Control method for magnetic suspension bearing of molecular pump

Abstract

The invention relates to the technical field of bearing control, in particular to a control method for a magnetic suspension bearing of a molecular pump, and provides the following scheme that the radial displacement observed quantity of a rotor relative to the magnetic suspension bearing is obtained; under the constraint condition of average pumping, zero-mean perturbation is injected into the radial X force channel and the radial Y force channel, and response data are collected; calculating a fractional delay parameter set and an uncertain domain thereof based on the response data; updating a time-varying control matrix according to the parameter set, and generating a control current instruction to drive opposite electromagnetic poles; according to the method, zero-mean constraint and energy balance are simultaneously realized in the time domain and the angle domain, the transverse nonreciprocal fractional delay is accurately identified and compensated, and the stability of the magnetic suspension bearing and the pumping performance of the molecular pump are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bearing control, in particular to a control method for a magnetic bearing of a molecular pump. BACKGROUND

[0002] In the prior art, the magnetic bearing of a molecular pump usually relies on a fixed parameter control method to maintain the radial suspension and stable operation of the rotor. However, due to the cross-influence of the radial X force channel and the radial Y force channel in long-term operation, the traditional control method often has difficulty in maintaining stability under different operating conditions, which easily leads to drift of the rotor track and reduces the operation accuracy. At the same time, the existing method lacks effective means to compensate for the time delay characteristics caused by gas molecules, resulting in insufficient decoupling performance between the radial force channels, unbalanced phenomenon, and thus affecting the pumping efficiency and vacuum quality. In addition, the existing control strategy relies on single signal correction and cannot take into account the constraints in the time domain and the spatial domain, which easily causes the control energy to gather in a local direction, thereby causing vibration enhancement and insufficient stability. In summary, the prior art is difficult to effectively suppress the lateral imbalance and fractional delay while ensuring the continuity of the average pumping, which limits the application performance of the magnetic bearing of the molecular pump under high-precision and high-stability operating conditions.

[0003] To solve the above problems, the present application designs a control method for a magnetic bearing of a molecular pump. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a control method for a magnetic bearing of a molecular pump to solve the problems of the prior art. The radial displacement observation of the rotor relative to the magnetic bearing is obtained. Under the constraint condition of average pumping, zero-mean perturbation is injected into the radial X force channel and the radial Y force channel, and response data is collected. Based on the response data, a set of fractional delay parameters and their uncertainty domain are calculated. The time-varying control matrix is updated according to the parameter set, and a control current instruction is generated to drive the opposing electromagnetic poles. The present application simultaneously realizes zero-mean constraint and energy balance in the time domain and the angle domain, accurately identifies and compensates for the lateral non-reciprocal fractional delay, and effectively improves the stability of the magnetic bearing and the pumping performance of the molecular pump.

[0005] To achieve the above purpose, the present application provides the following technical solutions: A control method for a magnetic bearing of a molecular pump, the molecular pump comprising a rotor, a magnetic bearing, a controller and a stator, wherein the controller is configured with a time-varying control matrix for controlling the radial X force channel and the radial Y force channel of the magnetic bearing, the method comprising: obtaining the radial displacement observation of the rotor relative to the magnetic bearing; According to the radial displacement observation, under the constraint condition of average pumping, zero-mean perturbation is injected into the radial X force channel and the radial Y force channel to obtain response data; A fractional delay parameter set is calculated according to the response data; A time-varying control matrix is updated according to the fractional delay parameter set to generate control current instructions.

[0006] The radial X force channel and the radial Y force channel represent a control loop that generates electromagnetic control forces in the X-axis direction and the Y-axis direction through oppositely arranged pairs of electromagnetic poles arranged radially orthogonally, wherein the X-axis and the Y-axis correspond to a reference coordinate system with the center of the magnetic suspension bearing as the origin.

[0007] According to the radial displacement observation, under the constraint condition of average pumping, zero-mean perturbation is injected into the radial X force channel and the radial Y force channel, including: The radial displacement observation is converted into a trajectory curve according to the frequency of the average pumping; The trajectory curve is corrected by constraint to obtain a reference trajectory curve, wherein the constraint correction includes a time mean constraint and a zero bias constraint; Extreme value detection is performed on the reference trajectory curve to obtain a local maximum value arranged in time sequence and a subsequent local minimum value corresponding to the local maximum value, and a local maximum value and a subsequent local minimum value are summarized to obtain an extreme value pair set; The extreme value pair set is taken as a trigger sequence to calculate a zero-mean perturbation corresponding to the trigger sequence.

[0008] The time mean constraint includes: A periodic sliding window aligned with the frequency is constructed, and the trajectory curve is slid according to the periodic sliding window; In each periodic sliding window, the direct current component corresponding to the trajectory curve is calculated and offset compensated to make the time integral in the periodic sliding window zero; The trajectory curve in the periodic sliding window is divided into multiple angle sectors with the frequency band of the cascade frequency of the magnetic suspension bearing as the angle basis, the energy distribution corresponding to each sector is counted, and the trajectory curve after offset compensation is nonlinearly scaled according to the energy distribution.

[0009] The zero bias constraint includes: Band-limited smoothing and phase-locked processing are performed on the trajectory curve after the time mean constraint to obtain a smoothed trajectory curve; The linear correlation degree of the smoothed trajectory curve with respect to the direction information of the radial displacement observation is calculated as a reference; The linear correlation degree is compared with a preset zero bias threshold, and if the linear correlation degree is greater than or equal to the zero bias threshold, a principal direction of the direction information is calculated by principal component analysis; The smooth trajectory curve is processed according to the deviation between the principal direction and the reference, so that the component corresponding to the smooth trajectory curve is attenuated according to the deviation, to obtain a reference trajectory curve.

[0010] The extreme value pair set is taken as a trigger sequence, and a zero mean perturbation corresponding to the trigger sequence is calculated, including: The angle sector corresponding to the magnetic suspension bearing is quadrantized and partitioned to obtain a reference quadrant, wherein the reference quadrant includes a first quadrant, a second quadrant, a third quadrant, and a fourth quadrant, and each quadrant is allocated according to the angle sector; A codebook composed of a plurality of vector atoms is preset in each quadrant, wherein the vector atom is a perturbation template for perturbing the radial X force channel and the radial Y force channel, the perturbation template at least includes one of a single-lobe short-time template and a double-lobe cancellation template, the vector atoms in the same codebook are equal in area and opposite in sign; According to the extreme value pair set, an atom is selected for each extreme value pair corresponding to a local maximum value and a subsequent local minimum value, wherein the atom selection includes determining the quadrant of the extreme value pair according to the radial direction of the reference trajectory curve at the time corresponding to the extreme value pair, and determining the corresponding vector atom according to the selection order; In the process of atom selection, the remaining amount of each quadrant codebook is recorded, and if the remaining amount is less than a preset amount threshold, the perturbation template of the remaining vector atom in the corresponding codebook is replaced with a double-lobe cancellation template; The vector atoms storing the local maximum value or the subsequent local minimum value are spliced by a greedy algorithm to obtain a candidate perturbation; The candidate perturbation is corrected according to the local maximum value and the subsequent local minimum value to obtain a zero mean perturbation.

[0011] A fractional delay parameter set is calculated according to the response data, including: A phase-locked reference corresponding to the zero mean perturbation is determined, and the response data is coherently demodulated and synchronously averaged with the phase-locked reference as a reference benchmark to extract amplitude information and phase information and combine them to obtain a corresponding phase-locked complex transfer point set; The phase-locked complex transfer point set is preprocessed by screening, and a fitting weight of the phase-locked complex transfer point set after screening preprocessing is calculated according to the signal-to-noise ratio of the remaining phase-locked complex transfer points, wherein the screening preprocessing includes consistency screening and frequency band avoidance; The phase-locked complex transfer point set is amplitude and phase fitted according to the fitting weight to obtain a fractional delay parameter set.

[0012] updating the time-varying control matrix according to the fractional delay parameter set, comprising: calculating a delay equalization matrix and a cross-coupling gain corresponding to the radial X force channel and the radial Y force channel according to the fractional delay parameter set, wherein the delay equalization matrix is calculated by performing decimal order delay approximation fitting on the delay amount in the fractional delay parameter set, and the cross-coupling gain is calculated by performing weighted fitting on the amplitude component and the phase component in the fractional delay parameter set and combining the amplitude-phase statistical result of the cross channel; generating a reverse-coupling feedforward channel according to the cross-coupling gain, embedding a corresponding displacement-current feedback in the time-varying control matrix according to the reverse-coupling feedforward channel to obtain a candidate time-varying control matrix; performing matrix fusion on the candidate time-varying control matrix according to the delay equalization matrix to obtain an updated time-varying control matrix.

[0013] The matrix fusion comprises: calculating the eigenvalue of the delay equalization matrix, and multiplying the eigenvalue with the matrix elements of the candidate time-varying control matrix item by item to obtain the updated time-varying control matrix.

[0014] The method further comprises: performing robustness judgment according to the uncertainty domain; performing multi-vertex constraint and dissipation inequality constraint on the candidate time-varying control matrix according to the robustness judgment to obtain a gain upper limit of the candidate time-varying control matrix; performing amplitude limiting on the compensation gain of the candidate time-varying control matrix according to the gain upper limit to obtain an updated candidate time-varying control matrix.

[0015] Compared with the prior art, the beneficial effects of the present application are: The present application realizes dynamic identification and compensation of the transverse non-reciprocal fractional delay by introducing zero-mean perturbation in the radial X force channel and the radial Y force channel and extracting the fractional delay parameter set combined with the response data, thereby avoiding long-term drift of the rotor track. In the process of updating the time-varying control matrix, delay equalization and reverse-coupling feedforward are introduced at the same time, so that the decoupling performance between the radial channels is improved, and the control accuracy is significantly enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0016] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, made with reference to the accompanying drawings: Figure 1 An exemplary application scenario provided for the embodiments of the present application is shown in the following figure: Figure 2 Problem principle schematic diagram provided for an embodiment of the present application; Figure 3 Further problem principle schematic diagram provided for an embodiment of the present application; Figure 4 Flowchart of a control method for a magnetic bearing of a molecular pump provided by an embodiment of the present application; Figure 5 Flowchart of zero-mean perturbation calculation provided by an embodiment of the present application.

[0017] Reference signs: 100, pump body; 101, rotor; 102, bearing; 103, stator; 104, forepump. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application.

[0019] In this document, the term “embodiment” means that a specific feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It will be apparent to those skilled in the art from the description herein that the embodiments described herein can be combined with other embodiments.

[0020] On the production line of ultra-high vacuum and high-cleanliness manufacturing, molecular pumps are rapidly evolving towards higher rotational speed, smaller air gap, and wider working condition window. In order to reduce the risk of particles and maintenance cost, more and more complete machines use magnetic bearings: The rotor operates in a non-contact state, and the inherent mechanical damping is almost evacuated. Any small transverse coupling or phase lag will be amplified into observable track deformation. This scenario is completely different from traditional bearings or conventional blowers. The control object is no longer a stable and symmetric spring-damping approximation, but a time-varying multi-input multi-output system that wanders with rotational speed and gas state changes.

[0021] In these devices, the dilemma faced by those skilled in the art does not come from information field problems such as data explosion or link congestion: The signal dimension and rate of the molecular pump are far from reaching the bottleneck of the communication link. On the contrary, the difficulty lies in the physical side: The discrete flight of rarefied gas makes the response of aerodynamic force to lateral displacement show a time delay of small order, and the average flight path along two orthogonal radial directions is not equal, resulting in a significant rise of non-reciprocal cross coupling in some speed-back pressure window; magnetic suspension also strips passive dissipation, so these asymmetric and time delay directly enter the closed loop. The common remedy is to increase the trap wave, improve the current loop bandwidth or pre-set the decoupling matrix, but under the double action of working condition drift and blade passage frequency, fixed parameters are often only effective in a narrow window, and cross-window appears elliptical trajectory expansion or even short-term instability. Further, the engineering site rarely allows a large amount of excitation signal to be injected during operation, because it will change the capture angle of the pumping path and the axial net thrust; similarly, installation space and cleanliness restrictions also discourage the addition of additional flow / pressure measurement points.

[0022] It is easy to understand that these realistic constraints are to see the non-reciprocal fractional delay of the object, without disturbing the average pumping; both in the X / Y two force channels to make cross compensation, and to resist the slow drift caused by speed, temperature rise, back pressure; both to improve the track quality, and not to break the safety red line of current loop and overall energy consumption.

[0023] It should be noted that the key to the control problem of the magnetic suspension molecular pump is: how to distinguish and eliminate the anti-symmetric time delay coupling that really destroys the stability domain in limited observable without increasing hardware and disturbing the pumping. The present application seals the energy injection risk with provable average constraints, guarantees the stability of the implementation layer with the approximate realizability of fractional delay, and responds to the slow drift of working conditions with multi-vertex robustness, so as to restore the track quality and stability margin to a reproducible and verifiable level without changing the pump structure and sensor configuration.

[0024] Reference Figure 1 , Figure 1 An example application scenario is provided for the embodiment of the present application.

[0025] Figure 1 An example of a magnetic suspension molecular pump is shown, which specifically includes a pump body 100, a rotor 101, a magnetic suspension bearing 102, a stator 103, and a forepump 104, Figure 1 The arrow in the middle indicates the flow direction of gas molecules.

[0026] It can be understood that when the whole machine is powered on, the controller (not shown in the figure) first realizes the non-contact suspension of the rotor 101 in the pump body 100 through the magnetic suspension bearing 102: the radial and axial positions are obtained by displacement sensing, and the control current generates electromagnetic force in the opposite electromagnetic poles, so that the rotor 101 is kept centered and posed in the micron-level air gap. After completing the stable suspension, the motor (usually integrated coaxially with the stator 103) accelerates the rotor 101 to the rated speed. Because there is no mechanical contact and lubricant, the rotor 101 can obtain high circumferential speed in a very low damping and clean environment, creating the necessary conditions for molecular pumping.

[0027] Further, the pumping process enters from the upper inlet, and the blades of the rotor 101 are arranged at a certain inclination angle and pass through the gas molecules at a high linear speed; the molecules in the free molecular / transition flow state slide and collide with the moving blades multiple times, obtaining the same momentum as the blade movement direction in the tangential direction, and being given an axial component velocity by the blade geometry. The stator 103 blades arranged with opposite or modified inclination angles follow, phase the tangential momentum obtained by the molecules at the rotor 101 and convert it into more stable axial migration, thereby transferring the molecules to the next stage of the blade cascade along the pump axis direction. The momentum transfer of such "rotor acceleration-stator orientation" is repeated at each stage, forming a net axial flux, so that the gas is pushed to the front stage pump 104.

[0028] During operation, the magnetic suspension bearing 102 continuously maintains the radial and axial positions of the rotor 101 in a closed loop manner, and suppresses the orbit deviation caused by imbalance, aerodynamic disturbance or through beat frequency; unlike traditional mechanical bearings, magnetic suspension avoids friction and particle sources, and the stability of the rotor 101 when passing through the critical speed can be guaranteed through the control strategy. Because the bearing 102 is non-contact, the small air gap between the rotor and the stator can be maintained for a long time, and combined with high blade tip linear speed and reasonable blade cascade angle arrangement, efficient pumping and compression of free molecular / transition flow state gas is realized.

[0029] Reference Figure 2 , Figure 2 The problem principle schematic diagram is provided for the embodiment of the application.

[0030] Figure 2 The basic configuration of the magnetic suspension bearing of the molecular pump and the action relationship of the radial force channel thereof are shown. As shown in Figure 2 , the molecular pump includes a rotor and a magnetic suspension bearing for supporting the rotor, the magnetic suspension bearing is arranged in a stator (not shown in the figure), and opposite electromagnetic pole pairs (not shown in the figure) are arranged in the radial direction of the magnetic suspension bearing, which respectively constitute a radial X force channel and a radial Y force channel. By applying differential current in the radial X force channel and the radial Y force channel, electromagnetic control forces in the X direction and the Y direction can be generated respectively, which are used for suspending and dynamically adjusting the displacement of the rotor in the radial direction.

[0031] Figure 2 Further shown is the displacement vector of the rotor in the radial plane, the change of which relative to the center of the magnetic suspension bearing will directly cause cross-coupling between the two radial force channels. During the operation of the molecular pump, additional disturbances to the displacement will be caused by the average pumping effect of the gas molecule flow, so that non-reciprocal fractional delay characteristics are exhibited in different directions, thereby causing imbalance in the control of the radial X force channel and the radial Y force channel.

[0032] Reference Figure 3 , Figure 3 Further problem principle schematic diagram provided for the embodiments of the present application.

[0033] It can be understood that, Figure 3 On the basis of Figure 2 Further explain the displacement offset, Figure 3 In the figure, the black dots represent gas molecules, the black solid lines represent the theoretical displacement of the gas molecules between the rotors, and the black dashed lines represent the actual displacement deflection of the gas molecules between the rotors.

[0034] In an ideal case, the cascade of the molecular pump and the rotor channel should ensure that the gas molecules are transported along a symmetric path by the rotor blades, so that the rotor is not subjected to additional unbalanced forces in the radial direction. However, during high-speed operation and pumping, the offset of the rotor itself in the radial direction will change the effective length of the molecular flight path, so that the arrival time of the molecular group in different directions is different, exhibiting transverse non-reciprocal fractional delay.

[0035] It is easy to understand that in the actual operation of the molecular pump, the rotor is in a high-speed pumping state, and the movement of the gas molecules between the rotor and the stator cascade is no longer an ideal continuous medium process, but presents a discrete flight characteristic, that is, the flight time of the molecules from one blade surface to the next blade surface is required, thereby exhibiting a delay effect in force transmission.

[0036] In theory, if the pump body geometry is completely symmetrical, the molecular flight paths in the X direction and the Y direction should be equal, and the actions of the two radial force channels remain reciprocal; but in reality, this is not the case. As Figure 3 shown, the black solid line is the theoretical motion trajectory of the gas molecules, and the black dashed line is the actual offset trajectory of the gas molecules. Due to the projection difference of the blade angle in the X and Y directions, the assembly eccentricity of the rotor and the stator, and the asymmetric bias of the gas flow in the pumping direction, the average flight path of the molecules when offset in the X direction is not the same as the average flight path when offset in the Y direction.

[0037] Further, when the rotor is offset in the X direction, the coupling force generated in the Y direction is no longer symmetrical with the coupling force generated in the X direction when the rotor is offset in the Y direction, destroying the ideal reciprocity, forming a transverse coupling matrix with an antisymmetric term, and carrying a fractional delay of a small number of sampling periods. Since the rotor of the magnetic suspension molecular pump is completely suspended in a micron-level track without mechanical contact constraints, this transverse non-reciprocal fractional delay effect directly enters the control loop, causing the radial X force channel and the Y force channel to be unbalanced in control performance, which is a technical difficulty that needs to be addressed.

[0038] Next, the control method for the magnetic suspension bearing of the molecular pump provided by the embodiments of the present application will be further described in combination with Figure 4 The control method for the magnetic suspension bearing of the molecular pump provided by the embodiments of the present application will be further described in combination with S1: Obtain a radial displacement observation of the rotor relative to the magnetic suspension bearing; In this embodiment, the radial displacement observation is collected by a displacement sensor arranged in the magnetic suspension bearing. The sensor can be an eddy current sensor, an optical displacement sensor, or other detection elements that can meet the micron-level resolution requirement. The collected data directly reflects the offset of the rotor relative to the geometric center of the magnetic suspension bearing.

[0039] As can be understood by those skilled in the art, the radial displacement observation can be collected according to the actual situation, and the specific sensor type can be selected according to the space condition of the pump body, the resolution requirement, and the vacuum compatibility. The present application does not make more limitations.

[0040] S2: According to the radial displacement observation, a zero-mean perturbation is injected into the radial X force channel and the radial Y force channel under the constraint condition of average pumping to obtain response data; In this embodiment, the controller applies a differential driving current to the X force channel and the Y force channel to construct a zero-mean perturbation sequence, ensuring that the time average of the current is zero in the entire injection window, and suppressing the influence on the average pumping through an angle domain balancing strategy. The triggering time of the perturbation is obtained by extreme value detection of the trajectory curve of the radial displacement observation, ensuring that the perturbation is applied at the most sensitive point of the rotor track to obtain response data with the highest identification efficiency. On the one hand, the perturbation does not change the average pumping performance of the pump, and on the other hand, the obtained response data contains sufficient dynamic information, which is convenient for subsequent modeling of the transverse non-reciprocal fractional delay characteristics.

[0041] As can be understood by those skilled in the art, the specific implementation form of the perturbation signal can be a pulse, a symmetric sequence, or an optimized sparse vector codebook, as long as it can meet the zero-mean and constraint conditions.

[0042] S3: calculating a fractional delay parameter set according to the response data; In this embodiment, the controller first performs phase-locked processing on the response data, extracts the amplitude and phase relationship under different injection frequencies in the rotation coordinate system synchronized in electrical angle, and obtains a series of complex transfer points. Subsequently, a frequency domain fitting method is used to calculate the difference in the molecular flight time in the X direction and the Y direction according to the variation trend of the phase with the frequency, and the fractional delay parameter set is obtained. Unlike traditional small signal modeling, the transverse non-reciprocal effect is directly identified by using the perturbation response here, which can capture the delay asymmetry caused by the molecular flow under real operating conditions.

[0043] Those skilled in the art can understand that the fitting method of the fractional delay parameter can adopt a multiple harmonic fitting, a Prony algorithm or a phase difference least square, which is not limited in the present application.

[0044] S4: updating a time-varying control matrix according to the fractional delay parameter set to generate a control current instruction; In this embodiment, the controller takes the fractional delay parameter as input, dynamically updates the time-varying control matrix, and compensates the cross coupling of the X / Y two radial force channels. Specifically, the uncertainty domain is considered when updating the matrix, so that when the delay parameter fluctuates, the control current instruction can still maintain zero mean and robust stability. The generated control current instruction acts on the opposing electromagnetic poles in the form of differential, and adjusts the position of the rotor in real time.

[0045] Before the specific technical content corresponding to the unfolding step, the embodiments of the present application need to be emphasized again.

[0046] In a typical molecular pump operating environment, the rotor is suspended in the magnetic field at a super high speed, and the radial support force is completely provided by the magnetic bearing. Due to the inevitable small orbit drift of the rotor during pumping, and the fact that the transfer of gas molecules between the narrow leaf channels is not an ideal continuous medium behavior, but presents a random and directional flight feature, the dynamic relationship between the rotor and the radial support force presents a nonlinear and time delay effect that changes with the operating condition. For the control loop, this coupling will not directly manifest as a prominent instability, but more commonly as a gradually accumulated difference between the displacement observation signal and the actual acting force. Especially when there is cross influence in the radial X force channel and the Y force channel, the difference manifests as directional imbalance and dynamic drift of the force matrix, so that the conventional fixed gain control or simple compensation based on displacement deviation is difficult to maintain long-term robustness.

[0047] The present application does not directly aim at identifying more accurately, but first restricts the "energy injection for identification" to a non-invasive process in the sense of pumping, that is, under the multiple restrictions of zero mean in the time domain, balance in the angle domain, and avoidance in the frequency domain, the collection and reconstruction of the closed-loop observable characteristics are completed; then based on the collected phase-locked complex characteristics, the parameterization of the transversely non-reciprocal fractional delay is completed, and the control matrix update is driven in a robust manner, thereby identifying and compensating in the same constraint framework.

[0048] It can be understood that the present application can identify under real pumping conditions without additional flow or pressure sensing, and does not cause accumulative bias to the average pumping.

[0049] Further, the present embodiment does not stop at the data fitting level of parameter identification, but directly uses the obtained fractional delay as the basis for updating the time-varying control matrix. This means that the controller can dynamically adjust the coupling compensation relationship between the radial X and Y force channels according to the real-time observed phase difference and uncertainty range, avoiding overcompensation or undercompensation caused by fixed parameters in traditional methods.

[0050] In engineering practice, this updating method is equivalent to embedding a real-time calibration path in the controller, so that the molecular pump can still maintain orbital stability and mechanical balance under high-speed operation and rapid changes in pumping load.

[0051] In one example, the radial X force channel and the radial Y force channel represent a control loop that generates electromagnetic control forces in the X-axis and Y-axis directions by oppositely arranged pairs of radial orthogonal electromagnetic poles, where the X-axis and Y-axis correspond to a reference coordinate system with the center of the magnetic levitation bearing as the origin.

[0052] Next, the technical content of the zero-mean perturbation of the present application method is further expanded.

[0053] Reference Figure 5 , Figure 5 The flowchart of the zero-mean perturbation calculation provided by the present embodiment is shown.

[0054] In one example, according to the radial displacement observation, a zero-mean perturbation is injected into the radial X force channel and the radial Y force channel under the constraint condition of average pumping, including: S2.1: converting the radial displacement observation into a trajectory curve according to the frequency of the average pumping; Specifically, to inject perturbation without changing the average pumping, double constraints are needed to be imposed on the trajectory curve: one is that the time average is zero, that is, no direct current bias is generated in any integral cycle window; the other is zero bias constraint, that is, the linear bias of the trajectory in the reference direction is suppressed to avoid forming net thrust in the pumping direction. The implementation path is: first estimate the time average of the trajectory curve in the current window and compensate the bias, then construct a decomposition orthogonal to the displacement direction as the reference, suppress the component in the same direction as the reference according to the threshold, and retain or moderately enhance the component orthogonal to the reference, and then perform energy equalization in the angular domain to prevent long-term accumulation of perturbation in a certain angular sector.

[0055] S2.2: constraint correction is performed on the trajectory curve to obtain a reference trajectory curve, wherein the constraint correction comprises a time average constraint and a zero bias constraint;

[0056] In an optional specific embodiment, the time average constraint comprises: A periodic sliding window aligned with the frequency is constructed, and the trajectory curve is slid according to the periodic sliding window; in each periodic sliding window, the direct current component corresponding to the trajectory curve is calculated and bias compensation is performed to make the time integral in the periodic sliding window zero; the trajectory curve in the periodic sliding window is divided into a plurality of angular sectors according to the frequency band of the blade passing frequency of the magnetic suspension bearing as the angular reference, the energy distribution of each sector is counted, and the trajectory curve after bias compensation is nonlinearly scaled according to the energy distribution.

[0057] In yet another optional specific embodiment, the zero bias constraint comprises: The trajectory curve after the time average constraint is subjected to band-limited smoothing and phase locking processing to obtain a smoothed trajectory curve; the linear correlation degree of the smoothed trajectory curve with respect to the reference is calculated based on the direction information of the radial displacement observation; the linear correlation degree is compared with a preset zero bias threshold, and if the linear correlation degree is greater than or equal to the zero bias threshold, the principal direction of the direction information is calculated by principal component analysis; the smoothed trajectory curve is subjected to suppression processing according to the deviation degree between the principal direction and the reference, so that the component corresponding to the smoothed trajectory curve is attenuated according to the deviation degree, to obtain a reference trajectory curve.

[0058] S2.3: extreme value detection is performed on the reference trajectory curve to obtain local maximum values arranged in time sequence and successive local minimum values corresponding to the local maximum values, and the local maximum values and the successive local minimum values are summarized to obtain an extreme value pairing set; Specifically, the extreme value detection is used to determine the trigger node of the perturbation, and requires that the peak-valley event can be stably identified and a reliable pairing relationship can be established even in the presence of noise and slow drift.

[0059] In the present embodiment, the detection process includes: after the reference trajectory curve is subjected to a slow trend separation, local maximum and local minimum are searched using an extreme value threshold with hysteresis; to avoid false positives due to fine fluctuations, a minimum interval and a minimum amplitude difference of the extreme values are set, and only candidate points that meet the interval and amplitude difference conditions are retained; then, each local maximum and its subsequent nearest local minimum are paired in time sequence to form an extreme value pairing set. The extreme value pairing set directly serves the subsequent pulse shaping, ensuring that each pair of trigger events has a natural cancellation basis at the time integration level.

[0060] Further, the extreme value detection is performed under a rotating coordinate reference, the continuity of the displacement direction is checked during peak and valley identification, and if the direction mutation exceeds a threshold, the candidate point is abandoned to ensure the pairing quality; when an isolated peak or an isolated valley occurs and cannot be paired, the event is temporarily stored and is preferentially paired in the next window, and if it still cannot be paired, it is incorporated into adjacent events according to the amplitude halving principle.

[0061] S2.4: Taking the extreme value pairing set as a trigger sequence, a zero-mean perturbation corresponding to the trigger sequence is calculated; It can be understood that in the molecular pump magnetic suspension bearing, the radial displacement will have multiple peak and valley values in one pumping cycle, and these extreme value points are naturally paired, and using the distribution relationship of one positive and one negative, a perturbation sequence with time integration of zero can be constructed, ensuring that the average pumping is not changed in a statistical sense.

[0062] However, if only the extreme value pairs are directly applied with equal amplitude positive and negative pulses, there are two problems that are difficult to overcome in actual operation: First, the flight time of gas molecules between the rotor blade passages is not strictly symmetrical, resulting in a residual direct current component after the perturbation passes through the object, so that the so-called zero mean is destroyed at the output end; Second, the rotor offset direction may be biased to a certain angle in different cycles, and if not limited, the long-term injected perturbation energy will accumulate in a certain direction, thus appearing as an additional disturbance force in the pumping process, thereby destroying the pumping continuity of the molecular pump.

[0063] In short, the traditional peak-valley cancellation method can only guarantee zero mean under ideal conditions, and cannot maintain stability when there is transverse non-reciprocity and gas flow bias.

[0064] In the present embodiment, a quadrant partitioning processing logic is proposed for the foregoing problems. Specifically, the radial plane of the magnetic suspension bearing is divided into four reference quadrants, and each quadrant corresponds to an angle sector. Each pair of extreme value events must first be mapped to the quadrant to which it belongs, and then a preset vector atom (i.e., a perturbation template) is selected to execute triggering in the quadrant.

[0065] It can be understood that by classifying the extreme points in different angular directions by quadrant, the distribution of perturbation energy in the angular domain can be counted in real time, and then the energy balance between the quadrants can be maintained in multiple cycles. The introduction of the quadrant partition makes the zero mean constraint no longer rely on one-to-one cancellation in the time domain, but also adds the balance constraint in the spatial domain, avoiding the accumulation of long-term bias. In addition, under the quadrant partition, an independent quota table can be set for each quadrant. When the cumulative energy of a certain quadrant approaches the upper limit, the subsequent events in the same direction will automatically switch to a reduced amplitude or use a double-lobe cancellation template, achieving dynamic adjustment.

[0066] In one example, the specific steps of S2.4 are as follows: S2.4.1: The angular sector corresponding to the magnetic suspension bearing is partitioned into quadrants to obtain a reference quadrant, wherein the reference quadrant includes a first quadrant, a second quadrant, a third quadrant and a fourth quadrant, each quadrant is allocated according to the angular sector; Specifically, to avoid the long-term accumulation of zero-mean perturbation in a certain direction in the spatial angle, an angular domain management is introduced. A reference coordinate system is established with the center of the magnetic suspension bearing as the origin, the radial X-axis as the zero-degree reference, four reference quadrants are divided in the counterclockwise direction, and each quadrant is further subdivided into several equal-width angular sectors. When the extreme event (local maximum or subsequent local minimum) occurs, the instantaneous direction of the displacement vector is mapped to the specific sector, thereby obtaining a stable spatial attribution.

[0067] In some optional embodiments, to suppress boundary jitter, an angular dead zone and a hysteresis zone are set at the boundary of the quadrant: when the direction falls within the dead zone, the adjacent two quadrants are weighted according to the proximity; when it falls within the hysteresis zone, the existing attribution of the same type of event in the previous window is used, to ensure the time continuity and traceability of the quadrant attribution.

[0068] In this embodiment, each reference quadrant is associated with an angular domain quota table, which records the cumulative energy, remaining quota, and the latest trigger time of the quadrant in the current injection window; the quota table is initialized according to the balance target at the beginning of the window, and is updated immediately after each event mapping. The event mapping generates a quadrant marker and a sector marker, which are used for subsequent template selection and energy balancing.

[0069] S2.4.2: A codebook composed of multiple vector atoms is preset in each quadrant, wherein the vector atom is a perturbation template used to perturb the radial X force channel and the radial Y force channel, the perturbation template includes at least one of a single-lobe short-time template and a double-lobe cancellation template, the vector atoms in the same codebook are area-equivalent and sign-alternating opposite; Specifically, to achieve controllable and verifiable perturbation synthesis, a set of vector atoms is pre-configured in each quadrant. Each atom contains metadata such as direction, amplitude upper limit, duration, rise / fall edge speed limit, minimum trigger interval, phase code, etc., and provides two types of envelopes: single-lobe short-time templates (used to release minimum energy and improve recognition sensitivity) and double-lobe cancellation templates (used for in-place zero-sum and delay rectification suppression). Within the same codebook, positive and negative paired atoms are designed to be area-equivalent and sign-alternating, ensuring that a replacement can achieve zero-sum locally.

[0070] In this embodiment, the codebook manages amplitude and pulse width in discrete gears. Typically, several amplitude gears (low, medium, high of the rated differential current) and several pulse width gears (short, medium, long) are set, and sparse replacement templates are prepared for the sensitive frequency band. Template metadata is persistently stored, facilitating quick replacement, splitting, or cross-sector rearrangement when backtracking at the window end. After codebookization, each release of perturbation can be combined within the safety boundary, energy budget, and frequency band mask, reducing the uncertainty of temporarily generated waveforms.

[0071] S2.4.3: According to the extreme value pairing set, select an atom for each extreme value pair corresponding to a local maximum and a subsequent local minimum, wherein the atom selection includes determining the quadrant of the extreme value pair according to the radial direction of the reference trajectory curve at the time corresponding to the extreme value pair, and determining the corresponding vector atom according to the selection order; Specifically, the extreme value pairing set provides time sequence and event type. Atom selection follows two core rules: One is direction consistency, that is, the radial direction of the reference trajectory at the time of extreme value occurrence is read, and after mapping the event to a specific quadrant, only "positive atoms" (peaks) or "negative atoms" (valleys) can be selected from the codebook of the corresponding quadrant; The second is pairing consistency, that is, peak events and their subsequent valley events preferentially select the same pair of positive / negative area-equivalent templates to form a one-to-one cancellation locally.

[0072] When the event points across the sector boundary, a half-amplitude dependent strategy is adopted, which proportionally allocates the pair of atoms to the same templates in adjacent sectors, preserving the direction information and avoiding energy mutations caused by boundary jitter.

[0073] S2.4.4: During the atom selection process, record the remaining amount of each quadrant codebook. If the remaining amount is less than a preset amount threshold, replace the perturbation templates of the remaining vector atoms in the corresponding codebook with double-lobe cancellation templates; In this embodiment, the quadrant quota table is read during atomic selection. If the remaining capacity of a quadrant is low, the double petal cancellation template is preferred. If the remaining capacity is sufficient and the sector has not been covered for a long time, the selection probability of the single petal template is increased to improve the recognition effectiveness. To improve the triggering feasibility, the selector simultaneously checks the minimum trigger interval and the up / down edge speed limit. If it conflicts with the previous selected atom, it is automatically switched to a template with shorter pulse width or delayed execution. The atomic selection result is written into the scheduling queue with information such as timestamp, quadrant marker, template number, etc. to provide complete metadata for subsequent splicing and backtracking.

[0074] In this embodiment, to prevent long-term imbalance of space energy, the quadrant quota table is continuously refreshed during atomic selection and execution. Each selection deducts the corresponding area budget from the remaining capacity of the corresponding quadrant, and records the recent trigger time and cumulative energy. When the remaining capacity is below the threshold, subsequent events in the same direction are no longer allowed to select single petal high amplitude templates, but are automatically replaced with double petal cancellation templates to suppress the continuous rise of energy in the quadrant from the source. The replacement action is completed when the event is enqueued, avoiding the discovery of insufficient quota until the splicing stage.

[0075] In some optional embodiments, the threshold setting adopts a grading strategy: The first threshold enables the amplitude reduction, and the second threshold forces the switch to a double petal template and prohibits the energy from being moved to the end of the window for concentrated release. To avoid frequent switching near the threshold, the quota table uses hysteresis judgment for threshold comparison; for burst dense events, the minimum trigger interval limit is also linked to force the adjacent sectors to digest part of the energy.

[0076] S2.4.5: The vector atoms storing local maximum values or subsequent local minimum values are spliced by a greedy algorithm to obtain candidate perturbations; Specifically, the scheduling queue reads the selected atoms in timestamp order and uses a greedy strategy to place them on the injection time axis one by one: for each atom, it is preferentially attempted to be injected at the nominal trigger time; if it conflicts with the minimum interval or edge speed limit of the already placed atoms, its trigger time is adjusted or a short pulse width template in the same quadrant is used within the local time window until the engineering constraints are met. During the splicing process, the sensitive frequency band mask is checked simultaneously. When the atom envelope covers the sensitive frequency band, the combined modification of "delaying a small amount and shortening the pulse width" is preferred; if it cannot be avoided, the atom is marked as delayed to the next window for execution, and a small amount of placeholder energy is released in the current window with a low amplitude cancellation template to maintain the zero-sum structure of the whole window.

[0077] In the embodiment, the greedy splicing updates the full window statistics, including the time integral accumulation and the reference direction projection accumulation, every time an atom is incorporated, so that the deviation can be corrected in time when it is found early, without waiting until the end of the window to backtrack massively. For the two atoms of the same extreme value pair, if the nominal time interval is too short to cause a conflict, the peak side atom is retained according to the priority and the valley side atom is split into two half amplitude atoms, which are positioned at a small offset before and after, satisfying the interval constraint and maintaining the local zero sum of the event pair.

[0078] S2.4.6: modifying the candidate perturbation according to the local maximum and the subsequent local minimum to obtain a zero mean perturbation; In the embodiment, the backtracking correction follows the minimum change strategy: the amplitude range and the micro segment proportion are adjusted first, followed by the pulse width range, then the template replacement, and finally the cross sector rearrangement; the two statistics are recalculated immediately after each change until both indicators meet the requirements. After the correction, the final perturbation is mapped to the differential current command of the two radial force channels according to the X and Y components of the reference coordinates, and the amplitude clamping, the adjacent sampling point change rate limitation and the minimum interval limitation are applied to ensure the drivable side and long-term reliability.

[0079] Next, the technical content of the method of the application on the fractional delay parameter set is further expanded.

[0080] It can be understood that the fractional delay parameter set in the application can be understood as a quantitative result of the dynamic characteristics of the rotor radial displacement in different directions due to the time difference of gas molecules. The fractional delay parameter set not only includes the average delay amount extracted in the radial X direction and the radial Y direction, but also includes the non-reciprocity characteristics and the uncertainty range between the two directions.

[0081] In the actual operation process of the molecular pump magnetic suspension bearing, due to the assembly tolerance between the rotor and the stator, the projection difference of the spiral angle of the leaf channel in different directions, and the asymmetry of the gas flow in the pumping direction, the effective flight time experienced by the gas molecules when transferring from a certain radial offset state to the next blade is not exactly the same in the X direction and the Y direction. Therefore, in the control loop, the disturbance response of the radial displacement will exhibit a phenomenon of fractional delay with a fractional sampling period delay, which is called fractional delay.

[0082] In the embodiment, the construction process of the fractional delay parameter set includes two dimensions: on the one hand, the amplitude and phase extraction is performed on the phase-locked response data to obtain a complex transfer point set at different injection frequencies; on the other hand, the slope of the phase of the complex transfer point with respect to the frequency is fitted to obtain the average flight delay of the gas molecules in the X direction and the Y direction, and the difference between the two is extracted as the non-reciprocal fractional delay parameter.

[0083] In some optional embodiments, the fractional delay parameter set is not limited to a simple delay value, but can be extended to a multi-dimensional data structure containing complex features. For example, the set can include a dependence of the delay diagonal domain, i.e. whether the delays extracted in different quadrants have systematic differences; it can also include a time-varying trend quantization of the delay, to describe the drift speed and fluctuation amplitude of the delay parameter when the rotation speed slowly changes or the pumping pressure fluctuates. Through such an extended set structure, when updating the time-varying control matrix, the controller can no longer rely on a single nominal delay value, but can schedule and compensate according to the statistical characteristics of the entire set, thereby achieving more accurate and robust suppression of the transverse non-reciprocal fractional delay.

[0084] In one example, calculating a fractional delay parameter set according to the response data comprises: S3.1: determining a phase-locked reference corresponding to the zero-mean perturbation, and performing coherent demodulation and synchronous averaging on the response data with the phase-locked reference as a reference benchmark, to extract amplitude information and phase information and combine them to obtain a corresponding phase-locked complex transfer point set; S3.2: performing screening preprocessing on the phase-locked complex transfer point set, and calculating fitting weights of the phase-locked complex transfer point set after screening preprocessing according to the signal-to-noise ratio corresponding to the remaining phase-locked complex transfer points, wherein the screening preprocessing includes consistency screening and frequency band avoidance; S3.3: performing amplitude-phase fitting on the phase-locked complex transfer point set according to the fitting weights, to obtain a fractional delay parameter set.

[0085] Next, the technical content of the method of the present application regarding the time-varying control matrix is further expanded.

[0086] In one example, updating a time-varying control matrix according to the fractional delay parameter set comprises: S4.1: calculating delay equalization matrices and cross-coupling gains corresponding to the radial X force channel and the radial Y force channel according to the fractional delay parameter set, wherein the delay equalization matrices are calculated by performing decimal order delay approximation fitting on the delay amounts in the fractional delay parameter set, and the cross-coupling gains are calculated by performing weighted fitting on the amplitude components and phase components in the fractional delay parameter set and combining the amplitude-phase statistical results of the cross-channels; Specifically, the fractional delay parameter set includes the delay amount in radial X and Y directions, the amplitude and phase statistics of the cross-channel, and the confidence interval of each parameter. To eliminate the group delay mismatch caused by the inconsistency of the delay in two directions, a delay equalization matrix is constructed, which is essentially a fractional order delay approximation to the signals of the X→Y, Y→X cross paths and their respective direct channels, so that the equivalent phase curve within the bandwidth available to the controller is aligned with the nominal object. To counteract the anti-symmetrical coupling caused by the transverse non-reciprocity, the cross-coupling gain is calculated based on the amplitude and phase statistics in the set, which can be understood as a function of speed and thermal state. Delay equalization and cross-gain cooperate with each other, one solves the time structure mismatch, and the other suppresses the directional coupling, forming an embeddable parameter pair.

[0087] In this embodiment, delay equalization uses a combination of multi-order all-pass approximation and band-limited fitting: first, on the frequency point set matching the displacement ring closed-loop bandwidth, the phase-frequency curve is fitted according to the weight, and the weight is determined by the signal-to-noise ratio at the time of injection and the angular domain coverage; then select the order and coefficient, so that the residual phase error after equalization is limited to the preset tolerance; at the same time, the low confidence points are removed by the outlier suppression strategy to ensure the stability of the approximation. The calculation of the cross-coupling gain uses amplitude and phase joint fitting, and uses quadrant trimming statistics as a priori to avoid misjudging the spatial bias as an inherent coupling of the object; the fitting result constructs a two-variable interpolation table with speed and temperature rise as independent variables, and the table entries contain mean and upper and lower limits, which are used for subsequent robust constraints.

[0088] S4.2: generating a reverse-coupling feedforward channel according to the cross-coupling gain, embedding a corresponding displacement-current feedback in the time-varying control matrix according to the reverse-coupling feedforward channel, to obtain a candidate time-varying control matrix; Specifically, in the X channel, a feedforward correction derived from the Y displacement (or its trajectory shaping quantity) is superimposed, and in the Y channel, a feedforward correction derived from the X displacement is superimposed, and the amplitude and phase of the two are jointly set according to the gain table and the delay equalization result. Through this embedding, a controllable cancellation channel can be established for the transverse anti-symmetrical coupling without sacrificing the original damping and stiffness design.

[0089] In this embodiment, the feedforward channel uses double-buffering coefficient switching and amplitude limiting at the digital implementation level: One set of coefficients is used for online calculation, and the other set is used as output, and seamless switching is performed at the window boundary to avoid instantaneous jumps; each output feedforward correction is subjected to amplitude clamping, slope limiting and minimum interval constraint to ensure that the current loop can withstand. To reduce interference on the sensitive frequency band, a small shaping network with band-stop and phase advance is built into the feedforward path, so that the feedforward energy avoids the passage frequency and rectification harmonic neighborhood of the cascade, and is consistent with the delay equalization matrix in phase.

[0090] S4.3: matrix fusion of the candidate time-varying control matrix according to the delay equalization matrix, to obtain an updated time-varying control matrix; Specifically, equalization is applied in a series manner in the channel inner link, and equalization is applied in a parallel phase alignment manner in the cross feed branch, so that the equivalent group delay of the channel and the cross branch is consistent, and the amplitude correlation with the nominal feedback is maintained. The fusion process includes consistency checking, checking whether the influence of energy dissipation and current margin after equalization is within the tolerance, if not, reducing the cross feed weight or narrowing the equalization bandwidth according to the priority, until both checks pass at the same time. The fused matrix is used as the updated time-varying control matrix for real-time generation of differential current instructions of the X and Y channels.

[0091] In some optional embodiments, the matrix fusion includes: calculating eigenvalues of the delay equalization matrix, and multiplying the eigenvalues with matrix elements of the candidate time-varying control matrix item by item to obtain the updated time-varying control matrix.

[0092] In one example, calculating the fractional delay parameter set according to the response data further includes calculating an uncertainty domain of the fractional delay parameter set, and the method further includes: making a robustness judgment according to the uncertainty domain; performing multi-vertex constraint and dissipation inequality constraint on the candidate time-varying control matrix according to the robustness judgment, to obtain a gain upper limit of the candidate time-varying control matrix; clipping a compensation gain of the candidate time-varying control matrix according to the gain upper limit, to obtain an updated candidate time-varying control matrix.

[0093] Although the embodiments of the present application have been shown and described above, it should be understood that the above-described embodiments are exemplary and should not be construed as limiting the present application, and those of ordinary skill in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. A control method for a magnetic levitation bearing of a molecular pump, characterized by, The molecular pump comprises a rotor, a magnetic suspension bearing, a controller and a stator, wherein a time-varying control matrix for controlling radial X force channels and radial Y force channels of the magnetic suspension bearing is configured in the controller, and the method comprises: acquiring a radial displacement observation of the rotor relative to the magnetic suspension bearing; according to the radial displacement observation, injecting zero-mean perturbations into the radial X force channels and the radial Y force channels under the constraint condition of average pumping to obtain response data; calculating a set of fractional delay parameters according to the response data; updating the time-varying control matrix according to the set of fractional delay parameters to generate control current instructions.

2. The control method for a magnetic levitation bearing of a molecular pump according to claim 1, wherein The radial X force channels and the radial Y force channels represent a control loop which generates electromagnetic control forces in the X-axis direction and the Y-axis direction through oppositely arranged pairs of radially orthogonal electromagnetic poles, wherein the X-axis and the Y-axis correspond to a reference coordinate system with the center of the magnetic suspension bearing as the origin.

3. The control method for a magnetic levitation bearing of a molecular pump according to claim 1, wherein According to the radial displacement observation, injecting zero-mean perturbations into the radial X force channels and the radial Y force channels under the constraint condition of average pumping, comprises: transforming the radial displacement observation into a trajectory curve according to the frequency of the average pumping; performing constraint correction on the trajectory curve to obtain a reference trajectory curve, wherein the constraint correction comprises a time mean constraint and a zero bias constraint; performing extreme value detection on the reference trajectory curve to obtain locally maximum values arranged in time sequence and successive locally minimum values corresponding to the locally maximum values, and collecting the locally maximum values and the successive locally minimum values to obtain a set of extreme value pairs; calculating zero-mean perturbations corresponding to the set of extreme value pairs as a trigger sequence.

4. The control method for a magnetic levitation bearing of a molecular pump according to claim 3, wherein The time mean constraint comprises: constructing a periodic sliding window aligned with the frequency, and sliding the trajectory curve according to the periodic sliding window; in each periodic sliding window, calculating a direct current component corresponding to the trajectory curve and performing bias compensation to make the time integral in the periodic sliding window zero; dividing the trajectory curve in the periodic sliding window into multiple angle sectors according to the frequency band of the through frequency of the blade row of the magnetic suspension bearing as an angle basis, counting the energy distribution corresponding to each sector, and performing nonlinear scaling on the trajectory curve after bias compensation according to the energy distribution.

5. The control method for a magnetic levitation bearing of a molecular pump according to claim 3, wherein The zero bias constraint comprises: performing band-limited smoothing and phase locking processing on the trajectory curve after the time mean constraint to obtain a smoothed trajectory curve; calculating the linear correlation degree of the smoothed trajectory curve relative to the reference according to the direction information of the radial displacement observation; comparing the linear correlation degree with a preset zero bias threshold value, if the linear correlation degree is greater than or equal to the zero bias threshold value, calculating the main direction of the direction information through principal component analysis; performing pressure drop processing on the smoothed trajectory curve according to the deviation degree between the main direction and the reference, so that the components corresponding to the smoothed trajectory curve are attenuated according to the deviation degree to obtain a reference trajectory curve.

6. The control method for a magnetic levitation bearing of a molecular pump according to claim 3, wherein Calculating zero-mean perturbations corresponding to the set of extreme value pairs as a trigger sequence, comprises: The corresponding angle sector of the magnetic suspension bearing is quadrantized and partitioned to obtain a reference quadrant, wherein the reference quadrant includes a first quadrant, a second quadrant, a third quadrant, and a fourth quadrant, each quadrant is allocated according to an angle sector; A codebook composed of a plurality of vector atoms is preset in each quadrant, wherein the vector atom is a perturbation template for perturbing the radial X force channel and the radial Y force channel, the perturbation template at least includes one of a single-lobe short-time template and a double-lobe cancellation template, the vector atoms in the same codebook are equal in area and opposite in sign; According to the extreme value pairing set, an atom is selected for each extreme value pair corresponding to a local maximum value and a subsequent local minimum value, wherein the atom selection includes determining the quadrant of the extreme value pair according to the radial direction of the reference trajectory curve at the time corresponding to the extreme value pair, and determining the corresponding vector atom according to the selection order; During the atom selection process, the remaining amount of each quadrant codebook is recorded, and if the remaining amount is less than a preset amount threshold, the perturbation templates of the remaining vector atoms in the corresponding codebook are replaced with double-lobe cancellation templates; The vector atoms storing the local maximum value or the subsequent local minimum value are spliced by a greedy algorithm to obtain a candidate perturbation; The candidate perturbation is corrected according to the local maximum value and the subsequent local minimum value to obtain a zero-mean perturbation.

7. The control method for a magnetic levitation bearing of a molecular pump according to claim 1, wherein The fractional delay parameter set is calculated according to the response data, including: A phase-locked reference corresponding to the zero-mean perturbation is determined, the response data is coherently demodulated and synchronously averaged with the phase-locked reference as a reference benchmark, amplitude information and phase information are extracted and combined to obtain a corresponding phase-locked complex transfer point set; The phase-locked complex transfer point set is screened and preprocessed, and the fitting weight of the screened and preprocessed phase-locked complex transfer point set is calculated according to the signal-to-noise ratio of the remaining phase-locked complex transfer points, wherein the screening and preprocessing includes consistency screening and frequency band avoidance; The phase-locked complex transfer point set is amplitude-phase fitted according to the fitting weight to obtain the fractional delay parameter set.

8. The control method for a magnetic levitation bearing of a molecular pump according to claim 7, wherein The time-varying control matrix is updated according to the fractional delay parameter set, including: The delay equalization matrix and the cross-coupling gain corresponding to the radial X force channel and the radial Y force channel are calculated according to the fractional delay parameter set, wherein the delay equalization matrix is calculated by approximating and fitting the delay amount in the fractional delay parameter set to a decimal order, and the cross-coupling gain is calculated by weighting and fitting the amplitude component and the phase component in the fractional delay parameter set and combining the amplitude-phase statistical results of the cross-channel; A reverse-coupled feedforward channel is generated according to the cross-coupling gain, and a corresponding displacement-current feedback is embedded in the time-varying control matrix according to the reverse-coupled feedforward channel to obtain a candidate time-varying control matrix; The candidate time-varying control matrix is matrix fused according to the delay equalization matrix to obtain an updated time-varying control matrix.

9. The control method for a magnetic levitation bearing of a molecular pump according to claim 8, wherein The matrix fusion includes: The eigenvalues of the delay equalization matrix are calculated, and the updated time-varying control matrix is obtained by multiplying the eigenvalues with the matrix elements of the candidate time-varying control matrix item by item.

10. The control method for a magnetic levitation bearing of a molecular pump according to claim 8, wherein The method further includes calculating an uncertainty domain of the fractional delay parameter set according to the response data, and performing a robustness judgment according to the uncertainty domain. The method further includes performing a multi-vertex constraint and a dissipation inequality constraint on the candidate time-varying control matrix according to the robustness judgment, to obtain a gain upper limit of the candidate time-varying control matrix. The method further includes performing amplitude limiting on a compensation gain of the candidate time-varying control matrix according to the gain upper limit, to obtain an updated candidate time-varying control matrix. ​

Images