as speed control of the motor 3, stable control of the bearing (electromagnet control), and monitoring and protection of the operating state of the entire vacuum pump in practical applications, and then automatically adjusts the operating parameters of the vacuum pump according to the actual working conditions to ensure its efficient and stable operation.

[0043] Embodiment 2

[0044] On the basis of Embodiment 1, the combined magnetic bearing unit includes a radial magnetic suspension bearing 4 respectively arranged outside the driving shaft 1 at both ends of the motor 3, and an axial magnetic suspension bearing 5 respectively arranged outside the two radial magnetic suspension bearings 4. A fixed disk 6 is arranged on the driving shaft 1 cooperating with the axial magnetic suspension bearing 5; the radial magnetic suspension bearing 4 and the axial magnetic suspension bearing 5 and the impeller 2 form a balance system.

[0045] The combined displacement detection unit includes a radial displacement sensor (not shown in the figure) arranged in cooperation with any one of the radial magnetic suspension bearings 4 and an axial displacement sensor 7 arranged in cooperation with any one of the fixed disks 6.

[0046] Compared with the conventional technology, in this embodiment, axial magnetic suspension bearings 5 are respectively arranged on the opposite outer sides of the two radial magnetic suspension bearings 4. The main purpose is to obtain more accurate axial displacement data and the difference between the two; on the basis that the radial magnetic suspension bearing 4, the axial magnetic suspension bearing 5 and the impeller 2 form a balance system, the axial magnetic suspension bearing 5 also plays more balancing functions; the balance system here means that during operation, the torque of a certain point on the impeller 2 and the driving shaft 1 is equal to or basically equal to the torque of the radial magnetic suspension bearing 4 and the axial magnetic suspension bearing 5 at this point.

[0047] In this embodiment, in order to obtain more accurate axial displacement data, in addition to arranging a radial displacement sensor in cooperation with each radial magnetic suspension bearing 4, an axial displacement sensor 7 is also arranged in cooperation with each axial magnetic suspension bearing 5. Specifically, it is arranged on the fixed disk 6 to obtain more detailed performance of the two ends of the driving shaft 1, including but not limited to vibration, deformation, runout, etc.

[0048] Embodiment 3

[0049] On the basis of Embodiment 2, any one of the axial displacement sensors 7 includes one or more measured light points arranged outside the fixed disk 6. A detection window 9 is arranged in cooperation with the measured light points, and the detection window 9 is arranged in cooperation with the controller.

[0050] Furthermore, an XOY coordinate system is established in cooperation with the detection window 9, with the origin of the XOY coordinate system corresponding to the position of the measured light point. When any one of the measured light points has a displacement, it is controlled and adjusted by the controller, and the displacement vectors of the two measured light points are obtained and fed back to the controller.

[0051] The radial displacement sensor is applied to the annular radial magnetic suspension bearing 4, and its control is easy to implement. In the prior art, the axial displacement sensor 7 often measures extremely small displacement changes through inductance. However, the data obtained in the electromagnetic environment may not be accurate, there may be certain jitter, and it is impossible to obtain tiny longitudinal displacements (such as vibrations) well. Therefore, in this embodiment, the measured light spot is directly set outside the fixed disk 6, and the detection window 9 can be correspondingly set in the packaging shell or the main housing 10;

[0052] The detection window 9 provides a relatively large range. In this relatively large range, the horizontal displacement and "jumping" conditions and their differences of the measured light spot at the end of the drive shaft 12 can be obtained better, and then better state characteristics of the drive shaft 1 and / or the impeller 2 can be obtained. During the high-speed movement of the drive shaft 1, in fact, only one measured light spot is needed to ensure that it is theoretically always captured by the same position of the detection window 9. This "capture" can be achieved by adjusting the detection frequency of the detection window 9;

[0053] Specifically, the detection window 9 is a rectangular window. Generally, its center point is used as the origin and calibrated with the corresponding measured light spot. Starting from the origin as the vector starting point, in principle, the most ideal state is that the measured light spot is always at the origin (or the initial position), but this is obviously impossible. When the measured light spot moves including the X-axis, Y-axis and between quadrants, it indicates that control adjustment is required, and the vector data is recorded after the control adjustment to collect historical data for the training of the subsequent improved prediction model. It should be particularly noted that when there are differences between two vectors at the same moment, the characteristic values represented by the differences are more important. This difference is reflected in the included angle and time delay of the displacement vectors at both ends of the drive shaft 1. Of course, the time delay here can also be expressed by the included angle.

[0054] Considering the closed environment inside the housing, there is almost no interfering light (background light) for the measured light spot, and more accurate detection results can be obtained.

[0055] Embodiment 4

[0056] Based on Embodiment 1 or 3, the improved prediction model includes a dual-branch structure;

[0057] One branch includes a number of convolutional layers connected in sequence. A temporal feature aggregation layer is set after each convolutional layer. After the last temporal feature aggregation layer, a fully connected layer, a softmax layer and an output layer are provided. One or more convolutional blocks are arranged in parallel within any of the convolutional layers;

[0058] The other branch includes a number of LSTM units, and the output of any LSTM is associated with the input of the corresponding temporal feature aggregation layer.

[0059] In this embodiment, an improved prediction model is proposed. In fact, the prediction model here fully considers the temporal characteristics of the recorded feature data and is different from the prior art. In this embodiment, the drive shaft 1 is simulated in a spatial environment, that is, the theoretical encapsulation housing is used as the spatial environment, and the shape of the drive shaft 1 in space at each moment can be obtained, so as to better obtain the acquisition of the abnormal characteristics of the drive shaft 1, especially when vibration and jump occur;

[0060] Based on this, 3 to 5 consecutive convolutional layers are set in this embodiment. In fact, each convolutional layer also includes 1 to 3 convolutional blocks arranged in parallel, generally 3, for extracting the feature expressions of each surface after the drive shaft 1 is placed in space. A temporal feature aggregation layer is set after each convolutional layer to obtain the associated features of any temporal sequence obtained based on LSTM in another branch, so as to better allocate weights. Finally, after the fully connected layer and the softmax layer, the output is output by the output layer. The output here is the probability distribution of classification. Specifically, it is the probability distribution of different operating problems. Based on this predicted value, the controller can quickly adjust in a feedforward manner to avoid the occurrence of serious problems and better realize the operation of the magnetic levitation vacuum pump.

[0061] In this embodiment, when training the data of the improved prediction model, the trend of continuous data is obtained as the training label; the data for training the improved prediction model includes the detection data of the continuous combined displacement detection unit and the corresponding control parameters. The detection data is used to represent the spatial position of the drive shaft 1 in the current spatial coordinate system; the paired detection data and control parameters with temporal characteristics are used as the training data pair. The detection data is input into the first branch, and the control parameters are input into another branch, and training is carried out in cooperation with the training label;

[0062] Furthermore, an error function is constructed to adjust the model parameters, including the cross-entropy loss L1 of the model under the true label and the regularization loss of the model parameters. The overall loss L satisfies,

[0063] L = λ 1 L1 + λ 2 ||Θ|| 2 2

[0064] where λ 1 and λ 2 are adjustable parameters between 0 and 1. The second term is the L2 regularization loss to prevent the model from overfitting.

[0065] Embodiment 5

[0066] On the basis of Embodiment 1, the casing includes a main casing 10 and a volute 11 which are cooperatively arranged. The volute 11 is provided with an air inlet 12 axially facing the impeller 2, and an air outlet 13 is provided at the position where the volute 11 cooperates with the blades of the impeller 2; an elastic member 16 is provided between the assembly sites of the main casing 10 and the volute 11.

[0067] A temperature sensor (not shown in the figure) is provided in the main casing 10, and the temperature sensor is cooperatively arranged with the controller; a plurality of diversion holes 14 are provided on the main casing 10.

[0068] In this embodiment, the main casing 10 and the volute 11 can be integrally arranged or can be installed separately to achieve joint operation, which is convenient for maintenance; in the case of being separated, the two are closely cooperated through an elastic member. Specifically, alternately corresponding trapezoidal positive and negative tooth-shaped structures 15 are coaxially arranged on the docking surfaces of the main casing 10 and the volute 11 as assembly sites. While ensuring the installation accuracy, an elastic member 16 is arranged inside the tooth-shaped structure 15, reserving enough vibration margin. After being connected by a fastener 8, the cooperation of the two is ensured while reducing the hard wear between the two. Of course, elastic sealing rings 17 are provided on one side of the two facing the main casing 10 to achieve a sealing effect.

[0069] Considering that a relatively large number of displacement sensors are used in this embodiment and there may be a certain amount of heat dissipation, temperature sensors are distributed in the main casing 10. Through the control and feedback of the temperature by the temperature sensors, the controller is coordinated to control the speed regulation of the drive shaft 1, so that the cold air outside can be introduced from the diversion holes 14 at an appropriate time; in practical applications, the diversion holes 14 are arranged in a "slant-inserted" manner and are consistent with the rectification direction of the blades, so that no air supply occurs when the drive shaft 1 is moving rapidly, and the channel is opened when the drive shaft 1 decelerates to a certain extent to achieve air supply and temperature reduction treatment; in order to reduce wind noise, the diversion holes 14 are generally only correspondingly arranged near the axial displacement sensor 7.

[0070] Embodiment 6

[0071] On the basis of the above embodiments, the present invention also relates to a control method for the magnetic levitation vacuum pump described above. The following is a specific description of the method:

[0072] S1 Assemble the magnetic levitation vacuum pump, initialize the combined displacement detection unit, and obtain an initial signal at the controller;

[0073] In particular, it is necessary to initialize the cooperation between the detection window 9 and the measured light spot and obtain an initial signal. If the measured light spot is not captured, the working state of the vacuum pump and the working state of the entire axial displacement sensor 7 (including the detection window 9 and the measured light spot) are judged to troubleshoot faults;

[0074] S2 Obtain the feedback signal and control parameters of the radial displacement sensor; obtain the signal that the measured light spot of the axial displacement sensor 7 is captured in the detection window 9;

[0075] Obtain the displacement vectors of the two measured light spots and the relationship parameters between them, record the time and the corresponding control parameters, and manually mark them;

[0076] Specifically, if the measured light spot has a longitudinal displacement, it actually indicates that the drive shaft 1 has a dynamic energy consumption situation. It is necessary to record the time and manually mark it, and classify it into specific training labels accordingly; however, in actual applications, the relationship parameters of the displacement vectors of the two measured light spots, such as the included angle, are more important. When the two measured light spots simultaneously have a longitudinal displacement but the included angle is less than the preset value, only the current in the radial magnetic levitation bearing 4 needs to be synchronously controlled. However, if the included angle is greater than the preset value, the drive shaft 1 may even collide with the bearing, and this situation needs to be avoided in the subsequent feedforward control;

[0077] If the measured light spot only has a lateral displacement, this is a normal process adjustment, and only the time and the corresponding control parameters need to be recorded;

[0078] S3 Obtain sufficient non-empty detection data of the continuous combined displacement detection unit and the corresponding control parameters, obtain the manually marked training labels, and input them into the improved prediction model for training;

[0079] S4 Obtain several groups of non-empty detection data of the continuous combined displacement detection unit and input them into the improved prediction model. The improved prediction model outputs classification information to the controller, and the controller adjusts the control parameters for control adjustment.

[0080] When the controller of the present invention is working, it also relates to a computer-readable storage medium, on which a control program of the magnetic levitation vacuum pump is stored. When the program is executed by a processor, the control method of the magnetic levitation vacuum pump is implemented.

[0081] When the controller of the present invention is working, it may also relate to a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the control method of the magnetic levitation vacuum pump is implemented.

[0082] In the implementation of the present invention, it also involves the training and testing of the improved prediction model, which is easily understood by those skilled in the art, and those skilled in the art can implement it according to the actual situation.

[0083] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0084] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0085] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that realizes the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0087] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A magnetic levitation vacuum pump, characterized in that: It comprises a casing, a driving shaft is arranged in the casing, an impeller is arranged in the casing at one end of the driving shaft, a motor and a combined magnetic bearing unit are arranged outside the driving shaft behind the impeller, the combined magnetic bearing unit comprises radial magnetic bearings arranged outside the driving shaft at both ends of the motor respectively and axial magnetic bearings arranged outside the two radial magnetic bearings respectively, a fixed disk is arranged on the driving shaft that cooperates with the axial magnetic bearings; the radial magnetic bearings and the axial magnetic bearings form a balancing system with the impeller; A combined displacement detection unit is provided in conjunction with the combined magnetic bearing unit, and the combined displacement detection unit includes a radial displacement sensor provided in conjunction with any radial magnetic bearing and an axial displacement sensor provided in conjunction with any fixed disk; any of the axial displacement sensors includes one or more measured light points provided outside the fixed disk, and a detection window is provided in conjunction with the measured light points, and the detection window is provided in conjunction with the controller; A controller is provided in conjunction with the motor, the combined magnetic bearing unit and the combined displacement detection unit; The improved prediction model includes a dual-branch structure; One branch includes several convolutional layers connected in sequence, each convolutional layer is followed by a time series feature aggregation layer, and the last time series feature aggregation layer is followed by a fully connected layer, a softmax layer and an output layer; one or more convolutional blocks are arranged in parallel in any of the convolutional layers; The other branch includes several LSTM units, and the output of any LSTM is associated with the input of the corresponding temporal feature aggregation layer; Get the trend of continuous data as training labels; The data used to train the improved prediction model include detection data of the continuous combined displacement detection unit and corresponding control parameters, the detection data being used to characterize the spatial position of the drive shaft in the current spatial coordinate system; Using paired detection data and control parameters with time series characteristics as training data pairs, inputting the improved prediction model, and training with the training labels; The combined displacement detection unit obtains combined displacement detection data, and the controller adjusts the output of the combined magnetic bearing unit and the motor based on the trained improved prediction model.

2. A magnetic levitation vacuum pump according to claim 1, characterized in that: An XOY coordinate system is established in conjunction with the detection window, with the origin of the XOY coordinate system corresponding to the position of the measured light point. When any measured light point is displaced, the controller controls the adjustment and obtains the displacement vectors of the two measured light points and feeds them back to the controller.

3. A magnetic levitation vacuum pump according to claim 1, characterized in that: The casing comprises a main casing and a volute which are arranged in coordination with each other. The volute is provided with an air inlet in the axial direction of the impeller, and the volute is provided with an air outlet at the blades of the impeller. An elastic member is provided between the assembly points of the main casing and the volute.

4. A magnetic levitation vacuum pump according to claim 3, characterized in that: A temperature sensor is arranged in the main housing, and the temperature sensor is arranged in cooperation with the controller; a plurality of guide holes are arranged on the main housing.

5. A control method for a magnetic levitation vacuum pump according to any one of claims 1 to 4, characterized in that: The method comprises the following steps: S1 is equipped with a magnetic levitation vacuum pump, the combined displacement detection unit is initialized, and the controller obtains the initial signal; S2 obtains the feedback signal and control parameters of the radial displacement sensor; obtains the signal of the measured light point of the axial displacement sensor captured in the detection window; Obtain the displacement vectors of the two measured light points and their relationship parameters, record the time and corresponding control parameters, and mark them manually; S3 obtains sufficient non-empty detection data and corresponding control parameters of the continuous combined displacement detection unit, obtains manually labeled training labels, and inputs them into the improved prediction model for training; S4: obtaining a plurality of groups of continuous non-empty detection data of the combined displacement detection units and inputting them into an improved prediction model; the improved prediction model outputs classification information to a controller; and the controller adjusts control parameters to perform control adjustment.

Images