4results about How to "Avoid confrontation" patented technology

The invention belongs to the technical field of mobile communication, and particularly relates to a mobile communication terminal network switching method and system, and the method comprises the steps: collecting multi-dimensional network parameters and terminal state parameters; judging the working mode of the terminal and dynamically distributing parameter weights; pre-judging a source network state through a machine learning model, and screening a target network; establishing a pre-connection channel, and executing differentiated seamless switching; generating a feedback result, inhibiting ping-pong switching and optimizing the model; and multi-terminal collaborative switching is supported. The system comprises functional modules on a terminal side and an optional cloud auxiliary module. Core logic of multi-dimensional decision making, active pre-judgment and seamless switching language feedback optimization is adopted, switching precision, real-time performance and low power consumption are achieved, switching delay is smaller than or equal to 50 ms, the ping-pong switching occurrence rate is reduced by more than 80%, hardware does not need to be newly added, and the method is suitable for various mobile communication terminals and application scenes.

The utility model relates to a kind of filter screen mounting structure and railway vehicle air conditioner, including rectangular filter screen frame, the filter screen frame is inserted into card slot and is installed, at least the spring piece is installed on the side of the filter screen frame, the fixation between card slot is realized by the deformation of the spring piece, the both ends of the spring piece have mounting hole for being fixed with filter screen frame by fastener, the middle body part of the spring piece is composed of three bending, three segment shape is W-shaped as a whole, there are two downward recessed accommodating cavities on the filter screen frame, the two bending points of the bottom of the spring piece respectively sink into corresponding accommodating cavity after deformation downward.This utility model can make filter screen disassembly more smooth, spring piece is not easy to damage, can improve the maintainability and reliability of product.

The invention discloses a coating workshop resource scheduling method based on multi-source heterogeneous data and federated learning, and belongs to the technical field of coating workshop resource scheduling. According to the method, multi-source heterogeneous data such as equipment operation states, material circulation and environmental parameters are collected in real time; then constructing a federated learning framework, training a global resource scheduling model by each client by using locally collected data, and aggregating model parameters by a central server to generate an optimized global model; dynamically generating a resource scheduling instruction according to real-time workshop state data by utilizing the optimized model, and automatically adjusting an equipment task, a logistics path and a production queue; and finally, an instruction is issued to execution equipment, the production progress and key performance indexes are monitored in real time through a visual interface, early warning is carried out on abnormity, and closed-loop control is formed. According to the invention, data privacy of each production unit can be guaranteed, the resource utilization rate and the production efficiency of a coating workshop can be obviously improved, and the dependence on manual operation and the error rate can be reduced.

The invention discloses an ultrasonic sensor environment compensation method and system based on multi-modal feature fusion, and the method comprises the steps: starting a target ultrasonic sensor, collecting ultrasonic signal features through an ultrasonic detection unit, and synchronously collecting multi-dimensional environment features through an integrated environment sensing module, all the features are aligned and stored according to a preset timestamp. And calling a weight distribution mechanism, determining an environment feature weight benchmark in combination with the current working parameters of the sensor, performing dynamic adjustment, and fusing the weighted environment features and the ultrasonic signal features to generate a fusion feature set. And inputting the fusion feature set into a pre-training error compensation model, predicting an ultrasonic propagation feature offset and outputting a propagation speed correction parameter. And the sensor calculates a compensated measurement value according to parameters, dynamically updates a related process according to a preset period to realize continuous compensation, and synchronously updates model reference parameters to adapt to an adjustment scene during range calibration. And the high-precision detection requirement is met.