Magnetic bearing suspension recovery and drop protection system and method

Abstract

The application discloses a magnetic suspension bearing suspension recovery and falling protection system and method, and belongs to the technical field of magnetic suspension bearings. Mainly includes the main structure of the magnetic suspension bearing, the system identification module, the state discrimination module, the fault-tolerant control module, the hierarchical falling protection module and the cooperative decision and state monitoring module. The system identification module inverts system parameters and fault characteristics through closed-loop excitation and multi-source signal acquisition; the state discrimination module generates suspension state identification and protection trigger signals in combination with threshold judgment and feature fusion; the fault-tolerant control module generates fault-tolerant control current and suspension recovery force based on extended state observation, robust feedback correction and dynamic feedforward compensation; the hierarchical falling protection module sequentially executes three-stage protection actions of pre-contact buffer, rigid limiting and damping energy absorption; the application can improve the recovery capability of the magnetic suspension bearing under the working conditions of power failure, load disturbance and coil failure, and reduce the impact, temperature rise and wear in the falling process.

Description

Technical Field

[0001] This invention relates to the field of magnetic levitation bearing technology, and in particular to a magnetic levitation bearing suspension recovery and fall protection system and method. Background Technology

[0002] Magnetic levitation bearings achieve stable levitation between the rotor and stator without mechanical contact through controllable electromagnetic force. They offer a series of outstanding advantages, including no mechanical friction, no lubrication, no oil pollution, high speed, long lifespan, and active vibration control, and have gradually become core supporting components in high-end equipment. In equipment such as flywheel energy storage, high-speed molecular pumps, semiconductor heat treatment equipment, and rail transit traction motors, magnetic levitation bearings can significantly improve system efficiency, reduce maintenance costs, and extend service life, thus possessing extremely high engineering application value and promising prospects for industrial promotion.

[0003] However, magnetic levitation bearing systems are typical complex systems involving multiple physical fields coupled together, including electromechanical, magnetic, and control systems. Their closed-loop levitation height depends heavily on sensor signals, power amplification drive, coil excitation, control algorithms, and the stable and reliable operation of the power supply system. In actual industrial and extreme operating conditions, magnetic levitation bearing systems inevitably encounter the following typical failures: 1) Instantaneous power outages, power grid fluctuations, and drive module power failures cause the electromagnetic levitation force to disappear instantly, and the rotor falls rapidly under the influence of gravity, inertial force, or load impact; 2) Sudden unbalanced disturbances, load changes, and external impact excitations cause the rotor to deviate from its equilibrium position instantaneously, exceeding the controllable range of the magnetic bearing and causing levitation failure; 3) Short circuits or open circuits in the magnetic bearing coils due to insulation aging, overcurrent, or excessive temperature rise cause a local loss of electromagnetic force, resulting in rotor imbalance and deflection; 4) Noise, wire breaks, or failure of the displacement sensor lead to the loss of closed-loop control signals and instability of the control system; 5) When the rotor falls at high speed, it makes high-speed impact contact with the protective bearing, generating huge instantaneous contact force, severe frictional temperature rise, and severe wear, which can easily cause rotor rubbing, protective bearing breakage, and permanent damage to the entire machine.

[0004] Existing magnetic levitation bearing protection technologies generally suffer from the following significant drawbacks: First, traditional controllers are designed only for normal levitation conditions, primarily using algorithms such as PID, LQR, and sliding mode control. In scenarios involving large rotor displacement, large disturbances, and nonlinear contact force coupling caused by faults, they are prone to problems such as current saturation, limited output, and control gain mismatch, failing to achieve rapid, stable, and reliable levitation recovery. Second, fault diagnosis often relies on single-signal judgment, lacking the fusion of multi-source information such as coil current, rotor displacement, and speed fluctuations. This results in low fault identification accuracy and high latency, hindering precise fault type differentiation and rapid fault location. Third, coil fault tolerance often employs redundant coils and multi-phase redundant topologies, leading to complex structures, high costs, and large sizes. First, the engineering application of this technology is difficult, and it lacks a magnetic circuit reconstruction and magnetic compensation mechanism based on stator windings with tapped terminals. Second, the fall protection relies solely on a single TD bearing for mechanical limiting, which cannot effectively buffer the impact of high-speed falls. The contact force is large, the temperature rises rapidly, and the wear is severe, making it difficult to meet the requirements of high reliability and long life. Third, the suspension recovery control and fall protection mechanisms are independent of each other, lacking a unified collaborative decision-making and status monitoring mechanism. This makes it impossible to achieve integrated coordination of the entire process from fault perception to control switching, suspension recovery, fault-tolerant operation, and buffer landing, resulting in insufficient safety redundancy. Fourth, for axial gravity-type flywheel energy storage systems, there is a lack of specially designed graded buffering, step-by-step unloading, and damping energy absorption structures, making it impossible to achieve stable fall protection along the axial gravity direction.

[0005] In summary, developing a magnetic levitation bearing suspension recovery and fall protection system that integrates multi-source fault accurate diagnosis, dual-mode robust control switching, coil failure magnetic reconstruction, three-level buffering for energy absorption, and integrated collaborative decision-making has become a core technological bottleneck for improving the reliability, safety, and engineering applicability of magnetic levitation bearing equipment. It is of great significance for promoting the large-scale application of magnetic levitation technology under high-speed, heavy-load, and extreme working conditions. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a magnetic levitation bearing suspension recovery and fall protection system and method. Through the coordinated work of system identification, state judgment, fault-tolerant control, graded fall protection, and collaborative decision-making and state monitoring, it achieves timely identification of fault states, suspension recovery after instability, and graded buffer protection after recovery failure, thereby reducing rotor misalignment, contact impact, temperature rise and wear, and improving the safety, stability and reliability of the system.

[0007] The technical solution adopted in this invention is as follows: A magnetic levitation bearing levitation recovery and fall protection system, comprising: The main structure of the magnetic levitation bearing includes a stator core, stator windings, rotor shaft, rotor core, radial magnetic bearing, axial magnetic bearing, protective bearing, limiting structure, base, displacement sensor, speed sensor and temperature sensor; The stator winding is disposed on the stator core, the rotor core is fitted on the outside of the rotor shaft, and the radial magnetic bearing, axial magnetic bearing, protective bearing and limiting structure are respectively installed on the base and arranged around the rotor shaft; the rotor core and the stator core form a radial working air gap, and the rotor core and the axial magnetic bearing form an axial working air gap. The displacement sensor, speed sensor, and temperature sensor are fixedly installed on the base, with their detection ends facing the rotor shaft, rotor core, and stator winding, respectively. The system identification module is configured to generate the mass matrix, stiffness matrix, damping matrix, residual unbalance vector, and fault feature vector. The status determination module is configured to generate a floating status indicator and a protection trigger signal; The fault-tolerant control module is configured to generate real-time fault-tolerant control current and levitation restoring force to achieve rapid levitation recovery after instability. A graded fall protection module, which is configured to provide three levels of fall protection; Collaborative decision-making and status monitoring module; it is configured to achieve collaborative control of suspension recovery and fall protection; The system identification module, state discrimination module, fault-tolerant control module, graded fall protection module, and collaborative decision-making and state monitoring module are all electrically or mechanically connected to the main structure of the magnetic levitation bearing.

[0008] Furthermore, the protective bearing includes a permanent magnet protective bearing arranged along the axial direction and a ground contact protective bearing arranged along the radial direction; the permanent magnet protective bearing adopts a Halbach array to provide non-contact unloading force when the rotor deviates from the equilibrium position before mechanical contact occurs; the ground contact protective bearing is used to bear the remaining load after the displacement reaches the limit threshold, and together with the buffer pad, it limits the maximum radial and axial displacement of the rotor.

[0009] Furthermore, the stator winding is provided with tap terminals at preset number of turns positions; when the state discrimination module identifies a local winding open circuit, short circuit or overcurrent fault, the fault-tolerant control module isolates the faulty winding and adjusts the effective excitation turns and electromagnetic force distribution through tap switching, normal winding current redistribution and feedforward compensation, so that the resultant force and torque on the rotor tend to be balanced again.

[0010] Furthermore, the closed-loop excitation unit is used to inject a sinusoidal excitation signal into the magnetic bearing coil and simultaneously acquire displacement, current, temperature, and speed signals; Parameter inversion unit, used for electromechanical coupling dynamics model: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix of the magnetic levitation rotor system, respectively; x is the generalized displacement vector of the rotor; ẋ and ẍ are the first and second derivatives of x with respect to time t, respectively; and F(t) is the resultant force vector acting on the rotor. The resultant force includes at least electromagnetic force and external disturbance force, and may include contact force when contact occurs. The parameter inversion unit inverts the mass matrix m, stiffness matrix k, and damping matrix c through the least squares algorithm and the generalized eigenvalue decomposition algorithm.

[0011] The imbalance identification unit is used to generate residual imbalance vectors by inversely solving based on vibration residuals and displacement response. The fault feature extraction unit is used to generate fault feature vectors based on overcurrent, over-displacement, over-temperature, and speed fluctuation features through a feature fusion algorithm.

[0012] Furthermore, the state determination module includes: The speed tracking unit is used to calculate the speed in real time based on a phase-locked loop algorithm or a Kalman filter algorithm. The status judgment unit is used to compare displacement, current, and rotation speed with preset thresholds and generate four states: normal, warning, fault, and fall through logical judgment. The fault classification unit is used to generate Level 1 warning, Level 2 fault tolerance, Level 3 protection, and Level 4 fall protection levels based on fault duration, signal out-of-tolerance amplitude, and out-of-tolerance frequency using a fuzzy classification algorithm. The protection triggering unit is used to trigger corresponding protection actions based on status indicators and fault levels through timing logic.

[0013] Furthermore, the fault-tolerant control module includes: Extended state observation unit, used for real-time observation of disturbances and uncertainties based on the extended state observer algorithm: Where e is the displacement tracking error, xd is the desired rotor displacement, and x is the actual displacement measured by the displacement sensor; x1 and x2 are the first and second estimated states of the extended state observer, respectively, used to characterize the rotor displacement and the generalized velocity state including unmodeled dynamic effects; l1 and l2 are the extended state observer gains; f(x,ẋ,t) is the total disturbance function composed of system nonlinearity, parameter perturbation, and external disturbance; u is the control input after input gain normalization, t is time, and the "·" above the variable indicates the first derivative with respect to time.

[0014] A robust feedback control unit is used to generate fault-tolerant control current based on sliding mode robust control algorithm and feedback correction algorithm; The dynamic feedforward compensation unit is used to generate feedforward compensation force based on the unbalance vector and disturbance observations to suppress vibration and eccentric impact. The attitude correction unit is used to quickly pull the rotor back to the center equilibrium position based on a three-channel cooperative control algorithm.

[0015] Furthermore, the graded fall protection module includes: Pre-contact buffer unit, used to reduce falling speed and impact energy based on elastic damping buffer algorithm and velocity suppression algorithm; Rigid limiting constraint unit, used to limit maximum displacement based on displacement constraint algorithm and radial limiting algorithm; Damping energy absorption unit, used to convert falling kinetic energy into thermal energy and deformation energy based on viscoelastic damping energy absorption algorithm and kinetic energy conversion algorithm.

[0016] Furthermore, the collaborative decision-making and state monitoring module includes a multi-objective optimization unit, which is configured to construct the following multi-objective optimization function: Wherein, J is the comprehensive performance index of the suspension recovery and fall protection process; tr is the recovery time from fault triggering to rotor displacement recovery and sustained stability within the safety threshold; Fs is the peak impact force during the fall or contact process; ΔT is the maximum temperature rise during the protection process; W is the wear amount or equivalent wear index of the protective bearing and buffer structure; λ1, λ2, λ3, and λ4 are the non-negative weighting coefficients of the corresponding targets, and λ1+λ2+λ3+λ4=1. To eliminate the influence caused by the different dimensions and orders of magnitude of each target, tr, Fs, ΔT, and W are normalized before weighted summation. The collaborative decision-making unit is configured to implement coordinated actions for suspension recovery and fall protection based on time-series collaborative algorithms and logical judgment algorithms; The data recording unit is configured to store fault information, protection action sequence, impact force, displacement, and temperature data in real time. The human-machine interface unit is configured to display status, fault information, and protection actions, and supports manual reset and manual emergency stop.

[0017] A method for implementing a magnetic levitation bearing suspension recovery and fall protection system is also provided, which includes the following steps: Step 1: During the normal operation of the system, the radial displacement signal of the rotor, the excitation current signal of the magnetic bearing coil, and the instantaneous speed signal of the rotor are collected by the displacement sensor, the current detection unit, and the speed encoder, respectively. The displacement signal, current signal, and speed signal are synchronously sampled through a unified clock and then filtered and anti-interference processed to construct the system state vector. Step 2: Fault detection and status identification are performed based on the displacement signal, current signal, and speed signal: The displacement signal is compared with a preset safety threshold. When the displacement amplitude exceeds the threshold and the duration exceeds the set time window, it is determined that the rotor has abnormally deflected; the current signal is judged by a threshold. When the current exceeds the upper threshold or is lower than the lower threshold, it is determined to be an overcurrent fault or a coil open circuit fault, respectively; when multiple criteria meet the combined logic conditions, it is comprehensively determined that the system has experienced a levitation failure. Step 3: After determining that the levitation has failed, the collaborative decision-making and status monitoring module outputs a control switching command to switch the normal performance controller to a recovery controller. The recovery controller is designed based on the dynamic model of the magnetic levitation rotor system, and its system model is as follows: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix, respectively, x is the generalized displacement vector of the rotor, and F(t) is the resultant force vector acting on the rotor. Step four: The recovery controller outputs a control current command to gradually restore the magnetic bearing's ability to levitate and support the rotor; during the recovery control execution process, the graded fall protection module is simultaneously put into a preparatory state or a buffer state to achieve the synergistic effect of control recovery and structural protection. Step 5: Once the rotor displacement is detected to have entered the safe threshold range and remained stable for a preset time, a controller cut-back operation is executed. During the cut-back process, an exponential smoothing function is used to achieve a continuous transition of the control signal, expressed as follows: Where u(t) is the smooth control output actually applied to the power amplifier, ur(t) is the recovery controller output, up(t) is the normal performance controller output, α(t) is the continuous switching coefficient ranging from 0 to 1, and t is time; when switching from normal control to recovery control, α(t) changes continuously from 0 to 1, and when recovery is completed and the switchback is completed, α(t) changes continuously from 1 to 0; Step 6: If the system fails to complete the suspension recovery within the preset time, the recovery is deemed to have failed, and the system enters the fall protection phase. The graded fall protection module sequentially performs three levels of protection actions: pre-contact buffer, rigid limit, and damping energy absorption.

[0018] Furthermore, in step four, the recovery controller uses a μ-integrated robust control method to control the system. By performing structured modeling of the parameter uncertainties in the system, the optimal controller parameters are solved using the DK iterative algorithm. In step six, during the fall protection stage, the permanent magnet protection bearing set in the axial direction first provides axial support force to reduce the rotor's falling speed. Subsequently, the ground contact protection bearing set in the radial direction bears the remaining load and absorbs the impact energy through an elastic damping structure, thus achieving a buffered landing for the rotor.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: The system identification module used in this invention obtains mass, stiffness, damping, residual imbalance and fault characteristics through closed-loop excitation and multi-source signal inversion. The state discrimination module then combines threshold, duration and multi-feature fusion to generate state identifiers, which can reduce misjudgments caused by single signal noise and provide a basis for the selection of control parameters and protection levels.

[0020] The fault-tolerant control module of this invention uses an extended state observer to estimate unmodeled dynamics and external disturbances, maintains stability under model perturbation through robust feedback correction, and compensates for unbalanced forces and missing local electromagnetic forces through dynamic feedforward compensation and multi-channel attitude correction, thus improving the levitation recovery capability after instability; the control mode adopts continuous weight smooth switching, which can reduce control current abrupt changes and the resulting secondary oscillations.

[0021] This invention is based on a graded fall protection module. First, pre-contact buffering reduces rotor speed; then, protective bearings and limiting structures restrict maximum displacement; and finally, a damping structure absorbs the remaining kinetic energy. The three-stage protection distributes energy progressively, reducing the peak load borne by a single-stage protective bearing and correspondingly decreasing contact temperature rise and wear.

[0022] This invention uses a collaborative decision-making and status monitoring module to uniformly arrange the activation sequence of fault-tolerant control and fall protection based on the fault level and recovery result, so that suspension recovery and mechanical protection are no longer independent of each other; when recovery is successful, it smoothly returns to normal control, and when recovery fails, it promptly enters fall protection, thereby improving the safety, stability and reliability of the system.

[0023] The permanent magnet protection bearing of this invention adopts a Halbach array, which can provide non-contact unloading force when the rotor deviates from the balance position but before mechanical contact occurs, effectively reducing the impact load of the grounding protection bearing; the setting of tap terminals allows the magnetic circuit to be reconstructed by switching in the event of a partial winding fault, without the need to add a whole set of redundant windings, thus taking into account both fault tolerance and structural compactness. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the main structure of the magnetic levitation bearing in Embodiment 1 of this application; Figure 2 This is a flowchart of the suspension recovery control method in Embodiment 2 of this application; Figure 3 This is a flowchart of the fall protection control method in Embodiment 1 of this application; Figure 4 This is a switching logic diagram of the dual-mode controller in Embodiment 1 of this application; Figure 5 This is a framework diagram of the μ-synthesis robust algorithm implementation in Embodiment 2 of this application; Figure 6 This is a structural diagram of the coil fault detection circuit in Embodiment 1 of this application; Figure 7 This is a schematic diagram of the stator winding structure with tapped terminals in Embodiment 1 of this application; Figure 8 This is a structural diagram of the three-level fall protection buffer mechanism in Embodiment 1 of this application; Figure 9 This is a flowchart of the GA-PSO multi-objective optimization algorithm in Embodiment 1 of this application.

[0025] The components are: 1. Iron core; 2. Winding; 3. Tap terminal; 4. Insulation layer; 5. Main shaft; 6. Halbach array; 7. TD bearing; 8. Elastic damping. Detailed Implementation

[0026] The technical solution of the present invention will be further described in a non-limiting manner below with reference to the accompanying drawings and specific embodiments.

[0027] like Figures 1 to 9 As shown, the accompanying drawings are only for illustrating the corresponding structures, signal relationships, or control flows, and do not represent the actual dimensions, scale, or unique arrangement of each component.

[0028] Example 1: This embodiment first provides a magnetic levitation bearing suspension recovery and fall protection system, which includes a magnetic levitation bearing main structure, a system identification module, a state discrimination module, a fault-tolerant control module, a graded fall protection module, and a collaborative decision-making and state monitoring module. The magnetic levitation bearing main structure includes a stator core, a stator winding, a rotor shaft, a rotor core, a radial magnetic bearing, an axial magnetic bearing, a displacement sensor, a speed sensor, a temperature sensor, a protective bearing, a limiting structure, and a base. The stator winding is disposed on the stator core, and the rotor core is fitted onto the outside of the rotor shaft. The radial magnetic bearing, axial magnetic bearing, protective bearing, and limiting structure are respectively installed on the base and arranged around the rotor shaft. The rotor core and stator core form a radial working air gap, and the rotor core and axial magnetic bearing form an axial working air gap. The displacement sensor, speed sensor, and temperature sensor are fixedly installed on the base, with their detection ends facing the rotor shaft, rotor core, and stator winding, respectively. like Figure 1 As shown, Figure 1 The mechanical and detection components of the main structure of the magnetic levitation bearing of this application are specifically illustrated. The radial and axial magnetic levitation bearings in the figure support the rotor shaft, the displacement sensor detects rotor displacement, the coil winding corresponds to the aforementioned stator winding, and the protective bearing or device limits rotor offset in the event of levitation failure. The motor in the figure is a rotary drive component connected to the rotor shaft. The system identification module, state judgment module, fault-tolerant control module, graded fall protection module, and collaborative decision-making and state monitoring module work in conjunction with this main structure through the aforementioned electrical or mechanical connections.

[0029] The system identification module, state judgment module, fault-tolerant control module, and collaborative decision-making and state monitoring module are implemented by the controller. It receives speed, displacement, current, and temperature signals from the main structure of the magnetic levitation bearing via a signal acquisition interface, and outputs control commands to the magnetic levitation bearing coil drive unit via a control output interface. The mechanical protection components in the graded fall protection module are installed within the main structure of the magnetic levitation bearing and form a graded protection relationship with the rotor. Therefore, it is not... Figure 1 The mechanical structure diagrams are shown one by one. The input signals, processing procedures, output results, and control relationships between the above modules are explained in subsequent control flowcharts and corresponding text in specific implementation methods.

[0030] Among them, the system identification module is used to generate mass matrix, stiffness matrix, damping matrix, residual unbalance vector and fault feature vector through closed-loop excitation, signal acquisition and parameter inversion; Among them, the status discrimination module is used to generate a suspension status identifier and protection trigger signal based on real-time speed, displacement, current and temperature signals through threshold judgment, feature fusion and fault classification algorithm; Among them, the fault-tolerant control module is used to generate real-time fault-tolerant control current and suspension restoring force based on extended state observation, robust feedback correction, and feedforward dynamic compensation algorithm to achieve rapid suspension recovery after instability; Among them, the graded fall protection module is used to realize three-level fall protection based on the pre-contact buffer algorithm, rigid limit constraint algorithm, and damping energy absorption algorithm; Among them, the collaborative decision-making and status monitoring module is used to realize the collaborative control of suspension recovery and fall protection based on multi-objective optimization algorithm and time-series collaborative decision-making algorithm; Among them, the system identification module, state discrimination module, fault-tolerant control module, graded fall protection module, and collaborative decision-making and state monitoring module are all electrically or mechanically connected to the main structure of the magnetic levitation bearing.

[0031] In one embodiment of the present invention, the main structure of the magnetic levitation bearing includes: a stator core with a silicon steel sheet laminated structure, winding slots on the toothed portion, and a stator winding embedded and fixed in the winding slots; the stator winding adopts a concentrated winding or a distributed winding, and the winding lead-out end is electrically connected to the fault-tolerant control module; the radial magnetic bearing has a two-pole or four-pole structure, symmetrically arranged on both radial sides of the rotor shaft to provide radial levitation force; the axial magnetic bearing has a single-end thrust or double-end thrust structure, arranged at the axial end of the rotor shaft to provide axial levitation force; the protective bearing is a rolling bearing or a sliding bearing structure, with the inner ring sleeved on the outside of the rotor shaft and the outer ring fixed to the base; the limiting structure includes a radial limiting block, an axial limiting block, and a buffer pad, with the limiting gap being smaller than the working air gap and larger than the protective bearing gap; the base is an integral or assembled structure, with sensor mounting holes, winding lead-out holes, and cooling channels.

[0032] In one embodiment of the invention, the protective bearing includes a permanent magnet protective bearing arranged axially and a ground contact protective bearing arranged radially. The permanent magnet protective bearing may employ a Halbach array to enhance the magnetic field on the working side, providing non-contact unloading force when the rotor deviates from the equilibrium position but before mechanical contact occurs; the ground contact protective bearing bears the remaining load after the displacement reaches the limit threshold and, together with the buffer pad, limits the maximum radial and axial displacement of the rotor.

[0033] like Figure 8 As shown in the figure, the main shaft 5 corresponds to the aforementioned rotor shaft; the Halbach array 6 constitutes the non-contact unloading part of the permanent magnet protective bearing, providing pre-contact buffering before the rotor contacts the protective bearing; the TD bearing 7, i.e., the aforementioned ground contact protective bearing, provides rigid restraint when the rotor offset reaches the contact threshold; the elastic damping structure 8, located around the ground contact protective bearing, absorbs energy through elastic deformation and damping energy dissipation after contact. Thus, Figure 8 The three structural parts shown correspond to three levels of protection: pre-contact buffer, rigid limit, and damping energy absorption.

[0034] In one embodiment of the present invention, tap terminals are provided at preset number of turns positions in the stator winding. When the state discrimination module identifies a local winding open circuit, short circuit, or overcurrent fault, the fault-tolerant control module isolates the faulty winding and adjusts the effective excitation turns and electromagnetic force distribution through tap switching, normal winding current redistribution, and feedforward compensation, so that the resultant force and torque on the rotor tend to rebalance.

[0035] like Figure 7 As shown, Figure 7The tap structure shown belongs to the stator winding in the main structure of the magnetic levitation bearing of this application. In the figure, core 1 is the stator core, winding 2 is the stator winding mounted on the stator core, and tap terminal 3 is led out from the preset number of turns position of the stator winding. An insulation layer 4 is provided between the stator winding and the stator core. Tap terminal 3 is electrically connected to the winding switching unit of the fault-tolerant control module through an insulated lead. During normal operation, the fault-tolerant control module supplies power to the stator winding according to the predetermined effective number of turns. When a partial winding open circuit, short circuit, or overcurrent is detected, the winding 2 switching unit isolates the faulty winding segment and selects the corresponding tap terminal 3, thereby changing the effective excitation turns of that channel. This, combined with the current redistribution and feedforward compensation of the normal winding, restores the required electromagnetic force. Figure 7 The "tap coil 3" or "winding 2" in the text both refer to the stator winding with tap terminals.

[0036] In one embodiment of the present invention, the system identification module includes: a closed-loop excitation unit, used to inject a sinusoidal excitation signal into the magnetic bearing coil and simultaneously acquire displacement, current, temperature and speed signals; Parameter inversion unit, used for electromechanical coupling dynamics model: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix of the magnetic levitation rotor system, respectively; x is the generalized displacement vector of the rotor; ẋ and ẍ are the first and second derivatives of x with respect to time t, respectively; and F(t) is the resultant force vector acting on the rotor. The resultant force includes at least electromagnetic force and external disturbance force, and may include contact force when contact occurs. The parameter inversion unit inverts the mass matrix m, stiffness matrix k, and damping matrix c through the least squares algorithm and the generalized eigenvalue decomposition algorithm.

[0037] In one embodiment of the present invention, the closed-loop excitation unit, while maintaining stable rotor levitation, superimposes a single-frequency or swept-frequency sinusoidal signal with limited amplitude onto the bias current or control current of the magnetic bearing channel to be identified, and synchronously acquires the coil current, rotor displacement, rotational speed, and winding temperature of the channel according to a unified clock. The parameter inversion unit performs noise reduction, time alignment, and discretization processing on the acquired signals, obtains the excitation force based on the mapping relationship between coil current and electromagnetic force, constructs a regression matrix from the electromechanical coupling dynamics equation, and uses the least squares algorithm to obtain the mass, damping, and stiffness parameters; then, it uses the modal frequencies and damping ratios obtained from generalized eigenvalue decomposition to verify the identification results. The temperature signal is used to correct for changes in coil resistance and electromagnetic force coefficient. The identification results are sent to the fault-tolerant control module and the collaborative decision-making and status monitoring module to update control parameters and protection trigger conditions.

[0038] The imbalance identification unit is used to generate residual imbalance vectors by inversely solving based on vibration residuals and displacement response. The vibration residual is the frequency-synchronous vibration component retained after subtracting the desired displacement response and the known control response from the measured displacement response. Based on the speed and phase references provided by the speed tracking unit, the unbalance identification unit extracts the amplitude and phase of the one-time-frequency component from each radial displacement signal, and performs inverse solving using the identified system frequency response model to obtain a residual unbalance vector characterizing the magnitude, direction, and initial phase of the unbalance. The dynamic feedforward compensation unit generates a compensation force with the same frequency and opposite phase to the unbalance excitation force based on the residual unbalance vector.

[0039] The fault feature extraction unit is used to generate fault feature vectors based on overcurrent, over-displacement, over-temperature, and speed fluctuation features through a feature fusion algorithm.

[0040] The fault feature vector includes at least the displacement amplitude and rate of change of each channel, the coil current amplitude and duration of overcurrent or current loss, the rotational speed change rate, the winding temperature and temperature rise rate, the fault duration, and the fault channel identifier. The fault feature extraction unit normalizes the above features and then fuses them according to a preset combination logic or fuzzy membership degree: when the current of a certain channel is lower than the open-circuit threshold and the displacement in the corresponding direction continues to increase, it is determined to be a winding open-circuit fault in that channel; when the current exceeds the overcurrent threshold and the temperature rise rate exceeds the temperature rise threshold, it is determined to be a winding overcurrent or short-circuit fault; when multiple displacement channels change abruptly simultaneously but the coil current does not show a corresponding abnormality, it is determined to be an external impact or load disturbance. The fusion result is used to determine the fault type, fault channel, and fault level.

[0041] like Figure 6 As shown, the coil fault detection circuit is connected to the current detection channel of the magnetic bearing coil, and includes a sampling resistor R, a signal filtering unit, a first comparator, a second comparator, a fault counting unit, and an output interface. The coil detection voltage obtained by the sampling resistor is filtered and compared with the first reference voltage VREF1 and the second reference voltage VREF2 to generate overcurrent fault signals and open-circuit fault signals. The fault counting unit outputs the fault result only when the corresponding fault signal continuously reaches the preset counting condition, so as to reduce misjudgment caused by transient interference. The output interface sends the fault result to the status discrimination module, and the collaborative decision-making and status monitoring module performs fault classification, control switching, or protection triggering accordingly. Figure 6 The term "cooperative control center" refers to the cooperative decision-making and status monitoring module in this manual.

[0042] In one embodiment of the present invention, the state discrimination module includes: The speed tracking unit is used to calculate the speed in real time based on the phase-locked loop algorithm and the Kalman filter algorithm; The status judgment unit is used to compare displacement, current, and rotation speed with preset thresholds and generate four states: normal, warning, fault, and fall through logical judgment. The fault classification unit is used to generate Level 1 warning, Level 2 fault tolerance, Level 3 protection, and Level 4 fall protection levels based on fault duration, signal out-of-tolerance amplitude, and out-of-tolerance frequency using a fuzzy classification algorithm. The protection triggering unit is used to trigger corresponding protection actions based on status indicators and fault levels through timing logic.

[0043] The status discrimination module uses a unified clock to synchronously sample displacement, current, speed, and temperature signals, and performs filtering and feature extraction on each signal. Threshold judgment and fault classification adopt a time window and continuous counting mechanism, changing the status flag only when the out-of-tolerance amplitude, duration, or frequency meets preset conditions, thereby reducing instantaneous noise and false triggering caused by single sensor anomalies; when multiple features simultaneously meet the combinational logic, the corresponding fault-tolerant control or fall protection trigger signal is output.

[0044] The normal state corresponds to all monitored quantities being within the preset safety range, and the system maintaining normal control. Level 1 warning corresponds to minor or short-term deviations in monitored quantities; the system maintains normal control and increases the monitoring frequency. Level 2 fault tolerance corresponds to confirmed local winding faults, abnormal parameters, or recoverable instability, triggering fault-tolerant control. Level 3 protection corresponds to continued increase in rotor displacement or insufficient recovery margin; the graded fall protection module enters a preparatory state and activates pre-contact buffering. Level 4 fall protection corresponds to recovery timeout, displacement reaching the protection bearing contact threshold, or continued deterioration of the state, activating rigid limit switches and damping energy absorption. State escalation and degraded states employ hysteresis criteria with different thresholds and a continuous counting mechanism to avoid repeated state switching near the threshold. Furthermore, the fault-tolerant control module includes: Extended state observation unit, used for real-time observation of disturbances and uncertainties based on the extended state observer algorithm: Where e is the displacement tracking error, xd is the desired rotor displacement, and x is the actual displacement measured by the displacement sensor; x1 and x2 are the first and second estimated states of the extended state observer, respectively, used to characterize the rotor displacement and the generalized velocity state including unmodeled dynamic effects; l1 and l2 are the extended state observer gains; f(x,ẋ,t) is the total disturbance function composed of system nonlinearity, parameter perturbation, and external disturbance; u is the control input after input gain normalization, t is time, and the "·" above the variable indicates the first derivative with respect to time.

[0045] A robust feedback control unit is used to generate fault-tolerant control current based on sliding mode robust control algorithm and feedback correction algorithm; The dynamic feedforward compensation unit is used to generate feedforward compensation force based on the unbalance vector and disturbance observations to suppress vibration and eccentric impact. The attitude correction unit is used to quickly pull the rotor back to the center equilibrium position based on a three-channel cooperative control algorithm.

[0046] The fault-tolerant control module can be configured with different control parameters or control laws for normal suspension and fault recovery states. After the state discrimination module outputs the fault state, the collaborative decision-making and state monitoring module smoothly transitions the control output from normal control to recovery control according to a preset timing sequence; the extended state observation unit estimates parameter uncertainties and external disturbances, the robust feedback control unit suppresses large displacements and model perturbations, the dynamic feedforward compensation unit compensates for residual imbalances and identified disturbances, and the attitude correction unit coordinates the radial and axial control channels to gradually return the rotor to the center equilibrium position.

[0047] like Figure 4 As shown, the performance controller corresponds to the normal performance controller in the normal suspension state, and the collaborative control center corresponds to the collaborative decision-making and status monitoring module. When the fault or threshold triggering conditions are not met, the system maintains normal performance control; when the switching conditions are met, the controller switching process is initiated, and control is switched to the recovery controller using an exponential smooth transition, while the graded fall protection module enters a ready or semi-active state. The recovery controller continuously adjusts the coil current and drives the rotor to restore stability; when the rollback conditions are met, it returns to normal performance control through an exponential smooth transition; if the rollback conditions are not met, recovery control and protection preparation continue.

[0048] When a fault is detected in a winding, the fault-tolerant control module determines the affected channel based on the fault feature vector provided by the system identification module, limits or cuts off the drive current of that channel, and increases the compensation share of the remaining normal channels. If tap terminals are provided, tap switching is performed synchronously to restore usable electromagnetic force without increasing the entire set of redundant windings.

[0049] In one embodiment of the present invention, the graded fall protection module includes: Pre-contact buffer unit is used to reduce falling speed and impact energy based on elastic damping buffer algorithm and velocity suppression algorithm; Rigid limit constraint unit is used to limit the maximum displacement based on displacement constraint algorithm and radial limit algorithm to prevent rifling and collision with the structure. Damping energy absorption unit is used to convert falling kinetic energy into thermal energy and deformation energy based on viscoelastic damping energy absorption algorithm and kinetic energy conversion algorithm, thereby reducing impact load.

[0050] In one embodiment of the present invention, the working principle of the graded fall protection module includes: When suspension fails, the pre-contact buffer algorithm is first activated to suppress the falling acceleration; when the preset displacement threshold is reached, the rigid limit constraint algorithm is activated to limit the radial and axial displacement; the damping energy absorption algorithm is activated throughout the process to absorb the falling kinetic energy; during the protection process, the impact force, displacement and temperature are collected in real time, and the protection control parameters are updated through an adaptive algorithm.

[0051] The three-stage protection is activated sequentially according to the rotor displacement and motion state: the pre-contact buffer stage uses permanent magnet unloading force or controlled electromagnetic force to reduce the falling acceleration; the rigid limiting stage uses the protective bearing and limiting structure to limit the maximum offset and avoid rotor rubbing against the stator; the damping energy absorption stage uses the deformation and damping of buffer pads, elastic elements or viscoelastic materials to convert the remaining kinetic energy into deformation energy and thermal energy. Because the previous stage reduces the speed and load, the peak impact experienced by the subsequent stage is reduced accordingly, thereby reducing the temperature rise and cumulative wear of the protective bearing.

[0052] like Figure 3 As shown, when the state discrimination module outputs a fall protection trigger signal, the collaborative decision-making and state monitoring module issues a collaborative control command. The graded fall protection module sequentially enables the Halbach array to provide primary non-contact unloading, enables the ground contact protection bearing (TD bearing) to bear the remaining load and perform secondary rigid limiting, and enables the elastic damping structure to perform tertiary energy absorption until the rotor lands smoothly. After confirming landing, the system cuts off or limits the power supply to the magnetic levitation bearing coil, records the fault location, and outputs an audible and visual alarm. Only after manual inspection and fault troubleshooting are completed can the system be reset and re-enter the levitation process.

[0053] In one embodiment of the present invention, the collaborative decision-making and status monitoring module includes: The multi-objective optimization unit is used to construct a multi-objective optimization function with the objectives of minimizing recovery time, impact, temperature rise, and wear. Wherein, J is the comprehensive performance index of the suspension recovery and fall protection process; tr is the recovery time from fault triggering to rotor displacement recovery and sustained stability within the safety threshold; Fs is the peak impact force during the fall or contact process; ΔT is the maximum temperature rise during the protection process; W is the wear amount or equivalent wear index of the protective bearing and buffer structure; λ1, λ2, λ3, and λ4 are the non-negative weighting coefficients of the corresponding targets, and λ1+λ2+λ3+λ4=1. To eliminate the influence caused by the different dimensions and orders of magnitude of each target, tr, Fs, ΔT, and W are normalized before weighted summation.

[0054] like Figure 9As shown, the multi-objective optimization unit can use the GA-PSO algorithm, which combines genetic algorithm and particle swarm optimization, to solve for the control parameters of suspension recovery and fall protection. The individuals or particles to be optimized are used to encode the recovery time threshold, displacement trigger threshold, controller parameters, buffer damping parameters, or target weights. The fitness is calculated from the normalized tr, Fs, ΔT, and W, with system stability, maximum control current, allowable displacement, and allowable temperature rise as constraints. Each iteration performs genetic selection, crossover, and mutation, as well as particle velocity and position updates, forming a new generation of candidate solutions. After reaching the termination condition, the Pareto optimal solution set is output, and the collaborative decision-making unit selects the control parameters based on the current fault level and preset weights.

[0055] The collaborative decision-making unit is used to implement coordinated actions for suspension recovery and fall protection based on time-series collaborative algorithms and logical judgment algorithms. The data recording unit is used to store fault information, protection action sequence, impact force, displacement, and temperature data in real time. The human-machine interface unit is used to display status, fault information, and protection actions, and supports manual reset and manual emergency stop.

[0056] The collaborative decision-making and status monitoring module uses the suspension status indicator, fault level, recovery time, peak displacement, and temperature as decision inputs: Under recoverable faults, fault-tolerant control is prioritized, and the graded fall protection module is kept in standby mode; when the rotor displacement returns to the safe threshold within the preset recovery time and remains stable, normal control is smoothly restored; when the recovery timeout occurs or the displacement continues to increase, graded fall protection is immediately triggered. The data recording unit synchronously saves the triggering cause, control switching time, and protection response data for subsequent reset, fault tracing, and parameter updates.

[0057] In one embodiment of the present invention, the system achieves rapid levitation recovery and fall protection after the magnetic levitation bearing becomes unstable through the coordinated work of the main structure of the magnetic levitation bearing, system identification, state discrimination, fault-tolerant control, graded fall protection, and collaborative decision-making, thereby improving the safety, stability and reliability of the system.

[0058] Example 2: This embodiment provides a method for implementing the system, the method specifically including the following steps: like Figure 2As shown, the levitation recovery control method first identifies the fault type after startup. For external disturbances or power-off recovery faults, the collaborative decision-making and status monitoring module switches the normal performance controller to the recovery controller, and the magnetic levitation bearing outputs the recovery electromagnetic force to pull the rotor back to a safe position. For faults within a local winding, the failed winding and its location are first located, the PWM drive of the failed channel is limited or cut off, and if necessary, the stator winding tap terminals are switched and the current of adjacent normal windings is redistributed. At the same time, the Halbach array of the permanent magnet protection bearing provides unloading force, and then the recovery controller implements electromagnetic force reconstruction. The system determines whether the recovery is successful based on the rotor displacement amplitude and the duration of stability. If the stability criterion is met, it returns to normal performance control through an exponential smooth transition; if the criterion is not met and the deflection continues to increase, it switches to fall protection. Figure 2 The 100μm, 50μm and 2s shown are example criteria for one embodiment. The actual thresholds are calibrated based on the working air gap, the protective bearing gap, the rotational speed and the load.

[0059] Step 1: First, during the normal operation of the system, multi-source state data acquisition and processing are performed on the magnetic levitation bearing system. Specifically, eddy current displacement sensors arranged at multiple radial measuring points on the rotor are used to collect the radial displacement response data of the rotor in real time. This displacement response data can reflect the dynamic deflection state of the rotor under different operating conditions. Excitation current data of each magnetic bearing coil is collected by a current detection unit located at the output of the power amplifier. This current data is used to characterize the magnitude of the electromagnetic force. Simultaneously, a speed encoder acquires the pulse signals during rotor rotation, and the instantaneous rotational speed of the rotor is calculated by pulse counting and time measurement. These multi-source signals are synchronously sampled using a unified clock, and after low-pass filtering and anti-interference processing, a system state vector is constructed, providing a data foundation for subsequent fault detection and control decisions.

[0060] Step Two: Fault detection and status identification processing are performed based on the displacement, current, and speed signals. Specifically, the displacement signal is compared with a preset safety threshold. When the displacement amplitude exceeds the threshold and the duration exceeds a set time window, abnormal rotor deflection is determined. Simultaneously, the current signal is threshold-judged. When the detected current exceeds the upper threshold or falls below the lower threshold, it is determined to be an overcurrent fault or a coil open-circuit fault, respectively. In addition, the presence of sudden external disturbances in the system can be determined by calculating the rate of change of speed. When the above multiple criteria satisfy the combined logic conditions, a levitation failure is comprehensively determined.

[0061] Step 3: After determining levitation failure, the system enters the levitation recovery phase. Specifically, the collaborative decision-making and status monitoring module outputs a control switching command to switch the original performance controller to a recovery controller. The recovery controller is designed based on the dynamic model of the magnetic levitation rotor system, and its system model can be represented as: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix of the magnetic levitation rotor system, respectively; x is the generalized displacement vector of the rotor; ẋ and ẍ are the first and second derivatives of x with respect to time t, respectively; F(t) is the resultant force vector acting on the rotor, which includes the electromagnetic force Fe(t) and the external disturbance force Fd(t) in the non-contact state, and also includes the contact force Fc(t) when the rotor is in contact with the protective bearing, i.e., F(t) = Fe(t) + Fd(t) + Fc(t). After converting this dynamic equation into state-space form, a robust feedback controller and a dynamic feedforward compensator can be constructed based on it.

[0062] Step 4: In this embodiment, the robust feedback control unit can select sliding mode robust control or μ-synthetic robust control based on the representation of parameter uncertainty and fault conditions. These two methods are alternatives and not necessarily required to be used simultaneously. When the upper bound of the disturbance can be determined and large displacements need to be suppressed quickly, a sliding surface can be constructed based on the displacement error and its rate of change, and a feedback correction quantity can be generated by the sliding mode robust control law. When it is necessary to simultaneously handle structured parameter uncertainty and frequency domain performance constraints, μ-synthetic robust control can be used. Both methods receive the disturbance estimate output by the extended state observation unit and convert it into a coil control current after superimposing it with the dynamic feedforward compensation quantity. The following specific implementation process uses μ-synthetic robust control as an example; sliding mode robust control can be implemented according to the same signal input, compensation superposition, and current output relationship.

[0063] In practical implementation, the recovery controller preferably employs a μ-synthetic robust control method to control the system. This method involves structured modeling of parameter uncertainties in the system, considering factors such as rotor stiffness variations, electromagnetic force coefficient deviations, and speed disturbances. The optimal controller parameters are then solved using the DK iterative algorithm, ensuring system stability even under uncertainties. The recovery controller outputs control current commands to gradually restore the magnetic bearing's levitation support capability for the rotor.

[0064] like Figure 5As shown, the μ-synthetic robust algorithm first inputs the basic system information and establishes a state-space model, then structures the uncertainties into measurement uncertainties, dynamic uncertainties, and parameter uncertainties. Measurement uncertainties include noise and signal transmission interference from displacement and speed sensors; dynamic uncertainties include contact force disturbances in the grounding protection bearing and sudden load changes; parameter uncertainties include rotor natural frequency fluctuations, magnetic bearing stiffness changes, and air gap deviations. Subsequently, a performance weight matrix is ​​designed and a weighted closed-loop system is constructed. The scale matrix is ​​solved through a D-step, and the structural singularity μ is calculated. When μ does not meet the preset conditions, the controller parameters are optimized through a K-step and iterated repeatedly until μ meets the robust stability and robust performance requirements, generating a dual-mode controller for normal performance control and fault recovery control.

[0065] During the recovery control execution process, the graded fall protection module is simultaneously put into a preparatory or buffer state to achieve synergistic effects of control recovery and structural protection. Specifically, the permanent magnet protection bearing in the axial direction adopts a Halbach array structure. This structure, through a special magnetization direction arrangement, enhances the magnetic field on one side, thereby providing non-contact magnetic support when the rotor deviates. Simultaneously, the ground contact protection bearing (TD bearing) in the radial direction is used to limit the maximum rotor deviance. When the rotor deviates beyond a certain range, the TD bearing contacts the rotor, thereby preventing further rotor deviation. To reduce contact impact force, an elastic damping element can also be incorporated into the TD bearing structure to absorb impact energy through elastic deformation. Its contact force can be expressed as: Where Fc is the normal contact force between the rotor and the protective bearing or limiting structure, kc is the equivalent contact stiffness, δ is the normal contact deformation, n is the contact nonlinearity index, c is the equivalent contact damping coefficient, and δ̇ is the first derivative of the contact deformation with respect to time; when the Hertz contact model is used, n can be taken as 3 / 2.

[0066] Step 5: As the recovery control continues, the rotor gradually returns to its equilibrium position. Once the rotor displacement enters the safe threshold range and remains stable for a period of time, the system executes a controller cut-back operation. During the cut-back process, to avoid system oscillations caused by sudden changes in control current, an exponential smoothing function can be used to achieve a continuous transition of the control signal. Its expression is as follows: Where u(t) is the smooth control output actually applied to the power amplifier, ur(t) is the recovery controller output, up(t) is the normal performance controller output, α(t) is the continuous switching coefficient ranging from 0 to 1, and t is time; when switching from normal control to recovery control, α(t) changes continuously from 0 to 1, and when recovery is completed and switching back, α(t) changes continuously from 1 to 0.

[0067] Step Six: If the system fails to complete the suspension recovery within the preset time, the recovery is deemed a failure, and the system enters the fall protection phase. In this phase, the permanent magnet protection bearing first provides the main axial support force to reduce the rotor's falling speed; subsequently, the ground contact protection bearing (TD bearing) bears the remaining load and absorbs the impact energy through the elastic damping structure, thereby achieving a buffered landing of the rotor and significantly reducing the impact force and wear.

[0068] In one embodiment, to further improve the stability of the suspension recovery process, a vibration suppression mechanism can be introduced during the recovery control process. Specifically, the instantaneous rotational speed can be obtained by frequency tracking through a phase-locked loop (PLL) based on the real-time rotational speed of the rotor. Based on the relationship between this rotational speed and the system modal frequency, the notch filter parameters can be adaptively configured so that the center frequency of the notch filter always tracks the dominant modal frequency, thereby achieving real-time suppression of resonant vibration.

[0069] Furthermore, in one embodiment, the system mass matrix and stiffness matrix can be obtained through parameter identification methods based on the system dynamics model and sensor data. Then, the system modal characteristics are analyzed through eigenvalue decomposition to determine the critical speeds and modal damping ratios for each order. Based on this, the controller parameters are dynamically adjusted using a gain scheduling strategy to give the system higher damping characteristics near the critical speed, thereby improving the system's stability in the critical region.

[0070] In one embodiment, to further improve the overall system performance, the rotor counterweight can be optimized using a multi-objective optimization method based on the rotor residual imbalance distribution. Specifically, a primary objective function can be constructed with minimizing vibration energy across the entire speed range, and a secondary objective function can be constructed with minimizing counterweight mass. Combined with modal stability constraints, the optimal counterweight scheme can be solved using a genetic algorithm or the aforementioned GA-PSO hybrid algorithm, thereby achieving a synergistic improvement in system vibration performance and structural optimization.

[0071] The above description is an explanation of the present invention and not a limitation thereof. The scope of the present invention is defined by the claims. Within the scope of protection of the present invention, any form of modification may be made.

Claims

1. A magnetic levitation bearing suspension recovery and fall protection system, characterized in that, include: The main structure of the magnetic levitation bearing includes a stator core, stator windings, rotor shaft, rotor core, radial magnetic bearing, axial magnetic bearing, protective bearing, limiting structure, base, displacement sensor, speed sensor and temperature sensor; The stator winding is disposed on the stator core, the rotor core is fitted on the outside of the rotor shaft, and the radial magnetic bearing, axial magnetic bearing, protective bearing and limiting structure are respectively installed on the base and arranged around the rotor shaft; the rotor core and the stator core form a radial working air gap, and the rotor core and the axial magnetic bearing form an axial working air gap. The displacement sensor, speed sensor, and temperature sensor are fixedly installed on the base, with their detection ends facing the rotor shaft, rotor core, and stator winding, respectively. The system identification module is configured to generate the mass matrix, stiffness matrix, damping matrix, residual unbalance vector, and fault feature vector. The status determination module is configured to generate a floating status indicator and a protection trigger signal; The fault-tolerant control module is configured to generate real-time fault-tolerant control current and levitation restoring force to achieve rapid levitation recovery after instability. A graded fall protection module, which is configured to provide three levels of fall protection; Collaborative decision-making and status monitoring module; it is configured to achieve collaborative control of suspension recovery and fall protection; The system identification module, state discrimination module, fault-tolerant control module, graded fall protection module, and collaborative decision-making and state monitoring module are all electrically or mechanically connected to the main structure of the magnetic levitation bearing.

2. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The protective bearing includes a permanent magnet protective bearing arranged along the axial direction and a ground contact protective bearing arranged along the radial direction; the permanent magnet protective bearing adopts a Halbach array to provide non-contact unloading force when the rotor deviates from the balance position before mechanical contact occurs; the ground contact protective bearing is used to bear the remaining load after the displacement reaches the limit threshold, and together with the buffer pad, it limits the maximum radial and axial displacement of the rotor.

3. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The stator winding is provided with tap terminals at a preset number of turns position; when the state discrimination module identifies a local winding open circuit, short circuit or overcurrent fault, the fault-tolerant control module isolates the faulty winding and adjusts the effective excitation turns and electromagnetic force distribution through tap switching, normal winding current redistribution and feedforward compensation, so that the resultant force and torque on the rotor tend to rebalance.

4. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The closed-loop excitation unit is used to inject sinusoidal excitation signals into the magnetic bearing coil and simultaneously acquire displacement, current, temperature, and speed signals. Parameter inversion unit, used for electromechanical coupled dynamics model: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix of the magnetic levitation rotor system, respectively; x is the generalized displacement vector of the rotor; ẋ and ẍ are the first and second derivatives of x with respect to time t, respectively; and F(t) is the resultant force vector acting on the rotor. The resultant force includes at least electromagnetic force and external disturbance force, and may include contact force when contact occurs. The parameter inversion unit inverts the mass matrix m, stiffness matrix k, and damping matrix c using the least squares algorithm and the generalized eigenvalue decomposition algorithm. The imbalance identification unit is used to generate residual imbalance vectors by inversely solving based on vibration residuals and displacement response. The fault feature extraction unit is used to generate fault feature vectors based on overcurrent, over-displacement, over-temperature, and speed fluctuation features through a feature fusion algorithm.

5. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The state determination module includes: The speed tracking unit is used to calculate the speed in real time based on a phase-locked loop algorithm or a Kalman filter algorithm. The status judgment unit is used to compare displacement, current, and rotation speed with preset thresholds and generate four states: normal, warning, fault, and fall through logical judgment. The fault classification unit is used to generate Level 1 warning, Level 2 fault tolerance, Level 3 protection, and Level 4 fall protection levels based on fault duration, signal out-of-tolerance amplitude, and out-of-tolerance frequency using a fuzzy classification algorithm. The protection triggering unit is used to trigger corresponding protection actions based on status indicators and fault levels through timing logic.

6. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The fault-tolerant control module includes: Extended state observation unit, used for real-time observation of disturbances and uncertainties based on the extended state observer algorithm: Where e is the displacement tracking error, xd is the desired rotor displacement, and x is the actual displacement measured by the displacement sensor; x1 and x2 are the first and second estimated states of the extended state observer, respectively, used to characterize the rotor displacement and the generalized velocity state including unmodeled dynamic effects; l1 and l2 are the extended state observer gains; f(x,ẋ,t) is the total disturbance function composed of system nonlinearity, parameter perturbation, and external disturbance; u is the control input after input gain normalization, t is time, and the "·" above the variable indicates the first derivative with respect to time; A robust feedback control unit is used to generate fault-tolerant control current based on sliding mode robust control algorithm and feedback correction algorithm; The dynamic feedforward compensation unit is used to generate feedforward compensation force based on the unbalance vector and disturbance observations to suppress vibration and eccentric impact. The attitude correction unit is used to quickly pull the rotor back to the center equilibrium position based on a three-channel cooperative control algorithm.

7. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The graded fall protection module includes: Pre-contact buffer unit, used to reduce falling speed and impact energy based on elastic damping buffer algorithm and velocity suppression algorithm; Rigid limiting constraint unit, used to limit maximum displacement based on displacement constraint algorithm and radial limiting algorithm; Damping energy absorption unit, used to convert falling kinetic energy into thermal energy and deformation energy based on viscoelastic damping energy absorption algorithm and kinetic energy conversion algorithm.

8. The magnetic levitation bearing suspension recovery and fall protection system as described in claim 1, characterized in that: The collaborative decision-making and status monitoring module includes a multi-objective optimization unit, which is configured to construct the following multi-objective optimization function: Wherein, J is the comprehensive performance index of the suspension recovery and fall protection process; tr is the recovery time from fault triggering to rotor displacement recovery and sustained stability within the safety threshold; Fs is the peak impact force during the fall or contact process; ΔT is the maximum temperature rise during the protection process; W is the wear amount or equivalent wear index of the protective bearing and buffer structure; λ1, λ2, λ3, and λ4 are the non-negative weighting coefficients of the corresponding targets, and λ1+λ2+λ3+λ4=1. To eliminate the influence caused by the different dimensions and orders of magnitude of each target, tr, Fs, ΔT, and W are normalized before weighted summation. The collaborative decision-making unit is configured to implement coordinated actions for suspension recovery and fall protection based on time-series collaborative algorithms and logical judgment algorithms; The data recording unit is configured to store fault information, protection action sequence, impact force, displacement, and temperature data in real time. The human-machine interface unit is configured to display status, fault information, and protection actions, and supports manual reset and manual emergency stop.

9. A method for implementing the magnetic levitation bearing suspension recovery and fall protection system according to any one of claims 1 to 8, characterized in that, Includes the following steps: Step 1: During the normal operation of the system, the radial displacement signal of the rotor, the excitation current signal of the magnetic bearing coil, and the instantaneous speed signal of the rotor are collected by the displacement sensor, the current detection unit, and the speed encoder, respectively. The displacement signal, current signal, and speed signal are synchronously sampled through a unified clock and then filtered and anti-interference processed to construct the system state vector. Step 2: Fault detection and status identification are performed based on the displacement signal, current signal and speed signal: The displacement signal is compared with a preset safety threshold. When the displacement amplitude exceeds the threshold and the duration exceeds the set time window, it is determined that the rotor has abnormally deflected. The current signal is threshold-based. When the detected current exceeds the upper threshold or falls below the lower threshold, it is determined to be an overcurrent fault or a coil open circuit fault, respectively. When multiple criteria meet the combinational logic conditions, the system is comprehensively determined to have a floating failure. Step 3: After determining that the levitation has failed, the collaborative decision-making and status monitoring module outputs a control switching command to switch the normal performance controller to a recovery controller. The recovery controller is designed based on the dynamic model of the magnetic levitation rotor system, and its system model is as follows: Where m, c, and k are the mass matrix, damping matrix, and stiffness matrix, respectively, x is the generalized displacement vector of the rotor, and F(t) is the resultant force vector acting on the rotor. Step four: The recovery controller outputs a control current command to gradually restore the magnetic bearing's ability to levitate and support the rotor; during the recovery control execution process, the graded fall protection module is simultaneously put into a preparatory state or a buffer state to achieve the synergistic effect of control recovery and structural protection. Step 5: Once the rotor displacement is detected to have entered the safe threshold range and remained stable for a preset time, a controller cut-back operation is executed. During the cut-back process, an exponential smoothing function is used to achieve a continuous transition of the control signal, expressed as follows: Where u(t) is the smooth control output actually applied to the power amplifier, ur(t) is the recovery controller output, up(t) is the normal performance controller output, α(t) is the continuous switching coefficient ranging from 0 to 1, and t is time; when switching from normal control to recovery control, α(t) changes continuously from 0 to 1, and when recovery is completed and the switchback is completed, α(t) changes continuously from 1 to 0; Step 6: If the system fails to complete the suspension recovery within the preset time, the recovery is deemed to have failed, and the system enters the fall protection phase. The graded fall protection module sequentially performs three levels of protection actions: pre-contact buffer, rigid limit, and damping energy absorption.

10. The method as described in claim 9, characterized in that: In step four, the recovery controller uses a μ-integrated robust control method to control the system. It performs structured modeling of the parameter uncertainties in the system and solves for the optimal controller parameters using the DK iterative algorithm. In step six, the fall protection stage, the permanent magnet protection bearing in the axial direction first provides axial support to reduce the rotor's falling speed. Then, the ground contact protection bearing in the radial direction bears the remaining load and absorbs the impact energy through an elastic damping structure, achieving a buffered landing for the rotor.

Images