11results about How to "Realize self-learning" patented technology

The present application relates to a kind of facial expression recognition methods based on attention mechanism and self-distillation, comprising the following steps: step 1, constructing facial expression recognition dataset;Step 2, constructing facial expression recognition model, the model is composed of feature extraction module, adaptive channel attention module, remodeling module and self-distillation network;Step 3, the input facial expression image is preprocessed;Step 4, using the training set image in the above facial expression recognition dataset trains facial expression recognition model;Step 5, using the facial expression recognition model trained to the test set and validation set in dataset are inferred classification.The method designs a kind of reasonable and efficient facial expression recognition scheme, it is well guided model to pay attention to those important feature information by the mode of attention mechanism;By a new self-distillation form, refine compressed network knowledge, improve the robustness of shallow network output feature and reduce the complexity of model in inference stage.

The present application relates to the field of science and education display, and particularly relates to a display control method of a multi-axle three-dimensional dynamic pseudo-motion phenomenon display device, which comprises the following steps: S1, setting up a three-dimensional dynamic pseudo-motion phenomenon display device; S2, representing the pseudo-motion phenomenon effect of the display device by a control function G; S3, establishing a target function library, selecting the best function from the target function library, and obtaining the optimal solution of each coefficient in the control function G by using the least square method; S4, determining the rotation angular velocity range of the control function G; S5, combining the control function G to establish an input parameter database, constructing a machine learning control algorithm, and thus self-adjusting the associated parameters affecting the control function G. The present application can present the best display effect for different people in different environments.

The present application relates to the technical field of equipment maintenance, and more particularly to an intelligent predictive equipment maintenance system, comprising a data acquisition module, a data processing module, a maintenance decision module, a maintenance execution module, a user interaction module and a closed-loop optimization module. The data acquisition module collects equipment operating parameters through the deployment of a sensor array to generate standardized data streams; the data processing module extracts time sequence features and performs state deduction through a virtual model to output fault analysis results; the maintenance decision module calls enterprise resource information for multi-objective optimization to generate a dynamic maintenance plan; the maintenance execution module assists in guiding the execution of maintenance through augmented reality and records data; the user interaction module dynamically displays equipment status and risk information; and the closed-loop optimization module updates model parameters based on feedback data. The system realizes predictive maintenance under data driving through module collaboration, effectively reduces unplanned downtime and resource waste, and improves fault diagnosis accuracy and operation and maintenance efficiency.

The invention discloses a power system oscillation mechanism identification method and a model construction method, and relates to the field of power system low-frequency oscillation, and the method comprises the steps: collecting the active power data of a generator and the active power data of a tie line; performing linear normalization and principal component analysis dimension reduction processing on the data, and constructing a multi-dimensional time sequence to-be-tested sample; a sample is input into a pre-trained deep belief network (DBN) model, the model is formed by stacking a plurality of restricted Boltzmann machines (RBM), data deep features are extracted through unsupervised layer-by-layer greedy pre-training, and parameter optimization is performed in combination with supervised fine tuning. According to the method, negative damping oscillation and forced power oscillation can be automatically identified, the uncertainty of manual feature selection is avoided, the influence of noise and amplitude difference on the identification precision is effectively suppressed, and the adaptability of the model to an actual power grid complex disturbance condition is improved.

The invention relates to a visual identification system and method for goods sorting and transportation, and relates to the technical field of logistics transportation control, and the visual identification system comprises an image collection module which collects original goods images of goods in transportation according to different visual angles; the image processing module is used for cleaning and standardizing the original article image to obtain an article multi-view linked image and extracting a logistics distribution position; the model construction module is used for aggregating historical article images into a training set and generating a multi-view sorting and distinguishing model by combining with a target detection algorithm for iterative training; the screening identification module is used for judging an article multi-view linked graph according to a multi-view sorting judgment model, and determining an article sorting conclusion in combination with a logistics distribution position; the updating optimization module is used for filling an article multi-view linked graph into an article database according to an article sorting conclusion and optimizing a multi-view sorting judgment model; a sorting discrimination model is constructed through a multi-view image fusion algorithm and a deep convolutional neural network, and the precision and efficiency of intelligent logistics sorting are improved.

The invention relates to the technical field of major activity safety guarantee, and particularly discloses a major activity safety guarantee risk management system and method based on risk portray.The system comprises a data footstone layer, an analysis, study and judgment layer, a decision action layer and a verification optimization layer, and the data footstone layer constructs a dynamic hidden danger library following the four-dimensional principle; multi-source data are gathered through a standardized model, risks are quantified based on a weighted RPN formula, and a data injection and dynamic updating mechanism is matched. And the analysis, research and judgment layer generates a risk portrait report containing a structured list through scene modeling and multi-dimensional aggregation analysis. The decision action layer converts the portraits into tactical actions such as special plans, resource presetting and special drilling. And the verification optimization layer realizes risk monitoring regulation and knowledge precipitation through closed-loop management. According to the invention, the conversion from passive response to data-driven active defense is realized, the risk identification and pre-judgment precision is improved, the resource configuration and emergency efficiency are optimized, the system self-learning evolution is driven, and the risk recurrence rate of the same kind is reduced.

The invention relates to the technical field of ocean, and provides a method and system for optimizing wave parameters based on inner and outer double-layer iteration. The method for optimizing the wave parameters based on inner and outer double-layer iteration comprises the following steps: constructing a wave model based on a wave mode taking a spectral energy balance equation as a core; the wave model comprises a plurality of empirical parameters and adjustment coefficients, namely to-be-optimized parameters, and then to-be-optimized parameter vectors are formed; searching an optimal parameter vector by using an inner and outer double-layer iterative optimization method, so that the error between the output wave data and the observation wave data of the wave model is minimum; the error between the output wave data and the observation wave data is represented by adopting a regularization error function, and the regularization error function is a normalized weighted mean square error introducing a priori constraint regularization term. According to the method, a wave model parameter optimization system with intelligent adjustment, closed-loop feedback and continuous evolution characteristics can be constructed, and the method can be widely applied to high-precision parameter optimization and automatic calibration of a wave mode.

This invention discloses an intelligent control method and system for motorcycle calipers. The method involves real-time acquisition and preprocessing of multi-modal data, including vehicle speed, wheel speed, piston displacement, brake disc temperature, tilt angle, and lateral acceleration, using multiple sensors. Based on the data, braking demand and adhesion coefficient are calculated, and tilt angle and lateral acceleration are used to identify straight-line, curved, or emergency braking scenarios. The data is input into a pre-trained braking control machine learning model to generate a target braking force variation curve. Based on this curve, the drive motor and caliper are controlled for braking, and the model is periodically optimized based on braking feedback and historical data. A component degradation model is used to predict the braking efficiency decay trend, and the braking force is compensated and the model is updated. Closed-loop control is used to adjust the motor current in real-time to follow the target braking force. The brake disc temperature is continuously monitored, and thermal fade compensation is triggered when the temperature exceeds a threshold. This invention achieves adaptive, precise, and reliable intelligent braking for motorcycles.

The invention belongs to the technical field of automatic control, and particularly relates to a self-adaptive control method for the temperature of an all-electric melting glass kiln, which comprises the following steps: acquiring operation parameters of a plurality of heating sections; for any heating section, determining the electric heat conduction volatility according to the resistance difference at the current moment and the liquid level height deviation rate; according to the ratio of the discharging amount to the maximum yield of the kiln at the current moment, the ratio of the power to the rated power of the transformer and the charge conduction ratio, the thermal load response sensitivity is determined; and correcting the basic gain by using the electric conduction volatility and the thermal load response sensitivity, determining an adaptive power regulation value, and further controlling the voltage regulation of the transformer. According to the invention, through multi-dimensional physical parameter fusion and adaptive gain compensation, temperature control lag and oscillation caused by thermal inertia are effectively suppressed.

This invention provides a real-time visual trajectory correction system and control method for a robotic arm applied to electromagnetic induction heating forming of ship curved plates. The system includes the following steps: synchronously acquiring joint angles, images, and heating power and rate; real-time reconstruction of local point clouds in the heating area; using observed values ​​and process parameters as input, recursively estimating and predicting the height and normal vector of the curved plate surface; generating pre-correction commands based on online correction of the end-effector pose; simultaneously generating joint fine-tuning amounts based on visual ranging deviations; and finally, superimposing these to obtain the final joint control commands to drive the robotic arm. The system includes a data acquisition and point cloud reconstruction module, a deformation prediction module, a trajectory pre-correction module, a distance closed-loop fine-tuning module, and a command synthesis and execution module. This invention integrates process parameters into deformation prediction and coordinates with visual closed-loop control, achieving real-time and accurate trajectory correction even at extremely narrow operating distances.

This invention discloses a device operation process optimization system and device based on interactive behavior data, belonging to the field of human-computer interaction technology. The invention performs facial recognition on a designated device for the user to be operated on, recording all identified users as potential users. If several potential users exist, user filtering is performed to determine the primary user to complete the device operation; otherwise, the operation is directly performed based on the potential user. If user filtering is performed, multiple interaction judgments are conducted using camera interaction behavior data to determine whether to execute the human-computer interaction operation, thereby identifying the primary user. During the execution of the human-computer interaction operation, a simultaneous human-computer interaction operation judgment is performed to select the issued interaction behavior command. Based on the response of the potential users to the interaction behavior command, the corresponding primary user is determined to complete the device operation. This solves the problem in existing technologies where systems struggle to accurately identify target users, leading to incorrect identity matching and potential payment errors.