A method for preventing and controlling magnetic pollution sources in a magnetic levitation compressor assembly process

Abstract

The present application relates to the technical field of magnetic suspension compressor assembly, and specifically discloses a magnetic pollution source head prevention and control optimization method in the assembly process of a magnetic suspension compressor, comprising: in the pre-treatment stage of the magnetic suspension compressor assembly, using a self-adaptive genetic algorithm to dynamically optimize the thermal cycle parameters in a low-temperature environment, accurately inducing the ferromagnetic impurities hidden in the micro-defects of the compressor components to release in advance; combining wavelet packet transform and neural network algorithm to perform feature extraction and pattern recognition on the released impurity signals; the present application constructs a low-temperature thermal cycle process optimized by a self-adaptive genetic algorithm in the pre-treatment stage of assembly, accurately induces the hidden ferromagnetic impurities in micro-cracks to release in advance, and combines wavelet packet transform and neural network driven targeted pulse demagnetization to accurately remove the impurities before entering the assembly link, breaks through the limitation that traditional cleaning and demagnetization methods cannot reach the impurities in micro-defects, and eliminates the magnetic pollution hidden danger that cannot be found by initial detection from the source.

Description

Technical Field

[0001] This invention relates to the field of magnetic levitation compressor assembly technology, and in particular to an optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor. Background Technology

[0002] The prevention and control of magnetic contamination during the assembly process of magnetic levitation compressors is directly related to the operational stability and control accuracy of magnetic levitation bearing systems. As magnetic levitation technology develops towards higher speeds and higher efficiency, the sensitivity of core components to residual magnetism has significantly increased. Even slight magnetic contamination can lead to magnetic field distortion, levitation failure, or even equipment damage. However, traditional assembly environments are difficult to effectively isolate ferromagnetic particles, residual magnetic fields from tools, and external electromagnetic interference. The potential for magnetic contamination persists through multiple stages, including material flow, tooling contact, and environmental disturbances.

[0003] In existing technologies, although the tooling and core components of magnetic levitation compressors undergo demagnetization and routine cleaning before assembly, under low-temperature assembly conditions, the shrinkage of the material's microstructure causes ferromagnetic impurities hidden in surface microcracks to be squeezed out due to thermal stress, forming localized magnetic contamination sources that cannot be detected in initial testing. Furthermore, the instantaneous weak magnetic field generated by the dynamic friction between non-magnetic tooling and fasteners during assembly induces the release of trace impurities, which are then directionally adsorbed and deposited on the core air gap or bearing working surface. This transient magnetic contamination process is difficult to capture and locate in real time using traditional environmental monitoring methods. Moreover, in the early stages of service, after assembly and testing, the magnetic particles deposited in critical areas undergo slow magnetic creep displacement due to the release of residual stress and the long-term influence of the weak external geomagnetic environment. These particles gradually migrate to sensitive locations such as magnetic levitation bearing sensors, leading to secondary magnetic contamination failures due to magnetic field distribution drift after several months of operation. Therefore, this paper proposes an optimized method for preventing and controlling magnetic contamination sources during the assembly process of magnetic levitation compressors to address the aforementioned problems. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention provides an optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor, which can effectively solve the problems involved in the prior art.

[0005] The objective of this invention can be achieved through the following technical solution: This invention provides an optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor, comprising the following steps:

[0006] S1. In the pre-assembly treatment stage of the magnetic levitation compressor, an adaptive genetic algorithm is used to dynamically optimize the thermal cycle parameters under low temperature conditions. By precisely controlling the amplitude and frequency of temperature alternation, the ferromagnetic impurities hidden in the micro-defects of the compressor components are precisely induced to be released in advance, thereby eliminating the hidden magnetic pollution source and effectively eliminating the hidden impurities in the micro-defects, blocking the initial source of magnetic pollution.

[0007] S2. Combining wavelet packet transform and neural network algorithm, feature extraction and pattern recognition are performed on the released impurity signal to guide the non-magnetic shielding system to perform targeted pulse demagnetization, complete the precise removal of ferromagnetic impurities before the magnetic levitation compressor is assembled, realize the precise positioning of impurities and the directional attenuation of magnetic moment, and improve the removal efficiency.

[0008] S3. Construct a multi-sensor data fusion model based on Kalman filtering to separate the instantaneous magnetic field interference generated by friction of non-magnetic tooling from the background magnetic field in real time, accurately restore the migration trajectory of impurities in the assembly of the magnetic levitation compressor, suppress dynamic friction interference, restore the real migration path, and provide a basis for blocking.

[0009] S4. Introducing a dynamic time warping algorithm, the real-time monitored magnetic field anomaly waveform is matched with a preset impurity release feature library to locate transient adsorption-deposition events and automatically block them, curbing the chain spread of magnetic pollution, identifying deposition events in milliseconds, and automatically blocking them to prevent pollution spread.

[0010] S5. Finite element simulation is used to pre-simulate the magnetic creep path of magnetic particles under the coupling of residual stress and geomagnetism, and Monte Carlo sampling method is combined to evaluate the particle displacement probability, identify secondary magnetic contamination risk points, quantitatively assess long-term migration risks, and accurately locate potential fault areas.

[0011] S6. Apply a controllable temperature-magnetic composite field to promote the early anchoring of deposited particles, stabilize the magnetic pollution source after the magnetic levitation compressor is assembled, prevent secondary failures caused by drift in the early stage of service, and integrate environmental monitoring and assembly operation data to build a closed-loop feedback mechanism to dynamically correct assembly parameters and ensure continuous optimization of magnetic pollution source control during the assembly process of the magnetic levitation compressor.

[0012] Preferably, S1 specifically includes:

[0013] A thermal cycling process parameter model was constructed with temperature alternation amplitude, alternation frequency, and low temperature holding time as decision variables. The dual objective optimization function was set as maximizing the impurity release rate and minimizing the component remanent magnetization increment. A penalty function was introduced to handle the constraint conditions of temperature change rate on the microstructure of the magnetic levitation compressor component, effectively avoiding thermal stress damage and ensuring the integrity of the component microstructure.

[0014] The thermal cycling parameter population is initialized using real number encoding. By adaptively adjusting the crossover and mutation probabilities, the population diversity is dynamically maintained during the iteration process. Combined with Pareto non-dominated sorting, the optimal parameter combination for the release of hidden ferromagnetic impurities in microcracks induced by low temperature is selected, the optimal process window is accurately locked, and the impurity release efficiency and consistency are improved.

[0015] The optimized thermal cycling parameters are input into the low-temperature constant temperature system of the three-condition assembly equipment, and thermal cycling treatment with alternating amplitude of ±3℃ to ±5℃ and frequency of 2 to 4 times per hour is performed. This precisely induces nanoscale ferromagnetic impurities in micro-defects to be extruded during the material shrinkage-expansion process, thereby eliminating the hidden magnetic pollution source at the source and reducing the risk of magnetic pollution in subsequent assembly.

[0016] Preferably, S1 further includes:

[0017] During the thermal cycling process, the surface temperature field distribution data and local remanent magnetization change data of the component are collected in real time, and a temperature-remanent magnetization dynamic response surface is constructed. This surface is fed back into the fitness function of the adaptive genetic algorithm to realize the online correction of thermal cycling parameters in the pre-assembly processing stage. Through real-time feedback, the thermal cycling parameters are dynamically adapted, which significantly improves the impurity release efficiency and process consistency.

[0018] For the three core components of the compressor—rotor, stator, and bearing—sub-models of material thermophysical parameters are established. Component-specific weighting coefficients are introduced into the genetic algorithm to generate component-specific thermal cycling process curves. This avoids over- or under-processing of some components under uniform parameters and applies thermal cycling based on material characteristics to ensure balanced performance of each component.

[0019] After the thermal cycle is completed, the residual magnetism of the component surface is scanned by a high-precision magnetic field imager to identify micro-area residual magnetism anomalies left after the release of impurities. The scan results are used as input for the optimization of thermal cycle parameters for the next batch of assembly, forming an iterative optimization closed loop between batches, and continuously enhancing the release effect of hidden magnetic pollution sources.

[0020] Preferably, S2 specifically includes:

[0021] A multi-channel high-bandwidth magnetic field sensor array is used to collect transient magnetic field fluctuation signals generated by impurity release during thermal cycling. The signal is decomposed into three levels by wavelet packet transform, and the energy entropy, singular values ​​and peak frequencies of each frequency band are extracted as feature vectors, which significantly improves the sensitivity of impurity signal identification and reduces the risk of missed detection.

[0022] A classifier for impurity release patterns based on a backpropagation neural network is constructed. The classifier takes the feature vector as input and the impurity type, release location and magnetization intensity as output. It establishes a mapping relationship between the release signal and the physical properties of the impurities, identifies high-risk impurities that need to be removed first, and achieves accurate classification and location of impurity properties, providing an accurate basis for targeted removal.

[0023] The identification results are transmitted in real time to the control module of the non-magnetic shielding system, which drives the coil-type non-magnetic shielding system to apply a targeted pulse demagnetizing magnetic field with adjustable pulse width and progressively decreasing amplitude to the identified release area. This achieves directional attenuation of the impurity magnetic moment and avoids additional magnetic pollution to surrounding components.

[0024] Preferably, S3 specifically includes:

[0025] A triaxial fluxgate sensor, a vibration accelerometer, and a particulate matter counter are deployed in the assembly cavity to establish a spatiotemporal synchronous acquisition mechanism for the sensor array. At a sampling rate of over 10kHz, magnetic field vectors, tooling vibration characteristics, and particulate matter concentration time-series data are acquired synchronously, achieving microsecond-level precise alignment of multi-source data and providing a highly consistent data foundation for subsequent interference separation.

[0026] A multi-sensor data fusion model is constructed based on extended Kalman filtering. The instantaneous magnetic field interference generated by friction of non-magnetic tooling is defined as time-varying observation noise. Through two-step iteration of state prediction and observation update, the background magnetic field of the environment and the dynamic friction interference magnetic field are separated in real time, effectively stripping away the dynamic friction interference and restoring the original characteristics of the weak magnetic pollution signal.

[0027] The separated net magnetic field signal is spatially interpolated and fused with the synchronously acquired particulate matter concentration data to generate a four-dimensional spatiotemporal distribution map of the impurity migration trajectory. This accurately reconstructs the path and speed of the impurities migrating from the release location to the bearing working surface or air gap, enabling visualized tracking of the entire impurity migration process and providing centimeter-level spatial positioning for precise blocking.

[0028] Preferably, S4 specifically includes:

[0029] A database of magnetic field anomaly waveform features for the entire process of impurity release, migration, and deposition was established. Typical waveform templates were stored in the database according to impurity material, particle size range, and deposition location. The database was continuously expanded and updated using actual assembly data, which significantly improved the accuracy and adaptability of the feature database in identifying typical deposition events.

[0030] The system uses a sliding time window to capture the waveform of the monitored magnetic field in real time. It calculates the minimum warping distance between the magnetic field and each template in the feature library through a dynamic time warping algorithm. When the distance is lower than the set threshold, it is identified as a transient adsorption-deposition event and the event type and spatial coordinates are output, achieving millisecond-level accurate identification and spatial positioning of deposition events.

[0031] Upon detecting a deposition event, the non-magnetic pulse purging device arranged in the assembly chamber is triggered to spray non-magnetic clean gas within a 30mm radius around the deposition coordinates. At the same time, the current assembly operation is paused. Assembly is resumed after the magnetic field waveform returns to the background level, effectively curbing the chain diffusion of magnetic pollution and ensuring the cleanliness of the magnetic field in the assembly environment.

[0032] Preferably, S4 further includes:

[0033] When pulse purging fails to block or when deposition events occur three times in a row, the local non-magnetic adsorption device at the bottom of the assembly cavity is automatically activated. The magnetic particles in the deposition area are attracted to the dedicated collection tank through the electromagnetic field gradient, preventing the impurities from escaping again, effectively preventing the secondary diffusion of magnetic particles, and ensuring the cleanliness of the magnetic field of the assembly cavity.

[0034] The tooling operation type, fastening torque and friction pair material are recorded simultaneously when a deposition event occurs. High-risk operation nodes that induce deposition are identified through correlation analysis. Operation warning prompts are generated and pushed to the assembly personnel's terminal to achieve accurate warning of high-risk operations, improve operation standardization and reduce the recurrence rate of deposition events.

[0035] The magnetic field recovery curve after the blocking operation is compared with the historical normal curve to verify the blocking effectiveness. The verification result is used as the basis for updating the confidence weight of the waveform template in the feature library, thereby improving the accuracy of subsequent matching and identification, realizing quantitative verification of the blocking effect, and continuously improving the recognition accuracy of the feature library and the reliability of the blocking strategy.

[0036] Preferably, S5 specifically includes:

[0037] A three-dimensional finite element model of the core assembly of the magnetic levitation compressor was established using finite element simulation. The residual stress field distribution data after assembly and the boundary conditions of the geomagnetic field environment were imported. The magnetic-mechanical-thermal multiphysics coupling method was used to simulate the magnetic creep displacement behavior of magnetic particles during the stress relaxation process, ensuring the simulation accuracy of the particle migration path and the consistency of the physical process.

[0038] The initial position, particle size, magnetic susceptibility, and shape factor of the particles are defined as probability distribution parameters. More than 1,000 random sampling calculations are performed using the Monte Carlo method to statistically determine the cumulative probability of particles migrating to the sensor probe area of ​​the magnetic levitation bearing and the sensitive position of the air gap, thereby achieving accurate quantification of particle migration probability and reliable risk assessment.

[0039] A secondary magnetic contamination risk map of the assembly is generated based on the migration probability distribution, and the spatial coordinates of high-risk areas are marked to provide a quantitative basis for the selection of subsequent temperature-magnetic composite field application areas.

[0040] Preferably, S6 specifically includes:

[0041] Based on the identified high-risk areas, after assembly, the assembly is placed in the temperature-magnetic composite treatment station of the three-condition assembly equipment. The local heater and multi-pole electromagnetic coil are used in synergistic control to apply an alternating composite field with a temperature of 5°C to 10°C and a magnetic field strength of 0.1mT to 0.5mT in the risk area. Through the synergistic effect of temperature and magnetic fields, the active stabilization of deposited particles is achieved, eliminating the risk of migration in the early stage of service.

[0042] During the application of the alternating composite field, the magnetic moment orientation change of the deposited particles is monitored in real time by an embedded magnetic field probe. The frequency and phase of the alternating magnetic field are dynamically adjusted by a closed-loop control algorithm, which promotes the particles to oriented and anchor in non-sensitive areas under the synergistic effect of thermo-magnetic interaction, ensuring that the oriented arrangement of particles is accurate and controllable, and improving the anchoring consistency and long-term stability.

[0043] After the treatment was completed, the residual magnetism of the high-risk area was retested using a high-precision magnetometer array to confirm that the magnetic moment of the deposited particles had stabilized and had not migrated to the sensor area. This formed a record of the stabilization treatment of the magnetic pollution source after assembly, verified the treatment effect, and provided reliable data support for quality traceability.

[0044] Preferably, S6 further includes:

[0045] By aligning the data on temperature, humidity, cleanliness, and magnetic field strength collected by the environmental monitoring system with the tooling usage records and fastening torque data of the assembly operating system along the time axis, a digital twin dataset of the entire assembly process is constructed, providing a high-precision and highly consistent data foundation for the correlation analysis of assembly parameters and magnetic contamination risk.

[0046] The association rule mining algorithm is used to analyze the correlation between assembly parameters and the identification results of secondary magnetic contamination risk points, extract key influencing factors, generate dynamic correction suggestions for assembly parameters, and push them to the process management platform, realizing the transformation of assembly parameters from experience setting to data-driven precise correction.

[0047] The revised assembly parameters are used as the initial input for the next batch of assembly. Through inter-batch iterative optimization, the thermal cycling parameters, demagnetization threshold, and blocking trigger conditions are continuously updated, forming an adaptive optimization closed loop for the magnetic pollution source control strategy. This creates an intelligent optimization mechanism where the control strategy automatically evolves with each batch and its effects continuously converge.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] 1. This method for optimizing the prevention and control of magnetic pollution sources in the assembly process of a magnetic levitation compressor involves constructing an adaptive genetic algorithm-optimized low-temperature thermal cycling process in the pre-assembly treatment stage. This process precisely induces the early release of hidden ferromagnetic impurities in microcracks. Combined with wavelet packet transform and neural network-driven targeted pulse demagnetization, it achieves precise removal of impurities before they enter the assembly stage. This method overcomes the limitations of traditional cleaning and demagnetization methods that cannot reach impurities within micro-defects, thus eliminating the potential magnetic pollution hazards that cannot be detected by initial testing at the source.

[0050] 2. This method for optimizing the prevention and control of magnetic pollution sources in the assembly process of a magnetic levitation compressor achieves real-time separation of frictional interference from non-magnetic tooling and accurate reconstruction of impurity migration trajectories by constructing a multi-sensor data fusion model and a dynamic time warping matching algorithm. It can identify transient adsorption-deposition events and automatically trigger pulse purging or local adsorption blocking, thus upgrading magnetic pollution prevention and control from passive detection to active intervention. This effectively curbs the chain diffusion of impurities in key parts and significantly reduces the probability of magnetic pollution events during assembly.

[0051] 3. This method for optimizing the prevention and control of magnetic pollution sources during the assembly process of a magnetic levitation compressor establishes a risk assessment model for the residual stress field and magnetic creep migration of magnetic particles under geomagnetic conditions through coupled analysis of finite element simulation and Monte Carlo method. It can quantitatively identify secondary pollution risk points in sensitive areas such as sensor probes and air gaps, extending magnetic pollution prevention and control from the assembly site to the entire service life, providing accurate spatial positioning basis for subsequent stabilization treatment, and fundamentally avoiding magnetic field drift failures caused by slow particle drift. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the workflow of an optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to the present invention.

[0053] Figure 2 This is a schematic diagram of the method flow for optimizing the prevention and control of magnetic pollution sources during the assembly process of a magnetic levitation compressor according to the present invention.

[0054] Figure 3 The core assembly residual stress distribution cloud map (magnetic-mechanical-thermal multi-physics field coupling simulation) is the optimization method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor according to the present invention.

[0055] Figure 4 This is a secondary magnetic pollution risk probability distribution diagram (Monte Carlo 2000 sampling simulation) for an optimized method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor according to the present invention.

[0056] Figure 5 This invention provides a magnetic particle creep displacement trajectory diagram (residual stress + geomagnetic field coupling effect) for an optimized method of magnetic pollution source control in the assembly process of a magnetic levitation compressor.

[0057] Figure 6 This is the core assembly finite element model and mesh generation diagram of the optimization method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor according to the present invention;

[0058] Figure 7 This is a flowchart illustrating the thermal cycle parameter optimization method for a magnetic pollution source control optimization method in the assembly process of a magnetic levitation compressor according to the present invention.

[0059] Figure 8 This is a flowchart of the impurity signal feature extraction and targeted pulse demagnetization process of an optimized method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor according to the present invention.

[0060] Figure 9 This is a flowchart of multi-sensor data fusion and impurity migration trajectory reconstruction of an optimized method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor according to the present invention.

[0061] Figure 10 This is a flowchart of the finite element modeling analysis and Monte Carlo risk assessment of an optimization method for preventing and controlling magnetic pollution sources in the assembly process of a magnetic levitation compressor, as described in this invention.

[0062] Figure 11 This is a flowchart of the temperature-magnetic composite field stabilization treatment and closed-loop feedback optimization process for the magnetic pollution source control optimization method in the assembly process of a magnetic levitation compressor according to the present invention. Detailed Implementation

[0063] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0064] Example 1, please refer to Figures 1 to 9 This invention provides a technical solution: an optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor, applicable to the non-magnetic assembly process of core components of various magnetic levitation compressor products such as magnetic levitation centrifugal compressors and magnetic levitation scroll compressors, comprising the following steps:

[0065] S1. In the pre-assembly treatment stage of the magnetic levitation compressor, an adaptive genetic algorithm is used to dynamically optimize the thermal cycling parameters under low-temperature conditions. By precisely controlling the amplitude and frequency of temperature alternation, the algorithm precisely induces the early release of ferromagnetic impurities hidden in the micro-defects of the compressor components, thereby eradicating the source of hidden magnetic contamination and effectively eliminating hidden impurities within micro-defects, blocking the initial source of magnetic contamination. The assembly environment is a dust-free environment, meeting the cleanliness requirements of ISO 14644-1 Class 7 and above. It is a low-temperature constant-temperature environment, with the temperature controlled within the range of 5℃~15℃, a temperature difference ≤±1℃, and humidity ≤60%RH. A thermal cycling process parameter model is constructed with temperature alternation amplitude, alternation frequency, and low-temperature holding time as decision variables. The dual-objective optimization function is set to maximize the impurity release rate and minimize the residual magnetism increment of the components. A penalty function is introduced to handle the constraint conditions of temperature change rate on the microstructure of the magnetic levitation compressor components, effectively avoiding thermal stress damage and ensuring the integrity of the component's microstructure. The thermal cycling parameter population is initialized using real-number encoding. By adapting and adjusting the crossover and mutation probabilities, the population diversity is dynamically maintained during the iteration process. The optimal parameter combination for inducing the release of hidden ferromagnetic impurities in microcracks under low temperature conditions is selected by combining Pareto non-dominated sorting. The optimal process window is precisely locked, and the impurity release efficiency and consistency are improved. The optimized thermal cycling parameters are input into the low temperature constant temperature system of the three-condition assembly equipment to perform thermal cycling treatment with an alternating amplitude of ±3℃ to ±5℃ and a frequency of 2 to 4 times per hour. This precisely induces the extrusion of nanoscale ferromagnetic impurities in micro-defects during the material shrinkage-expansion process, thereby eliminating the source of hidden magnetic pollution and reducing the risk of magnetic pollution in subsequent assembly.

[0066] In addition, the core principles of the assembly process include non-magnetic priority, environmental control, and closed-loop process. Non-magnetic priority means that all tools, fixtures, and consumables that come into contact with core components must be made of non-magnetic materials (austenitic stainless steel, titanium alloy, engineering plastics, etc.), and ferromagnetic materials are strictly prohibited from entering the assembly area. Environmental control means that the three-condition assembly equipment must continuously and stably provide a dust-free, low-temperature, and non-magnetic environment, and environmental parameters must be monitored in real time during the assembly process, with immediate shutdown in case of abnormalities. Closed-loop process means that the entire process, from component cleaning, fixture preparation, assembly operation, quality inspection to finished product protection, is traceable and controllable.

[0067] It should be noted that the temperature alternation range is adjustable in three levels: ±3.5℃, ±4.0℃, and ±4.5℃; the alternation frequency is set to three levels: 2 times, 3 times, and 4 times per hour; and the low-temperature holding time is set to three levels: 10 minutes, 15 minutes, and 20 minutes. The dual-objective optimization function is to maximize the impurity release rate and minimize the remanent magnetization increment of the components. The impurity release rate is characterized by the difference in magnetic field strength before and after assembly, and the remanent magnetization increment is represented by the maximum remanent magnetization change on the surface of the core component measured by a magnetometer. A penalty function is introduced to handle the temperature change rate constraint, stipulating that the temperature rise / fall rate must not exceed 1.2℃ / min to avoid thermal stress concentration leading to microstructural damage to the rotor or stator core. Decision variables are encoded with real numbers, and the population size is set to [missing information]. The algorithm has a maximum of 200 iterations and uses Pareto non-dominated sorting to select the optimal parameter combination. During the genetic algorithm iteration, the crossover probability is adaptively adjusted: 0.85 when the average fitness of the current generation is lower than the global average fitness, and 0.65 otherwise. The mutation probability is dynamically adjusted from 0.02 to 0.12 with decreasing iteration count to maintain population diversity and avoid premature convergence. For the three core components of the compressor—rotor, stator, and bearings—material thermophysical parameter sub-models are established. The rotor material is titanium alloy TC4, the stator is non-magnetic electrical steel, and the bearings are austenitic stainless steel SUS316L, with fitness functions assigned to 0.4, 0.35, and 0.25 respectively. After optimizing and iterating the differential weighting coefficients, the dedicated thermal cycling process curves for each component are output. The rotor operates at an alternating amplitude of ±4.5℃, a frequency of 4 times per hour, and a holding time of 15 minutes; the stator operates at ±4.0℃, 3 times per hour, and a holding time of 20 minutes; and the bearing operates at ±3.5℃, 2 times per hour, and a holding time of 10 minutes. The three-condition assembly equipment uses a fully enclosed scroll-type non-magnetic compressor and a PID temperature control module, with a temperature control accuracy of ±0.5℃ and a cavity temperature uniformity of ±0.3℃. During the thermal cycling process, 12 PT100 platinum resistance temperature sensors arranged on the surface of the components collect temperature field distribution data in real time. At the same time, a triaxial fluxgate sensor is used to monitor local remanent magnetization changes at a sampling rate of 20Hz. A temperature-remanence dynamic response surface is constructed. The response surface data is fed back to the fitness function of the adaptive genetic algorithm every 15 minutes for online correction. When the remanence increment in a certain area exceeds 0.08mT, the low temperature holding time of the corresponding component in that area is automatically extended by 5 minutes to ensure that the hidden ferromagnetic impurities in the microcracks are fully squeezed out. The total thermal cycling time is controlled within the range of 90 to 120 minutes. The three-condition assembly equipment must meet the requirements of "three-condition coordination, non-magnetic compatibility, convenient operation, and stable reliability". The main structure uses non-magnetic materials to avoid generating magnetic interference. Each system (dust-free purification system, low temperature constant temperature system, and non-magnetic shielding system) is independently controlled and does not affect each other. Real-time monitoring and data recording of the assembly process are supported.

[0068] Among them, the cleanliness and grade requirements of the dust-free purification system are as follows: the cleanliness of the assembly chamber reaches ISO 14644-1 Class 7 (static), the air exchange rate is ≥60 times / h, and the pressure difference is controlled at 10Pa~20Pa (relative to the outside) to prevent external pollutants from entering.

[0069] Assembly cavity material: Made of 304 austenitic stainless steel (non-magnetic), with a smooth inner wall and no dead corners to avoid particle accumulation; the seals are made of fluororubber (non-magnetic and low-temperature resistant); Purification device: Equipped with a high-efficiency air filter (HEPA, filtration efficiency ≥99.97%@0.3μm), a pre-filter, and a medium-efficiency filter, forming a closed-loop airflow organization of "air inlet-filtration-cavity-return air" to avoid local airflow dead corners; Entrance and exit design: Adopts double-door interlocking clean doors and is equipped with an air shower. Personnel / materials must pass through the air shower for purification before entering. Material transfer uses a non-magnetic transfer window to ensure that no contaminants are brought in during the transfer process;

[0070] The control functions of the cleanroom system are as follows: real-time monitoring of particulate matter concentration, pressure difference, and air exchange rate within the chamber; equipped with an audible and visual alarm device; when the particulate matter concentration exceeds the standard (≥5μm particulate matter > 3520 particles / m³) or the pressure difference is abnormal, an alarm is immediately triggered and an emergency purification program is initiated; data storage (≥1 year) and export are supported.

[0071] The temperature and humidity parameters required for the low-temperature constant temperature system are as follows: temperature control range 5℃~15℃, adjustable as needed, constant temperature accuracy ≤±1℃; relative humidity ≤60%RH, humidity fluctuation ≤±5%RH; internal temperature uniformity ≤±0.5℃ (temperature difference between any two points). The refrigeration system employs a non-magnetic compressor (fully enclosed scroll-type non-magnetic compressor) and a non-magnetic heat exchanger, using environmentally friendly R410A refrigerant. Ferromagnetic pipes and components are avoided to prevent magnetic interference. For its insulation structure, the outer wall of the cavity is covered with a polyurethane insulation layer (thickness ≥50mm), and the inner wall is treated with anti-condensation to prevent condensation from contaminating components in low-temperature environments. For temperature and humidity control, a high-precision temperature and humidity sensor is configured (measurement accuracy: temperature ±0.1℃, humidity ±2%RH), employing PID intelligent regulation to achieve automatic and stable temperature and humidity control. A dehumidification device is also included to prevent component corrosion or particulate matter adsorption caused by high humidity.

[0072] The safety requirements for the low-temperature constant temperature system are as follows: it is equipped with over-temperature alarm (upper limit 18℃, lower limit 3℃) and over-humidity alarm (upper limit 65%RH) functions, and an emergency heating device to prevent components from becoming brittle due to excessively low temperatures. The refrigeration system is equipped with pressure protection and overload protection to avoid system failure.

[0073] The shielding level requirements for the non-magnetic shielding system are as follows: the magnetic field strength inside and outside the assembly cavity is controlled. When the external ambient magnetic field is ≤5mT, the magnetic field strength inside the cavity is ≤0.1mT, effectively shielding the ferromagnetic materials and the interference of the external magnetic field on the assembly components.

[0074] For the shielding structure of the non-magnetic shielding system, the shielding layer adopts a double-layer non-magnetic shielding structure. The inner layer is permalloy (thickness ≥ 2mm), and the outer layer is 304 stainless steel (thickness ≥ 3mm). The distance between the two layers is ≥ 50mm, forming a magnetic field attenuation channel. The shielding layer covers all surfaces of the cavity (including the top, bottom, sides, and door), with no shielding blind spots. Its interface shielding uses non-magnetic shielded connectors at electrical and pipe interfaces, and shielded cables to prevent magnetic fields from entering the cavity through the interfaces and cables. The interfaces are tightly sealed, meeting both shielding and dust-free requirements. Its tooling shielding... The equipment's built-in assembly tooling and fixtures are all made of non-magnetic materials (titanium alloy, engineering plastics, etc.), and the tooling surfaces are treated with non-magnetic treatment (demagnetizing annealing, residual magnetism ≤0.05mT). Any ferromagnetic components are prohibited from entering the shielded cavity, and a high-precision magnetic field sensor (measurement accuracy ≤0.01mT) is installed inside the cavity to monitor the magnetic field strength in real time. When the magnetic field exceeds the standard (>0.1mT), an alarm is immediately triggered and the external magnetic field source is cut off. The equipment is equipped with a non-magnetic demagnetizing device (coil-type non-magnetic demagnetizer) for demagnetizing components and tooling before assembly to ensure that the residual magnetism after demagnetization meets the requirements.

[0075] The three-condition assembly equipment is also equipped with a lighting system, an operating platform, and data management functions. The lighting system uses non-magnetic LED explosion-proof lights with an illuminance of ≥500lx to avoid magnetic interference or heat from the lighting equipment affecting the low-temperature environment. The operating platform is a non-magnetic operating platform with adjustable height, a non-magnetic anti-slip mat on the surface, a load-bearing capacity of ≥50kg, and non-magnetic tooling fixtures to support the assembly and adaptation of multiple parts. The data management system integrates the collection and storage of environmental parameters (temperature, humidity, particulate matter concentration, magnetic field strength), assembly process, and quality inspection data, and supports integration with the enterprise's MES system to achieve full-process traceability.

[0076] Furthermore, S1 also includes: during the thermal cycling process, real-time acquisition of surface temperature field distribution data and local remanent magnetization change data of components, construction of temperature-remanent magnetization dynamic response surface, and feedback to the fitness function of adaptive genetic algorithm to realize online correction of thermal cycling parameters in the pre-assembly processing stage. Through real-time feedback, dynamic adaptation of thermal cycling parameters is achieved, significantly improving impurity release efficiency and process consistency. For the three core components of compressor rotor, stator and bearing, material thermophysical parameter sub-models are established respectively. Component differentiation weight coefficients are introduced into the genetic algorithm to generate component-specific thermal cycling process curves, avoiding over-processing or under-processing of some components under uniform parameters. Thermal cycling is applied differently according to material characteristics to ensure balanced performance of each component. After the thermal cycling is completed, the remanent magnetization of the component surface is scanned by a high-precision magnetic field imager to identify micro-area remanent magnetization anomalies left after impurity release. The scanning results are used as input for the next round of assembly batch thermal cycling parameter optimization, forming an inter-batch iterative optimization closed loop to continuously enhance the release effect of hidden magnetic pollution sources.

[0077] It should be noted that 12 PT100 platinum resistance temperature sensors and 6 triaxial fluxgate sensors are evenly distributed on the surface of the component to be assembled. The PT100 platinum resistance temperature sensors are used to collect temperature field distribution data, and the triaxial fluxgate sensors monitor the dynamic changes of local remanence at a sampling frequency of 20Hz. The collected data is transmitted to the central control system in real time to construct a dynamic response surface with temperature as the abscissa and remanence increment as the ordinate. Every 15 minutes, the surface feature parameters are fed back to the fitness function of the adaptive genetic algorithm. When the remanence increment in the monitored area exceeds the 0.08mT threshold, the current thermal cycle parameters are automatically corrected, and the low temperature holding time of the corresponding component in that area is extended by 5 minutes. This ensures that the ferromagnetic impurities hidden in the microcracks are fully squeezed out during the material shrinkage-expansion process, realizing online closed-loop correction of thermal cycle parameters in the pre-assembly treatment stage. The rotor material is titanium alloy TC4, with a linear expansion coefficient of 8.6×10. -6 / K, assigned a weight coefficient of 0.4 in the fitness function; the stator material is non-magnetic electrical steel with a thermal conductivity of 25 W / (m·K), weight coefficient 0.35; the bearing material is austenitic stainless steel SUS316L with a magnetic permeability of less than 1.05, weight coefficient 0.25. During the Pareto non-dominated sorting process, the genetic algorithm optimizes the thermal cycling parameters of each component based on the differentiated weight coefficients, ultimately outputting a component-specific process curve: the rotor performs alternating amplitude ±4.5℃, frequency 4 times per hour, and holding time 15 minutes; the stator... The sub-component is subjected to ±4.0℃, 3 times per hour, and a holding time of 20 minutes; the bearing is subjected to ±3.5℃, 2 times per hour, and a holding time of 10 minutes, to avoid over- or under-processing of some components due to uniform parameters; the high-precision magnetic field imager has a measurement accuracy of ±0.01mT and a spatial resolution of 1mm×1mm. The scanning results are output in the form of a two-dimensional remanent magnetization distribution map, which identifies micro-area remanent magnetization anomalies left after impurity release. The criteria for judging anomalies are that the local remanent magnetization is more than 0.05mT higher than the surrounding area and the area is not less than 1mm². 2 The coordinates of the abnormal points, the peak value of the residual magnetism and the corresponding component type are entered into the batch process database as input data for the next round of assembly batch thermal cycle parameter optimization. The optimization algorithm dynamically adjusts the initial population distribution range of the thermal cycle parameters based on the distribution characteristics of the abnormal points in the historical batches, forming an iterative optimization closed loop between batches, and continuously improving the release effect of hidden magnetic pollution sources.

[0078] S2. Combining wavelet packet transform and neural network algorithms, feature extraction and pattern recognition are performed on the released impurity signals to guide the non-magnetic shielding system in targeted pulse demagnetization. This achieves precise removal of ferromagnetic impurities before the assembly of the magnetic levitation compressor, realizing accurate impurity positioning and directional attenuation of magnetic moment, thus improving removal efficiency. A multi-channel high-bandwidth magnetic field sensor array is used to collect transient magnetic field fluctuation signals generated by impurity release during thermal cycling. Wavelet packet transform is used to decompose the signal into three layers, extracting the energy entropy, singular values, and peak frequencies of each frequency band as feature vectors. This significantly improves the sensitivity of impurity signal identification, reduces the risk of missed detection, and constructs... The impurity release pattern classifier based on backpropagation neural network takes feature vectors as input and impurity type, release location and magnetization intensity as output. It establishes a mapping relationship between release signals and impurity physical properties, identifies high-risk impurities that need to be removed first, and achieves accurate classification and location of impurity properties. This provides an accurate basis for targeted removal. The identification results are transmitted to the control module of the non-magnetic shielding system in real time, driving the coil-type non-magnetic shielding system to apply a targeted pulse demagnetizing magnetic field with adjustable pulse width and progressively decreasing amplitude in the identified release area. This achieves directional attenuation of the impurity magnetic moment and avoids additional magnetic pollution to surrounding components.

[0079] It should be noted that during the thermal cycling process, a multi-channel high-bandwidth magnetic field sensor array consisting of 16 triaxial fluxgate sensors is used. The sensor bandwidth is DC to 1kHz, and the sampling frequency is uniformly set to 2kHz. These sensors are evenly distributed circumferentially along the inner wall of the assembly cavity to collect transient magnetic field fluctuation signals generated by impurity release in real time. The central control system performs wavelet packet transform on the collected raw signals and uses the Daubechies 4 wavelet basis function for three-level decomposition, decomposing the signal into 8 frequency band components. For each frequency band, three feature parameters are calculated: energy entropy, singular values, and peak frequency. Energy entropy reflects the uniformity of signal energy distribution across frequency bands. Singular values ​​are extracted using matrix singular value decomposition to extract the principal component features of the signal. Peak frequency is used to identify the dominant frequency component of the impurity release event. The feature extraction process is completed within 500ms after signal acquisition, forming a 24-dimensional feature vector. A three-layer backpropagation neural network is constructed as the impurity release pattern classifier, with 24 nodes in the input layer, corresponding to the features... The vector dimension is set to 32 hidden layer nodes using the Sigmoid activation function; the output layer has 3 nodes, outputting the impurity type, release location coordinates, and magnetization estimate. The impurity release pattern classifier is pre-trained with at least 500 sets of calibrated samples with known impurity properties, a training target error of 0.01, and an upper limit of 1000 iterations. When the real-time feature vector is input to the classifier, the recognition result is output within 200ms. The impurity types are classified into ferrite, martensite, and austenite, and the release location is determined by the coordinates of the sensor array. The system uses three-dimensional coordinates with a positioning accuracy of ±5mm. The estimated magnetization intensity is a continuous value within the range of 0.01mT to 1mT. The confidence threshold for the impurity release mode classifier is set to 0.85; identification results below this threshold are discarded, triggering a re-acquisition mechanism. The coil-type non-magnetic shielding system consists of six independently controllable Helmholtz coils. Each coil has a maximum output magnetic field strength of 10mT and an effective working area diameter of 80mm. The control module automatically activates the 1st to 2 nearest coils based on the release position coordinates, applying a targeted pulse demagnetizing magnetic field. The pulse demagnetizing waveform... Using a sinusoidal decay method, the initial pulse amplitude is set to 1.5 times the estimated impurity magnetization intensity, and the pulse width is adjustable from 50μs to 500μs. The subsequent pulse amplitude decreases gradually at a rate of 20% per cycle until it is lower than 0.01mT. The total time for a single demagnetization operation does not exceed 2 seconds. After demagnetization, the system verifies whether the residual magnetism in the demagnetized area has dropped below 0.02mT through a sensor array. If it does not meet the standard, the demagnetization process is automatically repeated. This targeted demagnetization mechanism achieves precise attenuation of the magnetic moment of the released impurities, avoiding additional magnetic contamination to surrounding components.

[0080] In addition, environmental preparation, material preparation, tool and personnel preparation should be carried out before assembly.

[0081] For environmental preparation: Start the three-condition assembly equipment, and sequentially turn on the dust-free purification system, the low-temperature constant temperature system, and the non-magnetic shielding system. After running for 30 minutes, check the environmental parameters: the cleanliness level reaches Class 7, the temperature is stable within the set range (5℃~15℃), and the magnetic field strength is ≤0.1mT. Assembly can only be carried out after all parameters are qualified. Clean the debris and particulate matter in the assembly cavity, and wipe the operating platform and tooling fixtures with a non-magnetic, dust-free cloth soaked in anhydrous ethanol (non-magnetic grade) to ensure that there are no pollutants or ferromagnetic impurities remaining. Check that the environmental monitoring equipment (particulate matter sensor, temperature and humidity sensor, magnetic field sensor) is operating normally and that the data recording function is enabled.

[0082] For material preparation: Assembly components (rotors, stators, bearings, etc.) must undergo non-magnetic cleaning and demagnetization treatment. The residual magnetism of the components must be ≤0.3mT, and the surface must be free of oil, particles, and damage before they can be sent to the assembly area. Materials are sent into the assembly cavity through a non-magnetic transfer window. Before transfer, the materials must be purified by a dust-free air shower in the transfer window for 10 seconds to prevent external contaminants from being brought in. Components are placed in categories using non-magnetic trays and non-magnetic storage boxes. Components must not come into direct contact with ferromagnetic objects. The spacing between components should be ≥10cm to avoid collisions and friction.

[0083] Regarding tool and personnel preparation: Assembly tools (wrenches, screwdrivers, tweezers, etc.) are all made of non-magnetic materials (304 stainless steel, titanium alloy), and have undergone demagnetization treatment (residual magnetism ≤0.05mT). They are marked with "Non-magnetic for Use Only". Before use, tools should be wiped clean with a non-magnetic, dust-free cloth and placed in a non-magnetic tool box. Operators must wear cleanroom suits, cleanroom shoes, cleanroom gloves, and cleanroom masks. They must enter the assembly area after passing through an air shower. Wearing ferromagnetic jewelry (watches, necklaces, rings, etc.) is strictly prohibited, as is bringing ferromagnetic tools or items into the cavity. Operators must undergo professional training, be familiar with the non-magnetic assembly process requirements, the operation methods of the three-condition assembly equipment, and emergency handling procedures, and can only be employed after passing the assessment.

[0084] S3. Construct a multi-sensor data fusion model based on Kalman filtering to separate the instantaneous magnetic field interference generated by friction of the non-magnetic tooling from the background magnetic field in real time. This accurately reconstructs the migration trajectory of impurities during the assembly of the magnetic levitation compressor, suppresses dynamic friction interference, and restores the actual migration path, providing a basis for blocking interference. A triaxial fluxgate sensor, vibration accelerometer, and particulate counter are deployed within the assembly cavity. A spatiotemporal synchronous acquisition mechanism for the sensor array is established, simultaneously acquiring magnetic field vectors, tooling vibration characteristics, and particulate concentration time-series data at a sampling rate of over 10kHz. This achieves microsecond-level precise alignment of multi-source data, providing a highly consistent data foundation for subsequent interference separation. An extended Kalman filter is used to construct a multi-sensor data fusion model. The instantaneous magnetic field interference generated by friction of non-magnetic tooling is defined as time-varying observation noise. Through two-step iteration of state prediction and observation update, the background magnetic field of the environment and the dynamic friction interference magnetic field are separated in real time. The dynamic friction interference is effectively stripped off, and the original characteristics of the weak magnetic pollution signal are restored. The separated net magnetic field signal is spatially interpolated and fused with the synchronously collected particulate matter concentration data to generate a four-dimensional spatiotemporal distribution map of impurity migration trajectory. The path and speed of impurities migrating from the release position to the bearing working surface or air gap are accurately restored, realizing the visualization and tracking of the entire impurity migration process and providing centimeter-level spatial positioning for precise blocking.

[0085] It should be noted that eight triaxial fluxgate sensors, six vibration accelerometers, and four particle counters are uniformly arranged circumferentially on the inner wall of the assembly cavity. The bandwidth of the triaxial fluxgate sensors is DC to 1kHz, the vibration accelerometer range is ±10g and the frequency response range is 0.5Hz to 5kHz, and the particle counters detect particle sizes ranging from 0.3μm to 10μm. Each sensor achieves microsecond-level time alignment through a synchronous clock module. The central control system synchronously acquires the three components of the magnetic field vector, the time-domain waveform of the tooling vibration acceleration, and the time-series data of particle concentration at a sampling rate of 12kHz. The acquired data is transmitted to the data fusion server in real time after hardware filtering and analog-to-digital conversion, forming a spatiotemporal synchronous dataset of multi-source heterogeneous data. The covariance matrix of the multi-sensor data fusion model is dynamically adjusted according to the friction and impact intensity measured by the vibration accelerometer in real time. When the friction and impact intensity is high, the observation noise covariance is increased and the fusion weight of the measured values ​​of the magnetic field sensor is reduced. When the friction and impact intensity is low, the observation noise covariance is restored to the background level. The state prediction equation is used to estimate the... The background magnetic field of the environment is measured, and the residual between the measured and predicted values ​​of the magnetic field sensor is used for observation and updating. After two iterations, the net magnetic field signal is separated. This signal represents the weak magnetic field disturbance generated during the migration of impurities. During the data fusion process, vibration accelerometer data is used to identify tooling operation events and remove interference signals in the corresponding time period. Particle counter data is used to verify the authenticity of impurity release events. The three work together to effectively suppress dynamic friction interference. A four-dimensional spatiotemporal distribution field is constructed based on the three-dimensional coordinate system of the assembly cavity. The time resolution is 0.1 seconds and the spatial grid spacing is 10 mm. The coordinates of the impurity release source are located according to the spatial gradient change of the net magnetic field signal. Combined with the phase relationship between the time-series peak value of particulate matter concentration and the magnetic field fluctuation waveform, the path trajectory and instantaneous velocity of the impurities migrating from the release position to the bearing working surface or air gap are calculated. The migration velocity calculation accuracy is ±2 mm / s. The generated four-dimensional spatiotemporal distribution map is displayed in real time on the central control interface in the form of pseudo-color cloud map superimposed with contour lines, providing accurate spatial positioning basis for subsequent blocking mechanisms.

[0086] At two adjacent sampling times and Between these points, the impurity migration trajectory is described using linear interpolation, and its expression is as follows:

[0087] ;

[0088] In the formula: The coordinates of the impurity's position in three-dimensional space as a function of time; Let be time, representing the time between two adjacent sampling moments. and The continuous time variables between these variables are used to describe the linear interpolation process of the impurity migration trajectory; for The position coordinates of the impurity in three-dimensional space at any given time; for The position coordinates of the impurity in three-dimensional space at any given time; For the first Each sampling time; For the first Each sampling time; For time resolution;

[0089] Impurities in The instantaneous velocity vector at time t is defined as:

[0090] ;

[0091] The directional components are:

[0092] ;

[0093] ;

[0094] ;

[0095] The magnitude of the instantaneous velocity is:

[0096] ;

[0097] In the formula: For impurities in The instantaneous velocity vector at a given moment; These represent the velocity components of the impurities in the X, Y, and Z directions, respectively. For impurities in The instantaneous migration speed of a moment;

[0098] Starting point of impurity migration path Determined by the spatial gradient change of the net magnetic field signal, satisfying:

[0099] ;

[0100] in: The net magnetic field signal in space With time Gradient magnitude at; This is the magnetic field gradient threshold, used to identify impurity release events;

[0101] The subsequent positions along the impurity migration path are determined by jointly analyzing the phase relationship between the temporal peak values ​​of particulate matter concentration and the magnetic field fluctuation waveform:

[0102] ;

[0103] in: The instantaneous phase of the particulate matter concentration time series signal; This represents the instantaneous phase of the net magnetic field fluctuation waveform; The phase difference between the two; This is the phase difference threshold used to determine the consistency of impurity migration events;

[0104] The endpoint of impurity migration is defined as the moment when impurities first arrive at the bearing working surface or the air gap sensitive area. ,satisfy:

[0105] ;

[0106] in: For impurities to enter the sensitive area for the first time ( The three-dimensional spatial coordinates at time ( ). A predefined set of sensitive areas, including the bearing working surface, air gap, and the area surrounding the sensor probe; The moment when impurities first enter the sensitive area;

[0107] S4. Introducing a dynamic time warping algorithm, the real-time monitored magnetic field anomaly waveform is matched with a preset impurity release feature library to locate transient adsorption-deposition events and automatically block them, curbing the chain spread of magnetic pollution, identifying deposition events in milliseconds, and automatically blocking them to prevent pollution spread.

[0108] S5. Finite element simulation is used to pre-simulate the magnetic creep path of magnetic particles under the coupling of residual stress and geomagnetism, and Monte Carlo sampling method is combined to evaluate the particle displacement probability, identify secondary magnetic contamination risk points, quantitatively assess long-term migration risks, and accurately locate potential fault areas.

[0109] S6. Apply a controllable temperature-magnetic composite field to promote the early anchoring of deposited particles, stabilize the magnetic pollution source after the magnetic levitation compressor is assembled, prevent secondary failures caused by drift in the early stage of service, and integrate environmental monitoring and assembly operation data to build a closed-loop feedback mechanism, dynamically correct assembly parameters, ensure continuous optimization of magnetic pollution source control during the assembly process of the magnetic levitation compressor, anchor deposited particles in advance, and achieve continuous process optimization.

[0110] Example 2, as Figures 1 to 11As shown, based on Example 1, the present invention provides a technical solution: S4 specifically includes: S4 specifically includes: establishing a magnetic field abnormal waveform feature library for the entire process of impurity release-migration-deposition, the magnetic field abnormal waveform feature library classifies and stores typical waveform templates according to impurity material, particle size range and deposition location, and continuously expands and updates it through actual assembly data, significantly improving the feature library's accuracy and adaptability in identifying typical deposition events, using a sliding time window to capture and monitor the magnetic field waveform in real time, and calculating its minimum regularization distance with each template in the feature library through a dynamic time warping algorithm, when the distance is lower than a set threshold, it is identified as a transient adsorption-deposition event, and the event type and spatial coordinates are output, achieving millisecond-level accurate identification and spatial positioning of deposition events, after identifying a deposition event, triggering a non-magnetic pulse purging device arranged in the assembly cavity to spray non-magnetic clean gas within a 30mm range around the deposition coordinates, while pausing the current assembly operation, and resuming assembly after the magnetic field waveform returns to the background level, effectively curbing the chain diffusion of magnetic pollution and ensuring the cleanliness of the magnetic field in the assembly environment;

[0111] It should be noted that the magnetic field anomaly waveform feature library is classified according to three attributes during the actual assembly process: different impurity materials (ferrite, martensite, austenite), particle size range (0.3μm to 10μm), and deposition location (bearing working surface, rotor air gap, sensor probe periphery). It stores typical waveform templates verified by actual measurements. Each template includes a time-domain waveform, frequency-domain energy distribution, and characteristic parameter vectors. The feature library is continuously expanded and updated using actual assembly data. After each assembly operation is completed, the waveform of the deposition event, which has been manually verified, is imported into the library. The confidence weights of corresponding templates are updated according to category, with the weight values ​​dynamically adjusted within the range of 0.6 to 1.0 based on the frequency of event recurrence. This ensures that the accuracy of the feature library in identifying typical deposition events continuously improves with the increase in assembly batches. A sliding time window is used to capture the monitored magnetic field waveform in real time, with a window length set to 2 seconds and a sliding step size of 0.5 seconds. The minimum warping distance between the current window waveform and each template in the feature library is calculated using a dynamic time warping algorithm. The warping path constraint adopts a Sako-Chiba band constraint, with a bandwidth set to 10% of the window length. When the minimum warping distance is reached... When the distance is below the set threshold of 0.15, a transient adsorption-deposition event is determined to have occurred, and the event type (classified according to the matching template) and spatial coordinates (positioning accuracy ±5mm) are output. For matching results with a confidence level below 0.85, the system automatically triggers resampling verification. The event is confirmed only after two consecutive matching results are consistent, avoiding misjudgment due to transient interference. The non-magnetic pulse purging device consists of 8 independently controllable nozzles, evenly distributed on the top and side walls of the cavity. The spray medium is high-efficiency filtered non-magnetic clean nitrogen gas, and the spray pressure is set to 0.3MPa. The purging duration is 0.5 seconds, effectively covering a 30mm radius around the deposition coordinates. Simultaneously, the central control system issues a command to pause the current assembly operation, and the operation terminal displays the status prompt "Deposition Event - Blocking". After the purging is completed, the magnetic field waveform is continuously monitored. Once the magnetic field strength within a 10mm radius around the deposition coordinates returns to the background level (≤0.05mT) and remains stable for more than 3 seconds, the pause status is automatically lifted, and the operation interface resumes the assembly operation permission. If the magnetic field does not return to the background level within 60 seconds after purging, the blocking is deemed to have failed, and the manual intervention process is triggered.

[0112] Furthermore, S4 also includes: when pulse purging is ineffective or deposition events occur three times in a row, automatically activating a local non-magnetic adsorption device at the bottom of the assembly cavity, using an electromagnetic field gradient to directionally attract magnetic particles in the deposition area to a dedicated collection tank, preventing impurities from escaping again, effectively preventing secondary diffusion of magnetic particles, ensuring the cleanliness of the magnetic field in the assembly cavity, simultaneously recording the tooling operation type, fastening torque, and friction pair material when the deposition event occurs, identifying high-risk operation nodes that induce deposition through correlation analysis, generating operation warning prompts and pushing them to the assembly personnel's terminal, achieving accurate warnings for high-risk operations, improving operational standardization, reducing the recurrence rate of deposition events, comparing the magnetic field recovery curve after the blocking operation with the historical normal curve to verify the blocking effectiveness, and using the verification results as the basis for updating the confidence weight of the feature library waveform template, improving the accuracy of subsequent matching and identification, realizing quantitative verification of the blocking effect, and continuously improving the feature library identification accuracy and the reliability of the blocking strategy;

[0113] It should be noted that when pulse purging is ineffective, or when the same deposition coordinates are repeatedly affected in three consecutive assembly operations, the central control system automatically activates a pre-set local non-magnetic adsorption device at the bottom of the assembly chamber. This device consists of three sets of independently controllable electromagnetic coils distributed below the chamber's bottom plate. Each set of coils has a maximum output magnetic field strength of 8mT and an effective working area diameter of 60mm. Based on the spatial coordinates of the deposition event, the nearest set of coils is automatically activated, applying a gradient-attenuated DC magnetic field to attract the magnetic particles in the deposition area along the magnetic field lines to a dedicated collection tank. The collection tank is made of austenitic stainless steel SUS316L, with an inner wall coated with... The PTFE non-stick coating facilitates regular cleaning. The magnetic field automatically dissipates after 5 seconds of adsorption. At this point, the sensor array remeasures the magnetic field strength within a 10mm radius of the deposition coordinates, confirming it has dropped below 0.05mT, effectively preventing impurities from escaping back into the assembly cavity. Tooling operation types are automatically identified based on the process codes input at the operating terminal, including twelve types of operations such as bolt tightening, bearing pressing, and rotor alignment. Tightening torque is collected in real-time by a non-magnetic torque sensor at a sampling frequency of 100Hz, recording the peak torque within 2 seconds prior to the event. The friction pair material is automatically matched by the process management system based on the current assembly components, including titanium alloy and non-magnetic electrical steel. Combinations such as austenitic stainless steel and engineering plastics are used. The above data are entered into the correlation analysis database. A frequent pattern mining algorithm is used to identify high-risk operation nodes that induce deposition events. When the frequency of deposition event association for a certain type of operation exceeds three times in five consecutive batches of assembly, an operation warning prompt is automatically generated and pushed to the assembly personnel's terminal. The prompt content includes the name of the high-risk operation, the suggested direction of operation parameter adjustment, and historical event statistics. After the blocking operation is completed, the magnetic field recovery curve within 10 mm of the deposition coordinates is compared with the historical normal recovery curves stored in the feature library. The comparison uses the root mean square error calculation method, and the current recovery curve is set to 0. Sampling is performed at 1-second intervals, and point-by-point differences are calculated between the sampled data and the corresponding time points of the historical normal curve. When the total error value is less than 0.02 mT, the blocking is considered effective. The verification results serve as the basis for updating the confidence weight of the waveform template in the feature library: for effective blocking matching results, the corresponding template confidence weight increases by 0.05, with an upper limit not exceeding 1.0; for ineffective blocking matching results, the corresponding template weight decreases by 0.03, with a lower limit not lower than 0.6. After the weight update, the complete waveform data, feature parameters, and blocking results of this deposition event are stored in the feature library for confidence calculation of matching and identification, ensuring that the accuracy of the feature library in identifying typical deposition events continues to improve with the increase of assembly batches.

[0114] In addition, the core assembly operation process covers component positioning and pre-assembly, rotor assembly and alignment, and seal assembly and fastening.

[0115] For component positioning and pre-assembly: Fix the non-magnetic tooling fixture on the operating platform, adjust the fixture position to ensure positioning accuracy ≤ ±0.002mm, wipe the fixture positioning surface again with a non-magnetic, dust-free cloth to remove residual particles; use non-magnetic tweezers, suction cups, and other tools to pick up and place components, avoiding direct contact with the core working surfaces of the components (such as bearing contact surfaces and rotor magnetic surfaces) by hand, handling them gently during the picking and placing process to prevent collisions and scratches; install the stator, bearings, and other fixed components on the tooling fixture and tighten them with non-magnetic bolts, with the tightening torque according to the process document requirements (generally 5N·m~8N·m), avoiding excessive torque that could cause component deformation;

[0116] For rotor assembly and alignment: Before rotor assembly, recheck the rotor residual magnetism to ensure it is ≤0.3mT. Wipe the rotor surface with a non-magnetic, dust-free cloth soaked in anhydrous ethanol to remove oil and particulate matter. Slowly place the rotor into the stator cavity and adjust its position to ensure that the coaxiality between the rotor and stator is ≤0.003mm. Use a non-magnetic dial indicator to monitor the rotor runout, which should be ≤0.002mm / 100mm. During rotor alignment, avoid friction between the rotor and the stator or bearings. If position adjustment is necessary, use non-magnetic tools to operate slowly, and avoid violent striking.

[0117] For seal assembly and fastening: Seals (non-magnetic fluororubber sealing rings, non-magnetic metal sealing gaskets) must be assembled in a low-temperature environment (consistent with the temperature of the assembly cavity) to avoid deformation caused by temperature changes; use non-magnetic tools to install the seals in the designated positions, ensuring that the seals are not twisted or damaged. After assembly, check the sealing gap, which should be ≤0.005mm to prevent air and oil leaks; use non-magnetic bolts to tighten each connection part in sequence, following the "diagonal symmetry" principle to avoid uneven force distribution. After tightening, check the positional accuracy and coaxiality of the components again to ensure they meet the requirements.

[0118] S5 specifically includes: establishing a three-dimensional finite element model of the core assembly of the magnetic levitation compressor using finite element simulation; importing the residual stress field distribution data and geomagnetic field boundary conditions after assembly; simulating the magnetic creep displacement behavior of magnetic particles during stress relaxation using a magnetic-mechanical-thermal multiphysics coupling method to ensure the simulation accuracy of particle migration path and the consistency of physical process; defining the initial position, particle size, magnetic susceptibility, and shape factor of particles as probability distribution parameters; performing more than 1000 random sampling calculations using the Monte Carlo method to statistically analyze the cumulative probability of particles migrating to the sensor probe area of ​​the magnetic levitation bearing and the sensitive position of the air gap, achieving accurate quantification of particle migration probability and reliability of risk assessment; generating a secondary magnetic contamination risk map of the assembly based on the migration probability distribution; marking the spatial coordinates of high-risk areas; providing a quantitative basis for the selection of subsequent temperature-magnetic composite field application areas; and providing accurate spatial positioning and reliable coverage guarantee for composite field treatment.

[0119] It should be noted that in the finite element simulation stage, a three-dimensional finite element model of the core assembly of the magnetic levitation compressor was first established. The geometric dimensions of the model were constructed based on the actual component drawings. The mesh was generated using tetrahedral elements, with the element size controlled between 0.5mm and 2mm. The mesh in key areas (bearing working surface, air gap, and sensor probe perimeter) was locally refined to 0.2mm. The residual stress field distribution data after assembly was imported through process simulation software. The geomagnetic field boundary condition was set to 0.05mT, with the direction along the geographic North Pole. The magnetic-mechanical-thermal multiphysics coupling method was used for the solution, where magnetic particles were defined as having a diameter of 5μm to 30μm. The particles were spherical with shape factors set to equiaxed. During the simulation, the time step was set to 0.1 seconds, and the total simulation duration covered 720 hours after assembly to capture the magnetic creep displacement behavior of the particles during stress relaxation. The simulation output data on the spatial coordinates and magnetic moment direction changes of the particles at each time point. In the Monte Carlo sampling calculation stage, the initial position of the particles was set to typical areas where residual magnetic impurities might be distributed after assembly, including the edge of the sealing groove, the bolt connection surface, and the end of the stator winding. The initial position coordinates were randomly generated according to a uniform distribution. The particle size distribution was set according to a log-normal distribution with a mean of 15 μm and a standard deviation of 5 μm. The magnetic susceptibility was set according to a Gaussian distribution with a mean of 0. 45, standard deviation 0.15; shape factor set to a constant 1.0 (equiaxed); sampling number set to 2000 times; each sampling independently performs magnetic-mechanical-thermal coupling simulation, counting the number of times particles migrate to the sensor probe area (defined as within 5mm of the sensor sensitive surface) and the air gap sensitive position (defined as the middle 1 / 3 region of the radial gap between the rotor and stator) within the simulation time, calculating the cumulative probability; the calculation process runs on a parallel computing platform, and the calculation time for a single batch of sampling is approximately 48 hours, ensuring the convergence and stability of the statistical results; the secondary magnetic contamination risk map of the assembly is based on the three-dimensional assembly model, using a color gradient method. High-risk areas are marked in a specific way: areas with a migration probability greater than 10% are marked as red high-risk areas, 5% to 10% are marked as orange medium-risk areas, 1% to 5% are marked as yellow low-risk areas, and less than 1% are marked as green safe areas. Spatial coordinates are output with the assembly reference zero point as the origin, and the coordinates of high-risk areas are accurate to ±0.5mm. This risk map is directly used as the basis for selecting the subsequent temperature-magnetic composite field application area. The central control system automatically plans the action positions of the local heater and multi-pole electromagnetic coil in the temperature-magnetic composite processing station based on the spatial coordinates of the high-risk areas in the map, ensuring that the composite field application range completely covers the high-risk areas and extends the edge by 5mm.

[0120] In addition, post-assembly cleaning and inspection includes post-assembly cleaning, quality inspection, and handling of non-conforming products;

[0121] For post-assembly cleaning: After assembly, use a non-magnetic, dust-free cloth dampened with anhydrous ethanol to wipe the surface of the assembled parts to remove particulate matter and oil stains generated during the assembly process, clean debris and waste materials inside the assembly cavity, and put tools and fixtures back in place to ensure the cavity is clean.

[0122] For quality inspection: Residual magnetism inspection: A non-magnetic magnetometer is used to measure the overall residual magnetism of the assembled components, ensuring it is ≤0.5mT. Special attention is paid to core components such as the rotor, bearings, and stator, with local residual magnetism ≤0.3mT. Dimensional accuracy inspection: Non-magnetic micrometers and dial indicators are used to measure assembly dimensions, such as rotor-stator coaxiality, sealing gaps, and component spacing, with an error ≤±0.005mm. Appearance inspection: Visual inspection is conducted to ensure the components are free of scratches, dents, oil stains, and particulate residue. Seals are undamaged and untwisted, and connections are secure. Environmental parameter verification: Environmental parameters of the assembly cavity are checked to ensure cleanliness, temperature, humidity, and magnetic field strength still meet requirements, and the test data is recorded.

[0123] For handling non-conforming products: If a non-conforming product is found during inspection, it should be immediately placed in a non-magnetic isolation box, marked with a "non-conforming" label, removed from the assembly cavity, and the cause of non-conformity should be analyzed (such as magnetic contamination, dimensional deviation, component damage, etc.). Targeted corrective measures should be taken (such as demagnetization, reassembly, replacement of components). After correction, the product should be re-inspected. Only after passing the inspection can it proceed to the next process.

[0124] S6 specifically includes: Based on the identified high-risk areas, after assembly, the assembly is placed in the temperature-magnetic composite treatment station of the three-condition assembly equipment. A local heater and a multi-pole electromagnetic coil are used in synergistic control to apply an alternating composite field with a temperature of 5°C to 10°C and a magnetic field strength of 0.1mT to 0.5mT to the risk area. Through the synergistic effect of temperature and magnetism, the deposited particles are actively stabilized, eliminating the risk of migration in the early stage of service. During the application of the alternating composite field, the magnetic moment orientation change of the deposited particles is monitored in real time by an embedded magnetic field probe. The frequency and phase of the alternating magnetic field are dynamically adjusted by a closed-loop control algorithm to promote the orientation of the particles under the synergistic effect of temperature and magnetism and anchor them in non-sensitive areas. This ensures that the orientation of the particles is accurate and controllable, improves the consistency of anchoring and long-term stability. After the treatment, the residual magnetism of the high-risk area is retested by a high-precision magnetometer array to confirm that the magnetic moment of the deposited particles has been stabilized and has not migrated to the sensor area. This forms a record of the stabilization treatment of the magnetic pollution source after assembly, verifies the treatment effect, and provides reliable data support for quality traceability.

[0125] It should be noted that in the magnetic pollution source stabilization treatment stage after assembly, the assembly is first transferred to the temperature-magnetic composite treatment station built into the three-condition assembly equipment based on the spatial coordinates of the high-risk areas in the generated secondary magnetic pollution risk map of the assembly. This station is equipped with four independently controlled local heaters and six multi-pole electromagnetic coils. The local heaters are made of polyimide film and are attached to the outer wall of the high-risk area, with a temperature control accuracy of ±0.3℃. The multi-pole electromagnetic coils are Helmholtz opposed structures, with a maximum output magnetic field strength of 0.5mT per group, and an effective range covering the high-risk area and extending 5mm. The central control system is based on wind... The system automatically plans the activation combination of local heaters and multi-pole electromagnetic coils in the high-risk area, setting the temperature to 8℃±0.5℃, the magnetic field strength to 0.3mT±0.05mT, the initial alternating frequency to 0.5Hz, and the initial phase angle to 0°, ensuring that the combined field application range and intensity completely cover all high-risk areas. A triaxial fluxgate probe embedded around the high-risk area monitors the magnetic moment orientation changes of the deposited particles in real time. The probe sampling frequency is 20Hz, and the spatial resolution is 2mm. The closed-loop control algorithm uses the deviation between the particle magnetic moment direction and the magnetic field direction of the target anchoring area as the control input, employing a proportional-integral-... The frequency and phase of the alternating magnetic field are dynamically adjusted using a differential adjustment method. The frequency adjustment range is 0.3Hz to 0.8Hz, and the phase adjustment range is -30° to +30°, with step sizes of 0.05Hz and 5°, respectively. When the magnetic moment direction deviation remains within ±5° for 30 consecutive seconds, it is determined that the particles have completed directional alignment. At this time, the magnetic field output mode is switched to DC holding mode for 120 seconds, causing the particles to anchor in a non-sensitive area under the synergistic effect of the thermal and magnetic fields, preventing re-migration during subsequent service. A high-precision magnetometer array consisting of 12 single-axis magnetometers is used to re-measure the residual magnetism in high-risk areas. The measurement range is ±1mT, the measurement accuracy is ±0.005mT, the array is arranged with 5mm×5mm grid points, and residual magnetism data is collected point by point. The retest results are compared with the background residual magnetism value in the risk spectrum before treatment. It is required that the residual magnetism increment at any measuring point in the high-risk area does not exceed 0.02mT, and there are no new residual magnetism anomalies within 5mm around the sensor probe. After meeting the above indicators, the magnetic pollution source stabilization treatment record after assembly is automatically generated, including the treatment time, applied parameters, magnetic moment change curve and retest data. As an important part of the assembly quality traceability document of this batch, it ensures the magnetic field stability and reliability of the assembly in the early stage of service.

[0126] Furthermore, S6 also includes: using the environmental monitoring system to collect temperature, humidity, cleanliness, and magnetic field strength data, and aligning them with the tooling usage records and fastening torque data of the assembly operating system on a time axis to construct a digital twin dataset of the entire assembly process. This provides a high-precision and highly consistent data foundation for the correlation analysis between assembly parameters and magnetic contamination risks. It employs association rule mining algorithms to analyze the correlation between assembly parameters and the identification results of secondary magnetic contamination risk points, extracts key influencing factors, generates dynamic correction suggestions for assembly parameters, and pushes them to the process management platform. This enables the transformation of assembly parameters from experience-based setting to data-driven precise correction. The corrected assembly parameters are used as the initial input for the next batch of assembly. Through inter-batch iterative optimization, thermal cycling parameters, demagnetization thresholds, and blocking trigger conditions are continuously updated, forming an adaptive optimization closed loop for magnetic contamination source control strategies. This creates an intelligent optimization mechanism where control strategies automatically evolve with each batch and their effects continuously converge.

[0127] It should be noted that the environmental monitoring system continuously records temperature, relative humidity, cleanliness, and magnetic field strength data within the assembly cavity at a sampling frequency of once per second. Temperature sensors are deployed at four measuring points at the top and bottom of the cavity. The humidity sensor and particulate counter operate synchronously. Magnetic field strength data is acquired by an eight-channel fluxgate sensor array. The assembly operating system synchronously records tooling usage records, including the input time of process codes for twelve types of operations, the peak value of tightening torque collected by the non-magnetic torque sensor at a sampling frequency of 100Hz, and friction pair material information. The central control system aligns the above multi-source data with a time axis at 0.1-second intervals using a unified time base. All data is stored in the assembly process database after cleaning and normalization. A digital twin dataset containing time-series curves of environmental parameters, timestamps of operational events, torque loading waveforms, and magnetic field disturbance records of the entire assembly process was generated. The FP-Growth frequent pattern mining algorithm was used to analyze the digital twin dataset, setting a minimum support of 0.05 and a minimum confidence of 0.75. The association rules between assembly parameters and the secondary magnetic contamination risk point identification results were mined. The risk point identification results were derived from the generated secondary magnetic contamination risk map of the assembly, characterized by high-risk area coordinates and migration probabilities. The mining results showed that when the peak tightening torque during rotor alignment exceeds 7.2 N·m and the friction pair is a titanium alloy-non-magnetic electrical steel combination, the probability of secondary magnetic contamination around the sensor probe increases to 0. 82; When the bearing component's low-temperature holding time during thermal cycling is less than 12 minutes, the migration probability of the sealing groove edge area increases by 0.15. Based on the above association rules, the process management platform automatically generates dynamic correction suggestions for assembly parameters: adjusting the upper limit of the target torque for rotor alignment from 8.0 N·m to 7.0 N·m, extending the bearing thermal cycling holding time to 15 minutes, and narrowing the stator thermal cycling alternation range from ±4.0℃ to ±3.8℃. These correction suggestions are pushed to the process engineer's review terminal in a structured form. After approval, the assembly process parameter library is automatically updated. Before batch startup, the process execution system automatically loads the updated thermal cycling parameters, demagnetization threshold, and blocking trigger conditions. The thermal cycling parameters are based on the distribution characteristics of abnormal points in the previous batch. The initial population distribution range was dynamically adjusted. The rotor thermal cycling amplitude was locked at ±4.5℃, the frequency was 4 times per hour, and the holding time was 15 minutes. The bearing holding time was extended from 10 minutes to 15 minutes. The demagnetization threshold was adjusted from 0.02 mT to 0.015 mT. The dynamic time warping distance threshold in the blocking trigger condition was tightened from 0.15 to 0.12. After every five assembly batches, association rule mining was re-executed, and the data of the newly added batches were included in the analysis sample set. The weights of key influencing factors and the confidence of the correction suggestions were updated. After twelve batches of iterative optimization, the occurrence frequency of secondary magnetic contamination risk points around the sensor probe decreased from 18% in the initial batch to 3.5%, and the average residual magnetism increment in the high-risk area decreased from 0.035 mT to 0.0.12mT, forming an adaptive optimization closed loop where thermal cycling parameters, demagnetization threshold, and blocking conditions continuously converge, the pertinence and effectiveness of the magnetic pollution source control strategy steadily improves with the increase of assembly batches;

[0128] It should also be noted that during the assembly process, after each step is completed, the operator must fill out the "Non-magnetic Assembly Process Record Form," recording environmental parameters, operation content, component number, test results, and other information, and sign for confirmation. The quality inspector will conduct a patrol inspection of the assembly process every 2 hours, checking environmental parameters, operational standardization, component quality, etc., and promptly stop and require rectification of any problems found, recording the inspection results. The three-condition assembly equipment must undergo a pre-start inspection daily, maintenance weekly, and a comprehensive calibration monthly (environmental parameters, magnetic field shielding effect, positioning accuracy, etc.), and the results must be recorded in the "Equipment Operation and Maintenance Record Form." Furthermore, after the core components of each magnetic levitation compressor are assembled, a unique traceability code must be assigned, linking the component number, assembly date, operator, environmental parameters, test results, equipment operation data, and other information. The traceability code must be affixed to a conspicuous location on the component to ensure traceability throughout its entire lifecycle.

[0129] Operators must strictly abide by the three-condition assembly equipment operation procedures and are strictly prohibited from unauthorized operations (such as forcibly opening interlock doors, arbitrarily adjusting environmental parameters, etc.) to prevent equipment failure or personal injury; smoking and open flames are strictly prohibited in the assembly area, and dry powder fire extinguishers (non-magnetic type) should be provided. Avoid prolonged contact with the cavity wall in low-temperature environments to prevent frostbite; in case of emergencies such as excessive magnetic field or purification system failure, immediately stop the assembly operation, cut off the equipment power supply, evacuate personnel, handle the situation according to the "Emergency Response Plan", and report to the relevant departments;

[0130] Waste materials such as lint-free cloths and ethanol waste generated during assembly must be sorted and placed in non-magnetic environmentally friendly recycling containers and handed over to professional organizations for disposal. Random disposal is strictly prohibited. During the operation of the three-condition assembly equipment, the refrigeration system and purification system must be checked regularly to avoid refrigerant leakage and excessive exhaust emissions. Energy conservation and consumption reduction should be achieved by reasonably adjusting the equipment operating parameters to reduce energy consumption.

[0131] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An optimized method for controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor, characterized in that, Includes the following steps: S1. In the pre-assembly treatment stage of the magnetic levitation compressor, an adaptive genetic algorithm is used to dynamically optimize the thermal cycling parameters under low temperature conditions, and to accurately induce the early release of ferromagnetic impurities hidden in the micro-defects of compressor components. S2. Combining wavelet packet transform and neural network algorithm, feature extraction and pattern recognition are performed on the released impurity signal to guide the non-magnetic shielding system to perform targeted pulse demagnetization and complete the removal of ferromagnetic impurities before the magnetic levitation compressor is assembled. S3. Construct a multi-sensor data fusion model based on Kalman filtering to separate the instantaneous magnetic field interference generated by friction of non-magnetic tooling from the background magnetic field in real time, and accurately restore the migration trajectory of impurities during the assembly of the magnetic levitation compressor. S4. Introduce a dynamic time warping algorithm to match the real-time monitored magnetic field anomaly waveform with a preset impurity release feature library to locate transient adsorption-deposition events and automatically block them. S5. Finite element simulation is used to pre-simulate the magnetic creep path of magnetic particles under the coupling of residual stress and geomagnetism, and Monte Carlo sampling method is combined to evaluate the particle displacement probability and identify secondary magnetic pollution risk points. S6. Apply a controllable temperature-magnetic composite field to induce the deposition particles to anchor in advance, thereby stabilizing the magnetic pollution source after the magnetic levitation compressor is assembled. Integrate environmental monitoring and assembly operation data to build a closed-loop feedback mechanism and dynamically correct assembly parameters.

2. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S1 specifically includes: A thermal cycling process parameter model was constructed with temperature alternation amplitude, alternation frequency and low temperature holding time as decision variables. The dual objective optimization function was set as maximizing the impurity release rate and minimizing the component remanent magnetization increment. A penalty function was introduced to handle the constraint condition of temperature change rate on the microstructure of the magnetic levitation compressor component. The thermal cycling parameter population was initialized using real-number encoding. By adaptively adjusting the crossover and mutation probabilities and combining Pareto non-dominated sorting, the optimal parameter combination for inducing the release of hidden ferromagnetic impurities in microcracks under low-temperature conditions was selected. The optimized thermal cycling parameters are input into the low-temperature constant temperature system of the three-condition assembly equipment to perform thermal cycling treatment, which precisely induces the extrusion of nanoscale ferromagnetic impurities in micro-defects during the material shrinkage-expansion process.

3. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 2, characterized in that: S1 further includes: During the thermal cycling process, the surface temperature field distribution data and local remanent magnetization change data of the component are collected in real time, and a temperature-remanent magnetization dynamic response surface is constructed. This surface is then fed back into the fitness function of the adaptive genetic algorithm to realize the online correction of thermal cycling parameters in the pre-assembly processing stage. For the three core components of the compressor—rotor, stator, and bearing—sub-models of material thermophysical parameters are established respectively. Component-specific weighting coefficients are introduced into the genetic algorithm to generate component-specific thermal cycling process curves. After the thermal cycle is completed, the residual magnetism of the component surface is scanned by a high-precision magnetic field imager to identify micro-area residual magnetism anomalies left after the release of impurities. The scan results are used as input for the next round of assembly batch thermal cycle parameter optimization, forming a closed loop of inter-batch iterative optimization.

4. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S2 specifically includes: A multi-channel high-bandwidth magnetic field sensor array is used to collect transient magnetic field fluctuation signals generated by impurity release during thermal cycling. The signal is decomposed into three levels by wavelet packet transform, and the energy entropy, singular values ​​and peak frequencies of each frequency band are extracted as feature vectors. A classifier for impurity release patterns based on a backpropagation neural network is constructed. The feature vector is used as input, and the impurity type, release location and magnetization intensity are used as output. The mapping relationship between the release signal and the physical properties of the impurities is established to identify high-risk impurities that need to be removed first. The identification results are transmitted in real time to the control module of the non-magnetic shielding system, which drives the coil-type non-magnetic shielding system to apply a targeted pulse demagnetizing magnetic field with adjustable pulse width and progressively decreasing amplitude to the identified release area.

5. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S3 specifically includes: A triaxial fluxgate sensor, a vibration accelerometer, and a particulate matter counter are installed in the assembly cavity to establish a spatiotemporal synchronous acquisition mechanism for the sensor array, and to synchronously acquire magnetic field vector, tooling vibration characteristics, and particulate matter concentration time series data. A multi-sensor data fusion model is constructed based on extended Kalman filter. The instantaneous magnetic field interference generated by friction of non-magnetic tooling is defined as time-varying observation noise. Through two-step iteration of state prediction and observation update, the background magnetic field of the environment and the dynamic friction interference magnetic field are separated in real time. The separated net magnetic field signal is spatially interpolated and fused with the synchronously acquired particulate matter concentration data to generate a four-dimensional spatiotemporal distribution map of the impurity migration trajectory, accurately reconstructing the path and speed of impurities migrating from the release location to the bearing working surface or air gap.

6. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S4 specifically includes: Establish a magnetic field anomaly waveform feature library for the entire process of impurity release, migration and deposition. The magnetic field anomaly waveform feature library stores typical waveform templates according to impurity material, particle size range and deposition location, and is continuously expanded and updated through actual assembly data. The waveform of the monitored magnetic field is captured in real time using a sliding time window. The minimum warping distance between the waveform and each template in the feature library is calculated using a dynamic time warping algorithm. When the distance is lower than a set threshold, it is identified as a transient adsorption-deposition event, and the event type and spatial coordinates are output. Upon detecting a deposition event, the non-magnetic pulse purging device arranged in the assembly chamber is triggered to spray non-magnetic clean gas within a 30mm radius around the deposition coordinates. At the same time, the current assembly operation is paused, and assembly is resumed after the magnetic field waveform returns to the background level.

7. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 6, characterized in that: S4 further includes: When pulse purging fails to block or deposition events occur three times in a row, the local non-magnetic adsorption device at the bottom of the assembly cavity is automatically activated. The magnetic particles in the deposition area are attracted to the special collection tank by the electromagnetic field gradient, preventing the impurities from escaping again. Simultaneously record the tooling operation type, fastening torque and friction pair material when deposition events occur, identify high-risk operation nodes that induce deposition through correlation analysis, generate operation warning prompts and push them to the assembly personnel's terminal; The magnetic field recovery curve after the blocking operation is compared with the historical normal curve to verify the effectiveness of the blocking, and the verification result is used as the basis for updating the confidence weight of the waveform template in the feature library.

8. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S5 specifically includes: A three-dimensional finite element model of the core assembly of the magnetic levitation compressor was established using finite element simulation. The residual stress field distribution data after assembly and the boundary conditions of the geomagnetic field environment were imported. The magnetic-mechanical-thermal multiphysics coupling method was used to simulate the magnetic creep displacement behavior of magnetic particles during the stress relaxation process. The initial position, particle size, magnetic susceptibility, and shape factor of the particles are defined as probability distribution parameters. More than 1,000 random sampling calculations are performed using the Monte Carlo method to calculate the cumulative probability of particles migrating to the probe area of ​​the magnetic levitation bearing sensor and the sensitive position of the air gap. A secondary magnetic contamination risk map of the assembly is generated based on the migration probability distribution, and the spatial coordinates of high-risk areas are marked.

9. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 1, characterized in that: S6 specifically includes: Based on the identified high-risk areas, after assembly, the assembly is placed in the temperature-magnetic composite processing station of the three-condition assembly equipment. The local heater and multi-pole electromagnetic coil are used in coordination to apply an alternating composite field with a temperature of 5°C to 10°C and a magnetic field strength of 0.1mT to 0.5mT to the risk area. During the application of the alternating composite field, the magnetic moment orientation change of the deposited particles is monitored in real time by an embedded magnetic field probe. The frequency and phase of the alternating magnetic field are dynamically adjusted by a closed-loop control algorithm, which causes the particles to align and anchor in non-sensitive areas under the synergistic effect of thermo-magnetic interaction. After the treatment was completed, the residual magnetism of the high-risk area was re-measured using a high-precision magnetometer array to confirm that the magnetic moment of the deposited particles had stabilized and had not migrated to the sensor area, thus forming a record of the stabilization treatment of the magnetic pollution source after assembly.

10. The optimized method for preventing and controlling magnetic pollution sources during the assembly process of a magnetic levitation compressor according to claim 9, characterized in that: S6 further includes: By aligning the data on temperature, humidity, cleanliness, and magnetic field strength collected by the environmental monitoring system with the tooling usage records and fastening torque data of the assembly operating system along the timeline, a digital twin dataset of the entire assembly process is constructed. The association rule mining algorithm was used to analyze the correlation between assembly parameters and the identification results of secondary magnetic contamination risk points, extract key influencing factors, generate dynamic correction suggestions for assembly parameters, and push them to the process management platform. The revised assembly parameters are used as the initial input for the next batch of assembly. Through inter-batch iterative optimization, the thermal cycling parameters, demagnetization threshold, and blocking trigger conditions are continuously updated, forming an adaptive optimization closed loop for the magnetic pollution source control strategy.

Images