86results about How to "Effective alignment" patented technology

The invention provides a robust multi-modal image segmentation method and system based on instance perception query, and the method comprises the steps: obtaining original image data and text description data, and generating training data through semantic analysis and causal graph construction; extracting image features and text features by adopting a multi-layer coding network, and establishing a feature dependency relationship through causal reasoning; generating instance clustering features based on attention calculation, and obtaining fusion features in combination with a multi-scale fusion strategy; and calculating query confidence and generating candidate segmentation masks, and outputting a final segmentation result through post-processing optimization. According to the method, by introducing causal reasoning and instance awareness query mechanisms, the cross-modal feature alignment precision and the robustness of the segmentation result are improved, and the method has better segmentation performance and generalization ability in a complex scene.

The invention relates to the technical field of law enforcement intelligent alarm, in particular to a multi-mode intelligent analysis method and system based on law enforcement supervision. The method comprises the following steps: acquiring a law enforcement scene video file from a law enforcement supervision recorder platform, and performing audio and video segmentation and law enforcement behavior information identification analysis to obtain law enforcement scene audio behavior information data and law enforcement scene image behavior information data; performing timestamp and spatial position extraction processing on the law enforcement scene video file, performing multi-modal time-space synchronization association analysis, and performing law enforcement behavior sequence connection processing and event behavior intention recognition analysis at the same time to obtain law enforcement scene event behavior intention information data; and obtaining a preset law enforcement event behavior knowledge graph, and carrying out behavior normative score calculation and nonstandard behavior intelligent alarm optimization analysis to generate a law enforcement scene event nonstandard behavior optimization suggestion report. According to the invention, the accuracy of law enforcement nonstandard behavior supervision alarm can be improved.

The invention discloses a large-model-driven Web task automatic execution method and system, and relates to the field of artificial intelligence. According to the method, Web page content is perceived and analyzed through a large model, a state-action graph is constructed and maintained, a task sequence library is stored and managed, and semantic matching and automatic task execution are carried out. According to the method, the problems of high manual intervention requirement and low task processing efficiency in traditional Web task automation are solved, and the technical effects of remarkably improving the automation level, enhancing the intelligent decision-making capability, perfecting a knowledge memory mechanism and efficiently executing are achieved.

The invention discloses an intelligent eye-protection illumination sitting posture correction method and system, particularly relates to the technical field of illumination sitting posture correction, and is used for solving the problem of asynchronous illumination sitting posture analysis. According to the method, data from different sensors are ensured to be synchronized on the same time reference through a dynamic time warping algorithm, errors caused by asynchronization are eliminated, a self-attention mechanism is combined with position coding, the complex relation and time sequence dependence of the data are accurately extracted, the feature representation capability of the model is improved, and finally, the model is optimized. Time sequence data fusion is carried out on the multi-modal features according to the long-short-term memory network, the user state and demand are predicted in real time, accurate light adjustment and sitting posture monitoring are achieved, the robustness and adaptability of data analysis are improved, the user experience and health management efficiency can be remarkably improved in intelligent eye-protection illumination sitting posture correction, and the user experience is improved. And a more comfortable and healthy working and living environment is provided for the user.

The invention relates to an edge AI collaboration method and system based on cloud big model enhancement. The method comprises the steps of analyzing a natural language request through a natural language processing technology and converting the natural language request into sub-tasks; the processing capacity of each edge device is evaluated, and the subtasks are unloaded to the corresponding edge devices in a coordinated mode; edge AI model training is carried out on each edge device in a distributed mode; and performing synchronization mechanism and consistency constraint on each preliminary edge AI model by using a cross-device consistency maintenance algorithm, and aligning model parameters of the preliminary edge AI models among edge devices to complete edge AI collaboration. A natural language processing technology is utilized to deeply analyze a natural language request of a user, the natural language request is accurately converted into a sub-task which can be executed by an edge device, real-time dynamic allocation is carried out, the resource utilization rate can be optimized, and task execution delay is reduced; by performing synchronization mechanism and consistency constraint on the model, the overall performance and generalization ability of the model can be improved.

The invention discloses a cross-modal feature alignment and fusion method, and belongs to the field of multi-modal recommendation. The method comprises the following steps: firstly, capturing a semantic relationship between projects in the same modal by constructing a project semantic graph, and performing graph convolution operation on overall features and modal specific features on the basis to enhance the expression ability of the overall features and the modal specific features in the modal; potential information between features is captured and utilized, and the problem of incomplete matching between modes is relieved. And then, comparing the enhanced overall features with modal features in each modal by using a comparative learning method, and promoting cross-modal information transmission and complementation. And through conjoint analysis of project semantic graphs of different modalities, effective modeling of a complex relationship between the modalities is realized. According to the method, cross-modal semantic consistency is ensured, so that redundancy and noise in a modal fusion process are reduced, and the accuracy of multi-modal recommendation is improved.

The invention discloses a dialogue emotion recognition method based on context awareness and cross-modal attention sharing, and the method comprises the steps: collecting texts containing contexts and audio information in different dialogue scenes to form a data set, and completing the construction of the data set; preprocessing the constructed data set to extract corresponding modal features; the method comprises the following steps: respectively establishing an audio context sensing network, a cross-modal sharing attention network and an adaptive fusion network, and sequentially connecting the established networks to form an initial emotion recognition model; performing end-to-end training on the initial emotion recognition model, and storing the trained emotion classification model to form an emotion recognition model; performing feature extraction on the collected text and audio data by adopting a feature encoder to obtain text and audio features; and inputting the text and audio features into an emotion recognition model, and recognizing and outputting a predefined condition category by using the emotion recognition model. The method can effectively solve the problem of spatial heterogeneity, and has an excellent recognition effect.

The invention discloses a facial action unit identification method based on CLIP and multi-mode mask prompt learning. The method comprises the following steps: 1, carrying out facial action unit identification on the basis of the CLIP and the multi-mode mask prompt learning; the method comprises the steps that 1, a multi-modal shared AU prompt (AU-prompt) and an attention mask are designed based on an AU prior position, the AU-prompt comprises a plurality of learnable tokens, the attention mask is designed according to prior positions of different AU facial muscle changes, and a specific AU attention mask is obtained; 2, in a multi-mode prompt learning process, introducing an attention mask of a specific AU into different layers of a CLIP model text and an image encoder for prompt learning, and learning local detail features of the AU and global association features among different AUs to obtain local features and global features; and step 3, fusing the local features and the global features, calculating loss with the text features, and finally realizing AU identification. According to the method, the facial action unit can be effectively recognized under the condition that the labeled data is limited.

The invention relates to a text attention-based cross-modal representation model construction method and a text attention-based cross-modal representation model retrieval method. Two highly innovative modules, namely an attention guidance-based mask modeling module and a text enhancement module, are added in an online model. Wherein the attention-guided mask modeling module takes the attention in a text coding forward process as a reference and performs self-adaptive masking on words in text description, so that the mask language modeling module can capture semantics associated with masking words for alignment in a masking word prediction process; and therefore, the cross-modal characterization can be aligned more efficiently. The text enhancement module expands text description in a downstream task data set in a lightweight mode, the problems that the text description in the data set is single in word and errors exist are solved, and the retrieval accuracy is improved.

The invention relates to the technical field of material analysis, in particular to a multi-mode collaborative corrosion state monitoring method and system. Through a self-supervised contrast learning method of acoustic emission and electrochemical noise dual-mode signals and in combination with an innovative network architecture of Transformer and a multi-layer perceptron, feature alignment of acoustic emission and electrochemical noise signals is realized, the problem of dependence on a large amount of labeled data in corrosion monitoring of chlor-alkali production equipment is overcome, and a corrosion state on-line monitoring model is formed. Under a self-supervised contrast learning framework, the model can utilize unlabeled acoustic emission and electrochemical noise data to carry out autonomous learning, and the demand for expensive and scarce labeled resources is reduced. Only a small amount of manual annotation acoustic emission data is needed for fine tuning training, and online monitoring of the corrosion state can be achieved. Meanwhile, the on-site rapid and non-destructive corrosion state monitoring analysis of the chlor-alkali production equipment can be completed only through acoustic emission data, and the corrosion state monitoring technology is promoted to develop towards the more intelligent and automatic direction.

The invention discloses a tracking type photovoltaic power generation device which comprises a base, a rotating platform is rotatably arranged on the upper end face of the base, a driving mechanism used for driving the rotating platform to rotate is fixedly arranged at the bottom of the base, a vertical plate is fixedly arranged on the upper end face of the rotating platform, and a photovoltaic power generation panel is hinged to the upper end of the vertical plate. An angle adjusting mechanism used for driving the photovoltaic power generation panel to adjust the elevation angle is connected between the side face of the vertical plate and the bottom face of the photovoltaic power generation panel, solar sensors are fixedly arranged at the four corners of the photovoltaic power generation panel correspondingly, and sweeping assemblies used for removing dust on the surfaces of the solar sensors are fixedly arranged on the left side edge and the right side edge of the photovoltaic power generation panel correspondingly. According to the tracking type photovoltaic power generation device, through the automatic cleaning assembly and the intelligent control algorithm module, the requirement for manual intervention is reduced, manual regular cleaning of the solar sensor is not needed, and meanwhile, the tracking type photovoltaic power generation device can still stably operate in various complex environments, so that the tracking type photovoltaic power generation device has good practicability in different geographical environments and weather conditions.

The invention belongs to the technical field of federal learning, and particularly relates to a federal consistency representation learning system and method based on prototype guidance. The system comprises two main modules: an in-source characterization calibration module and a cross-source consistent characterization learning module. The method comprises the following steps: firstly, improving local training by using an in-source representation correction module, and correcting feature distribution on unbalanced data; this helps mitigate significant differences in the feature space due to dataset deviations. Meanwhile, it provides prototype information to a server, including cluster prototypes, cluster variances, and attention scores. The cross-source consistent representation learning module then learns generalized projections and classifiers using prototype information obtained from all customers. According to the algorithm, firstly, statistical knowledge is utilized to generate extended features, feature space is refined, and diversity is improved; then, the features of different sources are mapped to a unified space for comparison and classification, and interference of isolated points is eliminated according to attention scores;

The present invention relates to the technical field of motor stator wire embedding machines, and discloses a motor stator horizontal wire embedding machine, comprising a wire embedding machine body, two clamping arms are arranged on the right side of the wire embedding machine body, a wire embedding module located between the two clamping arms is arranged on the right side of the wire embedding machine body, two support frames are arranged on the right side of the wire embedding machine body, the tops of the two support frames are both provided with slide rails, the bottoms of the wire embedding machine body and the support frames are both provided with support seats, and a support mechanism slidably connected to the slide rails is arranged between the tops of the two support frames. The motor stator horizontal wire embedding machine effectively facilitates users to place the motor stator and adjust the feeding, effectively reduces the labor intensity of operators, effectively helps users to effectively position and align the motor stator, effectively reduces the misalignment of the wire embedding head, ensures the wire embedding quality, and is convenient for users to use.

The invention discloses a method and device for adjusting the posture of a vehicle in a battery swap station, the battery swap station and the vehicle, and belongs to the technical field of battery charging and swap. The method comprises the following steps: controlling a passing platform of the battery swap station to rise to a first preset height, so that a plurality of probes arranged on the passing platform are in contact with a plurality of probe detection surfaces at the bottom of a vehicle; and adjusting the posture of the vehicle according to the measured value measured by each probe and a pre-calibrated calibration value. According to the embodiment of the invention, the plurality of probes are matched with the plurality of probe detection surfaces at the bottom of the vehicle, so that the measurement data of the plurality of contact points between the probes and the probe detection surfaces are sensed, the levelness deviation of the vehicle can be accurately identified by comparing the measurement values measured by the probes with the pre-calibrated calibration values, and the attitude of the vehicle is adjusted according to the levelness deviation. Therefore, the vehicle attitude levelness is improved, and the battery replacement success rate is improved.

The invention discloses a feeding mechanism of an automatic cold header for machining piston pins. The feeding mechanism comprises a main machine (1), a feeding mechanism body (20) is arranged on the front portion of the main machine (1), the feeding mechanism is connected with an iron wire frame (2) through an iron wire, and a discharging mechanism (3) is arranged on one side of the main machine (1). The feeding mechanism body comprises a base (4), fixed wheels (5) are arranged on one side of the base (4), and moving wheels (6) are arranged on the other side of the base (4). Side bases (7) are arranged on two sides of the base (4), sliding bases (8) are arranged on the inner sides of the side bases (7), and the contact surfaces of the sliding bases (8) and the side bases (7) are in sliding fit. The moving wheels (6) are arranged on the upper sides of the sliding bases (8), a threaded barrel (9) is fixed below the sliding bases (8), a threaded column (10) is arranged in the threaded barrel (9), a bearing pedestal (11) is arranged at one end of the threaded column (10), a rotating ring (12) is fixed to the other end of the threaded column (10), the rotating ring (12) is provided with a circle of small holes (13), and a rotating handle (14) is arranged in one small hole (13) in an embedded mode.

The invention provides a data alignment method and system and a storage medium. The method comprises the steps that world time information is acquired, and a constraint mode of a trigger signal is selected; outputting a trigger signal to a corresponding sensing device according to the world time information and the constraint mode, so that each sensing device obtains sensing data based on the trigger signal; and based on the constraint mode, selecting a corresponding alignment mode to perform alignment processing on the sensing data. According to the method provided by the invention, the problem that a back-end perception fusion application cannot obtain accurate perception information due to the fact that the perception contents of the perception devices are not matched is solved, and the perception data collected by the multiple perception devices are effectively aligned.

The invention discloses an intelligent customer service cross-project experience migration method in combination with migration learning. The method comprises the following steps: S1, generating a source project feature set; s2, generating a source project optimization model containing a shared feature layer; s3, performing global optimization on shared feature layer parameters in the source project optimization model through Bayesian optimization; s4, generating an optimized target project feature set; s5, constructing a target project model; s6, performing fine tuning training on the target project model; s7, performing semantic consistency evaluation, response accuracy evaluation and business logic matching degree evaluation on the fine-tuned target project model on the target project data set; and S8, deploying the target project model passing the performance verification to a target project intelligent customer service system. According to the method, efficient deployment of the cross-project intelligent customer service system is realized, the migration efficiency and the service quality of the model are remarkably improved, and the method can be widely applied to intelligent customer service scenes of multiple industries.

The invention discloses a lead screw straightener, and relates to the technical field of lead screw straightening, the lead screw straightener comprises a frame body, the frame body comprises a bottom plate and two supporting frames, the top of the bottom plate is connected with a supporting plate, one side of the supporting plate is provided with a plurality of laser displacement sensors, a blanking plate is connected between the two supporting frames, and the lead screw straightener further comprises a driving mechanism assembled on one side of the supporting frames, in the automatic feeding link, through the contact effect of a poke rod and a contact block and in combination with linkage of a sliding plate and a compression spring, batched accurate feeding and rapid resetting of the ball screws are achieved, the whole feeding and clamping process is efficient and smooth, after straightening is completed, a connecting cylinder continues to rotate, and the automatic feeding and clamping device is used for automatically feeding the ball screws. A telescopic column and an abutting column are used for driving a first arc-shaped plate and a second arc-shaped plate to automatically open and release a ball screw, in the process, accurate reset of the first arc-shaped plate and the second arc-shaped plate is ensured through an energy storage and release mechanism of an arc-shaped spring, and preparation is made for next-round machining.

The invention discloses a sketch pedestrian re-identification method and device based on cross-modal reliable subjective features, and relates to the technical field of character sketch image processing, and the method mainly comprises the steps: obtaining a training set, a verification set and a test set according to character RGB and a sketch image; constructing a dual-path feature enhancement network by using the residual network, and training by using the training set to obtain a trained dual-path feature enhancement network; and according to the test set, utilizing the trained dual-path feature enhancement network to obtain a character sketch retrieval result. According to the sketch pedestrian re-identification method and device based on the cross-modal reliable subjective features, the accuracy and robustness of a cross-modal retrieval task can be improved.

The invention discloses a multi-source domain rolling bearing fault diagnosis method and system based on a diffusion model. The diagnosis method comprises the steps that rolling bearing vibration signals under different rotating speed working conditions are collected; in the forward process of the diffusion model, converting a source domain label into a one-dimensional vector, and adding noise to destroy original distribution to obtain yT approximate to Gaussian distribution; replacing yT with a Gaussian distribution variable # imgabs0 # in a reverse process, learning noise distribution characteristics added in forward distribution through a noise prediction network, denoising # imgabs1 # step by step to obtain # imgabs2 # in the same distribution as y0, inputting source domain data into the noise prediction network to extract data characteristics, inputting target domain data into the noise prediction network, and outputting the target domain data into the noise prediction network; global noise priori and local noise priori corresponding to the target domain data are obtained; and inputting source domain data and target domain data into a noise prediction network, mapping source domain data features to a target domain by using a multi-source domain adaptive algorithm, learning noise distribution features related to the target domain data, and gradually denoising # imgabs3 # to obtain a fault type of the target domain data.

The present invention provides an agricultural scientific research model training method, device, electronic device and storage medium, which belongs to the field of agricultural data processing technology. The method includes: obtaining a data set; encrypting each data set; obtaining a public identification data set based on the intersection of identification data sets, and then obtaining a target privacy data set; training based on the target privacy data set to obtain an agricultural scientific research model. The present invention determines the encrypted target privacy data set according to the identification data set, and realizes the effective alignment of data from different sources in an encrypted state, which is beneficial to improving the consistency and integrity of the target privacy data set and protecting the security of privacy data. The present invention obtains the aligned target privacy data set according to the public identification data set, and then trains the preset model to obtain the agricultural scientific research model. While protecting data privacy, it improves the practicality and security of agricultural scientific research data analysis, which is beneficial to improving the efficiency of model training in the field of agricultural scientific research.

The present invention provides a cross-prompt automatic composition scoring method, device and electronic device based on category adversarial joint learning, which belongs to the technical field of composition scoring, and obtains composition data of a composition to be evaluated; inputs the composition data into a feature extractor with a Hi-att network, and outputs composition features; defines a discriminator, and conducts adversarial training on the discriminator and the feature extractor; splices the composition features with manual features, and predicts the score through a plurality of different nonlinear feature transformations and calculates the score loss; uses regression loss, classification loss and adversarial loss to process the score, grade prediction and generalization ability of the composition to be evaluated. The present invention improves the performance of the model. By adopting a category-aware adversarial training method, the classification performance is improved, and the problem of misaligned alignment between source prompts caused by uneven category distribution is effectively avoided, which greatly enhances the practicality and accuracy of the model.

The present application relates to the field of sheet media processing, and particularly relates to a deviation rectifying device, a sheet media processing device and a control method of the deviation rectifying device; a rack of the deviation rectifying device is provided with a conveying channel, the conveying channel is used for conveying sheet media, and one side of the conveying channel is provided with a reference surface; a support frame of the deviation rectifying device is rotationally connected with the rack and has a first position and a second position, and a deviation rectifying wheel is rotationally arranged on the support frame and extends into the conveying channel; a first driving mechanism is used for driving the support frame to rotate between the first position and the second position; a second driving mechanism is used for driving the deviation rectifying wheel to rotate; when the support frame is located at the first position, an axis of the deviation rectifying wheel is arranged at a first included angle with the reference surface; when the support frame is located at the second position, an axis of the deviation rectifying wheel is arranged at a second included angle with the reference surface, and the first included angle and the second included angle are both less than or equal to 90°. The deviation rectifying device can improve the success rate of deviation rectification of sheet media.

The utility model discloses an automobile seat installing and positioning device which comprises four distance adjusting mechanisms, the four distance adjusting mechanisms are fixedly connected through connecting plates, each distance adjusting mechanism comprises a supporting rod and an inserting rod, the supporting rod is provided with an open cavity, the inserting rod is inserted in the open cavity in a sliding mode, and the distance adjusting mechanisms are fixedly connected through the connecting plates. The supporting rod and the inserting rod are connected through a positioning mechanism, the positioning mechanism comprises a supporting block fixedly connected to the outer wall of the supporting rod, a swing rod is rotationally connected to the supporting block, a traction rod is fixedly connected to one end of the swing rod, and a pressing plate is fixedly connected to one end of the traction rod; according to the seat mounting device, the position of a seat to be mounted is accurately positioned, effective alignment of the seat is guaranteed, and the mounting efficiency of the seat is improved.

The invention provides an image processing method and device, electronic equipment, a chip and a storage medium. The method comprises the following steps: carrying out superpixel segmentation on an image to be processed to obtain a plurality of superpixels; classifying different superpixels to obtain at least one depth-of-field region; wherein different depth-of-field areas correspond to different depths of field; based on the reference image, determining a target homography matrix of each field depth area; wherein the reference image is an image aligned with the image to be processed; and performing image warping processing on the corresponding depth-of-field region by using the target homography matrix of each depth-of-field region. According to the scheme provided by the invention, alignment of different depth-of-field planes can be realized at the same time, and the image alignment effect is improved.

The invention relates to the technical field of thrust chamber dissimilar metal welding, in particular to an electroforming nickel dissimilar metal vacuum laser welding process method which is used for thrust chamber axial butt joint welding seams, the welding seams are of a rotary body structure and a whole circle of circumferential welding seams, a non-full penetration structure mode is adopted, and base materials on the two sides of the welding seams are a high-temperature alloy section and an electroforming nickel section respectively. And vacuum laser welding is adopted, programming is conducted twice in the welding process, programming is conducted by deflecting a light spot to one side of the electroforming nickel section by 0.2-1.5 mm for the first time, and programming is conducted by deflecting the light spot to one side of the electroforming nickel section by 1-2 mm for the second time. According to the electroforming nickel dissimilar metal vacuum laser welding process method, the problem of welding deviation during welding of electroforming nickel and high-temperature alloy dissimilar steel is solved, the quality and reliability of welding seams are improved, and the electroforming nickel dissimilar metal vacuum laser welding process method is suitable for manufacturing of key components such as thrust chambers.

The invention discloses a disease classification method and system based on knowledge enhanced image text alignment, and belongs to the technical field of artificial intelligence and digital medical treatment. In order to solve the problems that in an existing method, only medical entity information is considered to promote alignment between a medical image and a text report, and graph structure information of a knowledge graph is ignored, firstly, data of the medical image and the text report are preprocessed; connecting the processed medical image and report data with a medical knowledge graph UMLS to construct an external knowledge graph, constructing a disease classification network model, and training the disease classification network model by using the processed medical image and report data and the external knowledge graph; and finally, performing disease classification on the medical image of the new patient by using the trained disease classification network model. The diagnosis efficiency is improved by outputting the category information of the medical image of the diagnosed patient.

The invention provides a magnetic clamping auxiliary laser centering device for a drilling machine. Comprising a laser transmitter, a magnetic shell, an end cover, a bolt, a sealing cover, a supporting block, a clamping claw and a limiting ring. The supporting block is of a split type combined structure and comprises a first supporting block body, a second supporting block body and a third supporting block body, the lower portion of the third supporting block body extends out of the magnetic shell, a laser transmitter is assembled in the center of the third supporting block body, and the laser transmitter and the center of the supporting block are collinear; the gripper jaw is arranged in a spacing cavity of the supporting block, and repulsive magnetic force is in contact with the drill bit and centered and locked; the limiting ring is assembled on the supporting block to limit the clamping jaw; the end cover is concentric with the magnetic shell and assembled at one end of the magnetic shell, the supporting blocks are contained in the magnetic shell, and the laser transmitter assembled at the extending position of the third supporting block extends out of the magnetic shell. The sealing cover is concentric with the magnetic shell and assembled at the other end of the magnetic shell; and the bolt is used for detachably assembling the end cover and the sealing cover on the magnetic shell. The center point and the center line of the drill bit are indicated through laser.

The invention discloses a power transmission line infrared image instance segmentation method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the training of a first teacher model according to a visible light image and label information, and generating a first trained teacher model; according to the first infrared image and the label information, training a second teacher model to generate a second trained teacher model; according to the second infrared image and the first pseudo label information, training the first student model to generate a first trained student model; according to the second infrared image and the second pseudo label information, training the second student model to generate a second trained student model; according to the second infrared image and the third pseudo label information, performing fine tuning training on the first trained student model and the second trained student model to generate a second convergence teacher model; and performing instance segmentation on the infrared image of the power transmission line through the second convergence teacher model. According to the invention, the instance segmentation precision of the infrared image can be improved.

The invention belongs to the technical field of bearing lubrication state monitoring, and particularly relates to a bearing lubrication state real-time monitoring method based on hydrodynamic model correction, and the method comprises the steps: obtaining an original vibration acceleration signal and operation state data, carrying out the variational mode decomposition of the original vibration acceleration signal, and extracting a cavitation feature component; calculating a cavitation kinetic energy impact index according to the extracted cavitation characteristic component and the operation state data; dynamically correcting the equivalent fluid density of the lubricating oil according to the cavitation kinetic energy impact index and the contact area pressure of the bearing estimated by the running state data; and solving the minimum oil film thickness of a lubricating oil film by utilizing the corrected equivalent fluid density, and determining a lubrication safety margin index by combining the temperature rise rate of the bearing. According to the invention, the defect of monitoring result distortion caused by single-phase fluid hypothesis is overcome, and the accuracy of bearing lubrication state monitoring and early risk early warning is improved.