28results about How to "Reduce imbalance" patented technology

The invention discloses an aero-engine vibration signal generation method under extreme imbalance, and belongs to the technical field of aero-engine fault signal generation. The method comprises the following steps: acquiring an original vibration signal at a target position of the aero-engine, and performing preprocessing operation; taking the preprocessed real vibration signal as a training sample, constructing a forward noise adding process, and training the improved one-dimensional diffusion generation network to learn a reverse denoising mapping relation; and inputting random Gaussian noise into the one-dimensional diffusion generation network after training convergence, and generating a target vibration signal matched with a real vibration signal feature through a reverse diffusion denoising process. Through targeted improvement of the diffusion generation network, the problem of data imbalance is effectively relieved, the model deployment cost is reduced, the signal generation capability of the model for minority types of faults is improved, and the method is suitable for fault signal generation scenes of aero-engines under complex working conditions.

The invention discloses a multi-source data fusion-based air volume dynamic adjustment method of a fume hood for a clean environment, particularly relates to the technical field of clean environment control, and aims at solving the problem that air volume adjustment of an existing fume hood cannot give consideration to front air suction safety, pressure difference stability and overall energy conservation. A basic operation data set is formed by collecting and filtering window opening and cabinet air speed, an air volume demand value and an initial priority parameter are calculated according to the data set and a room use state mark to form a demand data set, and a deviation memory parameter and a fluctuation influence parameter are calculated by fusing room pressure difference data and the availability of an exhaust fan. Obtaining a stable control priority adjustment coefficient through a time decay score learning model, correcting the priority, distributing the target exhaust air rate and the exhaust fan output, forming a target distribution result, comparing the target distribution result with the actual air valve opening degree and the exhaust fan output state, calculating the adjustment amount, and issuing a control instruction according to the priority; and the deviation record feedback updating coefficient and the demand value are verified in the control period to realize self-adaptive adjustment.

The invention discloses a few-sample fault diagnosis method based on TFCD fusing time sequence characteristics, and the method comprises the steps: inputting a preprocessing state signal into a trained fault diagnosis model, so as to enable the fault diagnosis model to carry out the fault diagnosis of a core part. The fault diagnosis model is obtained by training a plurality of training samples generated by a trained TFCD model, and the TFCD model extracts discriminative time sequence features by introducing a long short term memory network, and embeds the discriminative time sequence features as static condition vectors into a DDPM model. According to the method, the model can effectively capture the long-range dependency relationship and the dynamic mode characteristics in the vibration signal, so that the generated sample keeps good coherence in the time domain, meanwhile, the specific periodic impact mode of the fault is reserved, and the physical rationality of the generated sample is remarkably improved. Therefore, accurate diagnosis under cross-working conditions can be realized.

The present application belongs to the technical field of installation detection of clothes processing equipment, and discloses an installation detection method, an installation detection device and clothes processing equipment. The installation detection method comprises the following steps: after the installation of the clothes processing equipment is completed, a detection operation process of the clothes processing equipment is performed; during the detection operation process of the clothes processing equipment, the balance state of the clothes processing equipment is judged according to the pressure values of a plurality of detection positions at the bottom of the clothes processing equipment, and balance adjustment prompt information is reported; and installation prompt information is reported according to the noise value and the vibration acceleration value of the clothes processing equipment. The installation detection method can detect the clothes processing equipment after the installation is completed, analyze the causes of the eccentricity problem of the equipment, and facilitate quick adjustment.

The invention discloses a lactobacillus delbrueckii subsp. Bulgaricus CCFM1520 for transforming curcumin to produce tetrahydrocurcumin to relieve nervous anxiety, and belongs to the technical field of microorganisms. The lactobacillus delbrueckii subsp. Bulgaricus CCFM1520 screened by the invention has the capability of converting curcumin, and can utilize curcumin and generate tetrahydrocurcumin. The invention further provides a synbiotic preparation prepared from the lactobacillus gasseri CCFM1481 and curcumin, the synbiotic preparation can be used as an anti-fatigue functional medicine, the individual behavioral memory level can be improved, the imbalance of body neurotransmitters (DA, 5-HT and NE) can be slowed down, the expression of body neuroprotection related genes (BDNF, TrkB, DRD1 and TH) can be improved, and the synbiotic preparation has a huge application prospect.

The invention relates to the technical field of remote sensing image fine-grained target detection, in particular to a two-way guided remote sensing image fine-grained target detection method. Comprising the steps of 1, acquiring image data, and dividing the image data into a training set and a test set; 2, constructing a target detection network, and obtaining image multi-scale features from the preprocessed image through a feature extraction network; then, a coarse-grained detector screens out a target candidate area and judges whether the target candidate area is a foreground or a background; then, a fine-grained detector carries out finer classification and positioning on the candidate areas; and step 3, carrying out performance evaluation on the test set so as to test the detection capability of the model on the fine-grained target in an actual scene. According to the method, for the problems of class sample imbalance and easy confusion of fine-grained classes in fine-grained target detection, the remote-sensing image fine-grained target detection capability is effectively improved by constructing a layered detection network architecture and combining an instantaneous confidence coefficient signal and class accumulation learning quality.

The application relates to a video instance segmentation method based on a dynamic convolution decomposition of a lightweight attention mechanism; the method first inputs a video frame, and a backbone network independently extracts a feature map of each frame in the video; then the feature map extracted by the backbone network enters an HQT encoding and decoding module, the position change of an instance in each frame is accurately positioned by combining an output head, and three prediction branches are used to supervise model training; in the application, a convolution layer of an encoding network adopts dynamic convolution decomposition, a more compact model is obtained, model training is easier, the required parameter quantity is greatly reduced, the training speed is improved, and the training time is shortened; a lightweight HQT encoding and decoding module is provided, which helps to reduce model parameters and improve efficiency; the loss function is improved, model training is more stable, the convergence speed and convergence precision are improved, and the problems of imbalance between foreground and background in samples and imbalance between foreground categories under a long tail condition are relieved.

The invention belongs to the technical field of power distribution network leakage current detection, and particularly relates to a CNN and FiLM-based leakage current type intelligent identification method. Comprising the following steps: S1, collecting real-time operation data of a typical power supply area, obtaining leakage current waveforms and related environment characteristic parameters under different working conditions, and constructing an original leakage current sample data set; s2, a CNN-based leakage current classification model is constructed, a FiLM condition modulation module is introduced, a multi-task learning framework is used at the tail of the network, a main task is leakage current type identification, and an auxiliary task is grounding system discrimination; s3, using a weighted cross entropy loss function to alleviate a class sample imbalance problem; an OneCycleLR dynamic learning rate scheduling strategy is introduced; s4, evaluating the performance of the model on the test set, and using the accuracy and the confusion matrix as evaluation indexes; according to the method, high-precision identification of multiple types of faults such as single-phase grounding, arc type electric leakage and direct current system electric leakage is realized, different grounding systems can be adapted, and the accuracy and robustness of system diagnosis are improved.

The present application relates to the field of industrial data processing, and provides an industrial data virtual sample generation method based on physical constraints and attention mechanisms, which constructs feature data in response to an industrial data virtual sample generation task; the feature data is input into a virtual sample generation model for training, and corresponding candidate virtual samples are generated after model training, the virtual sample generation model includes a classification encoder, a mean-variance encoder, a Gaussian mixture latent space and a decoder, the candidate virtual samples are sequentially subjected to category validity screening, physical constraint compliance screening and distribution similarity screening to construct a high-quality candidate sample pool, and a diversity saturation critical point is determined based on feature entropy change trend analysis, and an optimal virtual sample set is output. Through the collaborative fusion of the attention mechanism, Gaussian mixture latent modeling and physical constraint loss, combined with multi-dimensional screening and entropy change monitoring mechanism, unified control of virtual sample generation in a small sample industrial scene is realized.

A data re-labeling classification algorithm based on clustering learning comprises the following steps: after preprocessing the expression image data in the original sample, a feature extraction module is used to extract a feature vector to form sample data; a dynamic clustering module splits a sample data set of a minority class into several sub-class data sets, and a sample data set of a majority class remains unchanged; the extracted feature vector is used as the input of the dynamic clustering module; if the sample data belongs to the sample data set of the minority class, the sample data is assigned to the nearest sub-class data set according to the distance between the sample data and the clustering center, and then a pseudo label is assigned to all sample data; the sub-classes are trained by using the pseudo label; and a label mapping module maps the prediction result of the sub-class data set to which the sample data belongs back to the real label space based on the mapping relationship learned in the training process. The application can improve the classification and recognition performance of an unbalanced expression data set, and fully excavates and utilizes the potential information of the sample data set.

This invention relates to the field of image processing, specifically providing an image segmentation processing method, control device, and readable storage medium, aiming to solve the problem of how to achieve more accurate segmentation of subtle categories in images. To this end, the image segmentation processing method of this invention includes: extracting features from the image to be segmented to obtain extracted features and stitched features, wherein the stitched features are obtained by further extracting features from the extracted features; and selecting, according to preset conditions, to perform image segmentation using either the stitched features or the combined stitched features and extracted features, thereby obtaining a segmentation category map. This solution preserves the high resolution of the image and improves the accuracy of image segmentation. During segmentation, in addition to using the stitched features output by the feature extraction network, feature information from the feature extraction stage is also incorporated, providing more referenceable feature information when optimizing the segmentation results, further ensuring the refinement of the segmentation results.

The invention provides a traffic flow rapid sensing method based on multi-modal fusion. Comprising the following steps: S1, a multi-source heterogeneous sensor carries out collaborative data acquisition and space-time alignment; s2, edge side data preprocessing and feature level fusion; s3, multi-modal target perception and feature extraction based on an improved network model; s4, performing multi-modal decision-making level fusion and accurate state recognition; s5, performing global traffic state monitoring and event identification; and S6, performing model optimization and edge efficient deployment. According to the invention, through fusion of a heterogeneous sensor space-time cooperation mechanism, improved YOLOP network construction, a hierarchical decision fusion engine and an edge intelligent optimization technology, high-precision rapid perception and edge real-time processing of traffic flow under an all-weather condition are realized.

The gapless viscous damper comprises a main cylinder, an auxiliary cylinder and a guide rod, the auxiliary cylinder is installed on the right side of the main cylinder, the guide rod is movably installed in the main cylinder and extends into the auxiliary cylinder, and a piston used for sliding on the inner wall of the main cylinder is installed on the guide rod. The clamping assembly is used for assembling the guide rod and the piston, the first clamping key, the second clamping key and the third clamping key are installed between the guide rod and the piston, an original thread or welding mode of the piston is changed into a groove three-clamping-key mode, the concentricity of the guide rod is greatly guaranteed through the connecting mode structure, the welding portion is reduced, and the service life of the guide rod is prolonged. And the environmental protection requirement is reduced. The assembling is simple, and the production efficiency is high. And when concentricity is ensured, the damper can distribute pressure of an oil film / silicone oil more uniformly in the working process, so that vibration energy is more effectively absorbed and attenuated, the vibration amplitude is reduced, and the vibration reduction effect is improved. And meanwhile, the unbalance phenomenon of the damper in the working process is reduced.

This application relates to an improved sample-free, incremental learning method for predicting materials microscopy images. The method includes: constructing a materials microscopy image prediction model; training the model using a set of materials microscopy images; extracting initial features from a first dataset using a feature extractor; expanding the initial features in a feature and label alignment module and determining the parameters of the linear layer while preserving the feature backbone of the feature extractor; optimizing the feature extractor and determining the parameters of the second linear layer in a feature optimization and alignment module; performing incremental alignment and saving the prototype of the new class; performing label alignment on the expanded pseudo-features based on the parameters of the second linear layer in a forgetting compensation and testing module; and using the trained materials microscopy image prediction model to predict the materials microscopy image to be predicted. This method can improve prediction accuracy.

The application discloses a personal image privacy protection method, system and electronic equipment. The method provided by the application allows a user to specify a certain person in a protected image through language expression, generates multi-scale visual features with sufficient fusion of image and text information by using a lightweight deep neural network to perform parallel coding on input reference information and personal images, generates a stable specified personal privacy protection image mask through a multi-scale feature fusion and mask positioning enhancement module in the decoding process, and in addition, introduces a balanced binary cross-entropy loss to solve the pixel imbalance problem in training, optimize the network performance, and improve the personal image privacy protection effect. The application can solve the problems of excessive protection of content and pixel imbalance in the training of the reference personal image privacy protection network in the existing personal image privacy protection technology.

The invention provides an automatic judgment system and method for a mechanical threat level of a power transmission line, belongs to the technical field of safety operation and maintenance of the power transmission line, and constructs an automatic judgment system comprising a data acquisition module, a dimension decomposition and data preprocessing module, a visual language large model training and reasoning module and a result output and early warning processing module. The method comprises the following steps: acquiring a line image and basic attribute data, forming a structured data packet, analyzing an unstructured image into structured tags of different core dimensions, then completing multi-modal feature extraction and multi-dimensional decision analysis to output a threat level, and finally realizing display, early warning and storage. According to the method, dimension deconstruction is carried out on a complex mechanical threat judgment problem, a visual language large model technology and professional knowledge of power transmission line operation and maintenance are deeply fused, refined grading judgment of power transmission line mechanical threats is realized, the accuracy and interpretability of judgment are improved, and the judgment efficiency is improved. And the automation and intelligence level of operation and maintenance of the power transmission line is effectively improved.

The invention provides a mode perception prediction model for building load imbalance data, relates to the field of power load prediction, and recognizes typical operation modes in building load data through a K-means clustering algorithm, including startup impact, morning, afternoon and shutdown dormancy. A time window feature enhancement strategy is introduced to extend an original one-dimensional feature space to a three-dimensional feature space, a mode-aware oversampling method PP-SMOTE is constructed, an optimal balance factor is obtained through a balance factor optimization strategy, and the quality of data generated by the PP-SMOTE is verified; the PP-SMOTE is applied to a LightGBM prediction model, and the optimal balance between prediction precision and data distribution characteristics is realized through a method for determining the optimal data balance degree through a multi-balance-factor comparison experiment.

The application discloses a large-span stiff skeleton arch bridge temperature-response intelligent mapping method and belongs to the technical field of bridge structure health monitoring. S100: data acquisition; S200: data preprocessing, generating a data set, and dividing into a training set and a validation set; S300: constructing a Phy-SETCN model based on the training set / validation set, using MAE, RMSE, R 2 The strain data obtained by mapping the Phy-SETCN model is evaluated; S400: real-time detection of bridge temperature-induced strain and abnormal response early warning through the Phy-SETCN model. The application can be widely applied to a large-span bridge structure health monitoring system, effectively identifies temperature-induced responses through a high-precision temperature-strain mapping model, provides a reliable basis for bridge safety evaluation, and reduces operation and maintenance costs.

The invention belongs to the field of intelligent biological prediction, and particularly relates to an RNA-small molecule binding affinity prediction method based on multi-view learning. The method comprises three stages of data set up and down sampling processing, feature construction and extraction based on RNA and small molecule sequences, and binding affinity prediction. According to the method, data set distribution is optimized through an up-down sampling method, multi-view feature information is constructed through sequence information of RNA and small molecules, a deep network is established to extract and depth biological information features, and finally, an MLP regression device is adopted to predict an RNA-small molecule binding affinity value, so that drug discovery work is assisted to be achieved. In order to solve the problem of imbalance of the existing data set, the invention provides a method of combining up and down sampling for processing, and the problem of imbalance is effectively relieved.

The application discloses a MMC loss balancing control method and system based on upper and lower limits of switching factors, and specifically is: based on the sub-module switching state, the actual sampling obtained sub-module capacitor voltage is adjusted, and is defined as the first sub-module corrected capacitor voltage; the switching factor of each sub-module is calculated; based on the switching factor of the sub-module and the upper limit and the lower limit thereof, the second corrected capacitor voltage is calculated; based on the upper limit and the lower limit of the sub-module capacitor voltage, the third sub-module corrected capacitor voltage is calculated; based on the third corrected capacitor voltage and the MMC modulation strategy, the corresponding sub-module is selected to be put in and cut off. The application can effectively balance the switching frequency and switching loss distribution among the sub-modules of the modular multilevel converter under a low switching frequency, thereby effectively improving the operation reliability of the converter, meanwhile, without the need of additionally increasing sensors and without the need of changing the hardware structure of the MMC system, the control is simple and easy to implement, and has strong economy and practicality.

The utility model discloses a mining explosion-proof type concrete mixing tank radiator which comprises a mounting frame, a rotating shaft is rotatably connected to the middle of the mounting frame, a fan blade is fixedly connected to one end of the rotating shaft, and a hydraulic motor is fixedly connected to the side, opposite to the fan blade, of the mounting frame. One end, opposite to the fan blades, of the rotating shaft is fixedly connected with an output shaft of the hydraulic motor; a filter screen is arranged on the side, close to the hydraulic motor, of the mounting frame, a fixing cover is fixedly connected to the side, close to the fan blades, of the mounting frame, and a plurality of blocking pieces are arranged on the front side of the fixing cover. By means of the structure, the small hydraulic motor is connected to a hydraulic system loop in series, the hydraulic motor drives the fan blades to well dissipate heat of the hydraulic system, meanwhile, circuit explosion accidents can be avoided, and the using safety under the mining condition is guaranteed; and the filter screens and the blocking pieces are arranged on the two sides of the fan blades, attachment of dust and other impurities to the fan blades can be reduced, and fan blade unbalance and abrasion caused by dust accumulation are reduced.

The application provides an extended training data set, and a method for classification using a classification model trained by the extended training data set. The method for extending the training data set includes generating merged data based on original data and auxiliary data, the auxiliary data being associated with an application scenario of the classification model and coming from a source different from a source of the original data; and updating a category of the merged data using the classification model to generate a first extended data set for training the classification model. The scheme of the application can expand the number of labeled data samples and improve the labeling accuracy of the data, so that the classification performance of the trained classification model is optimized.

The invention discloses a remote sensing image classification and target detection method fusing three-way decision and multi-agent reinforcement learning, and belongs to the technical field of remote sensing image processing. The method comprises a classification process and a target detection process: in the classification process, firstly, feature importance is calculated through grey correlation analysis, then a multi-agent system is constructed to optimize three decision threshold values of a TwGrey feature selection algorithm, and after an optimal feature subset is obtained, the optimal feature subset is input into an SVM classifier to complete classification; according to the target detection process, firstly, a multi-agent collaborative improved FPN is constructed to strengthen multi-scale features, then, a high-quality anchor frame is screened through a sequential three-way decision model, and finally, a detection head is input to complete target detection. The system is correspondingly provided with a classification module, a target detection module, a multi-agent reinforcement learning module and a data interaction module, and cooperative work among the modules is achieved. The method effectively solves the problems of high remote sensing image classification dimension, multi-scale target detection and sample imbalance, improves the processing precision and stability, and can be widely applied to the fields of urban planning, geological disaster monitoring and the like.

The invention relates to the technical field of hard alloy column tooth passivation, and particularly discloses a hard alloy column tooth passivation machine tool which comprises a frame body, an inner cylinder is rotationally arranged at the upper end of the frame body, a cylinder cover is hinged to the upper end of the inner cylinder, a clamping assembly is arranged on the bottom wall of the inner cylinder, a liquid storage tank is arranged on one side of the frame body, and a water pump is arranged on one side of the liquid storage tank. The water inlet end of the water pump communicates with the liquid storage tank, the water outlet end of the water pump communicates with the middle of the bottom wall of the inner cylinder, and an electrolytic passivation mechanism is arranged on the inner bottom wall of the inner cylinder; the cylindrical teeth are placed in the inner cylinder, the clamping assembly can stably clamp the cylindrical teeth, then the cylinder cover is closed, the water pump is started, the water pump on one side of the frame body pumps electrolyte in the liquid storage box into the middle of the bottom wall of the inner cylinder, and at the moment, the electrolytic passivation mechanism can make the polished ends of the cylindrical teeth generate uniform passivation films; and the scraping effect and scraping efficiency of the passivation film are improved, so that the passivation effect of the column teeth is improved, the corrosion resistance of the column teeth is improved, and the service life of the column teeth is prolonged.

The invention discloses an MMC loss balance control method and system based on upper and lower limits of a switching factor, and the method specifically comprises the steps: adjusting a sub-module capacitor voltage obtained through actual sampling based on a sub-module switching state, and defining the sub-module capacitor voltage as a first sub-module correction capacitor voltage; calculating a switching factor of each sub-module; based on the switching factor of the sub-module and the upper limit and the lower limit thereof, calculating to obtain a second corrected capacitor voltage; based on the upper limit and the lower limit of the sub-module capacitor voltage, calculating to obtain a third sub-module correction capacitor voltage; and based on the third correction capacitor voltage and an MMC modulation strategy, selecting a corresponding sub-module for switching on and switching off. According to the method, the switching frequency and switching loss distribution among the sub-modules of the modular multilevel converter can be effectively balanced under the low switching frequency, so that the operation reliability of the converter is effectively improved, meanwhile, additional sensors are not needed, the hardware structure of an MMC system does not need to be changed, control is simple and easy to implement, and high economical efficiency and practicability are achieved.

The invention relates to a high-speed electric machine for a motor vehicle, comprising a stator and a rotor which is rotatably mounted relative to the stator along an axis, the stator comprising a stator yoke and stator teeth, the stator teeth being connected to the stator yoke in a force-fitting and / or form-fitting manner, the rotor comprises sheets which form a laminated core and are laminated in the axial direction, the laminated core comprises at least two receiving spaces in the radial direction relative to the axis, a permanent magnet is arranged in each receiving space, the corresponding permanent magnet is inserted in the corresponding receiving space in a force locking mode, and the permanent magnets are arranged in the receiving spaces. The laminated core with the permanent magnets has a circular outer contour in the radial direction with respect to the axis, wherein the rotor is enclosed by a retaining strip for fixing the permanent magnets in the laminated core.

This application provides a method and system for regulating the electrode current of a calcium carbide furnace, relating to the technical field of electrode current regulation in calcium carbide furnaces. The method includes: collecting and estimating unmeasurable state parameters of the electrode system online using an extended Kalman filter algorithm based on three-phase current data and the lifting / lowering speed data of the three-phase electrodes; obtaining a three-phase current setpoint; and, based on the deviation between the three-phase current setpoint and the collected three-phase current data, the online identified three-phase coupling coefficient matrix, and the unmeasurable state parameters, performing rolling optimization in the prediction time domain using coupled model predictive control, considering three-phase coupling constraints and system time lag, generating a predictive control quantity for the lifting / lowering speed of the three-phase electrodes; and outputting the predictive control quantity for the lifting / lowering speed of the three-phase electrodes to the electrode lifting / lowering actuator to adjust the electrode position. By employing the above method, the technical effect of improving the accuracy of three-phase current control is achieved.

The application discloses a lung CT image analysis method and electronic equipment based on deep learning, which comprises the following steps: constructing a data set; establishing an expert pool, including vectorizing CT image data and constructing an expert pool expert segmentation submodel containing multiple expert segmentation submodels for processing data queues with specific image distribution characteristics; constructing a feature extractor for gating, mapping the CT image into a high-dimensional semantic vector reflecting the confidence and distribution label of the abnormal area; using a gating network to adaptively segment the CT image, and selecting the optimal expert segmentation submodel from the expert pool according to the input mixed image feature vector. The application improves the segmentation accuracy of heterogeneous lesions, realizes adaptive selection of different image phenotypes, and outputs three-dimensional segmentation masks and attribute prediction vectors, supporting subsequent quantitative evaluation and clinical decision-making; and the application alleviates the sample imbalance problem and can accurately route rare lesion forms.