8results about How to "Reduce bandwidth costs" patented technology

The invention discloses a damper real-time performance evaluation and health diagnosis system and method. Current traditional damper monitoring faces many limitations. The system comprises a sensor array module, a data acquisition and preprocessing module, a data fusion processing module, a performance evaluation module, a health diagnosis module, an early warning module and a communication and storage module. The data fusion processing module is used for performing automatic fusion processing on the preprocessed data to generate optimized fusion data; the performance evaluation module is used for calculating various performance indexes of the damper through the fused data; the health diagnosis module diagnoses the fault type of the damper according to the fused data and the performance evaluation result, analyzes the fault reason and evaluates the risk level of the fault; the early warning module is used for issuing early warning information; and the communication and storage module is used for data transmission and storage. According to the invention, the accuracy, real-time performance and reliability of damper monitoring are improved.

The invention discloses a battery maintenance method based on a dynamic sparse neural network and a large language model, and belongs to the technical field of battery management. The method comprises the following steps: acquiring sensor time sequence data and an operation and maintenance text maintenance log of a battery pack, and respectively carrying out standardized preprocessing on the sensor time sequence data and the operation and maintenance text maintenance log; extracting time domain and frequency domain features of the preprocessed sensor time sequence data to obtain sensor features; carrying out semantic coding on the preprocessed text maintenance log by adopting a large language model subjected to battery domain knowledge fine adjustment to obtain text semantic features; performing feature fusion on the sensor features and the text semantic features based on a cross attention mechanism to generate fusion features; and inputting the fusion features into a dynamic sparse neural network which adaptively adjusts the sparseness of each layer of network according to the real-time health state of the battery, and outputting a battery fault probability and a fault root cause weight.

The present disclosure relates to a file processing method and device and a storage medium. The method comprises: in the case of detecting a starting instruction for a to-be-started service container in a container cluster, obtaining a file acquisition request for the to-be-started service container; parsing a target file domain name and a target download address from the file acquisition request; determining whether a candidate file domain name matched with the target file domain name exists in a configuration file of the container cluster; in the case of the candidate file domain name matched with the target file domain name existing in the configuration file, determining a candidate cache address corresponding to the candidate file domain name matched with the target file domain name as a target cache address; determining a target cache space from the container cluster based on the target cache address; and determining a required target file from the target cache space based on the target download address. In this way, the file can be directly acquired from the container cluster quickly and conveniently, and the file pulling speed in the container cluster environment is effectively improved.

This invention relates to the field of port data processing, and more particularly to a green port energy digital management platform based on digital twins. When constructing the digital twin, this invention utilizes an incremental synchronization mechanism and spatiotemporal event identifiers to ensure that the green port energy digital management platform only transmits incremental data that has changed. This not only saves on expensive cloud resources and dedicated line bandwidth costs, but also allows the green port energy digital management platform to maintain smooth operation even when dealing with massive amounts of data. Furthermore, through the incremental synchronization mechanism, this invention ensures that the green port energy digital management platform only updates the digital twin when changes occur, and combined with an efficient event bus, latency is drastically reduced from minutes to seconds.

The application provides a front-end reasoning deployment method and system based on a TensorFlow lightweight model, comprising: in a client, creating a background thread independent of a main thread; in the background thread, performing a model loading step, the model loading step comprising: judging whether a new version of a deep learning model needs to be obtained from a server through a local cache verification mechanism, and obtaining or using the deep learning model cached locally according to the judgment result; the main thread and the background thread interact through an asynchronous communication mechanism to perform reasoning calculation based on the deep learning model in the background thread, and send the reasoning result from the background thread to the main thread. The application realizes low-delay real-time reasoning: the reasoning process is completely completed in the local client of the user, and data does not need to be uploaded to the server, eliminating the delay caused by network transmission and greatly improving the real-time performance of artificial intelligence functions.

Embodiments of the present application disclose a log management method and device, a storage medium and an electronic device. The method comprises: extracting a first instruction address in a call stack, the first instruction address being an address corresponding to a first instruction of a preset operation in the call stack; analyzing the first instruction address to obtain a target original address, the target original address pointing to an original instruction address of the first instruction in a symbol image; extracting second instruction addresses in turn; for each second instruction address, determining a first feature value and a second feature value, the first feature value being a distance difference value of an original instruction address of a corresponding instruction in the symbol image relative to the target original address, and the second feature value being an instruction quantity difference value of the corresponding instruction relative to the first instruction; and generating a log according to the target original address, the first feature value and the second feature value corresponding to each second instruction address. The present application can reduce the amount of log data.

The invention discloses a two-stage compression denoising imaging method for a large-area array Geiger mode APD (avalanche photo diode). The method comprises the following steps: intercepting photon events in a pre-sampling time window at the front part of a data block to carry out global histogram statistics, fitting and estimating time domain characteristics of background noise based on Poisson distribution, screening and compressing full-frame data, and eliminating most noise; carrying out local joint histogram statistics and peak value extraction on the compressed data based on a pixel space neighborhood, and reconstructing a preliminary distance image and a preliminary intensity image; if the effective data in the neighborhood of the data meets the flat area condition, standard median filtering is executed; otherwise, starting a robust filtering algorithm based on absolute median difference MAD for processing. According to the method, the cooperation of front-end data efficient compression and rear-end accurate denoising and edge preserving is realized, and the method has the advantages of high compression ratio, high adaptability and excellent imaging quality, and is particularly suitable for real-time three-dimensional imaging of a large-area-array GM-APD laser radar system.

This application discloses a video file distribution method and apparatus. The method involves: acquiring change messages for various media assets on a video platform; updating the media asset copyright information of each media asset based on the change messages; traversing video files that have passed the video platform's review; determining the target copyright playback region information of the video file based on the target media asset copyright information matching the video file; and triggering the video file to be pre-warmed to the CDN node of the corresponding target copyright playback region according to the target copyright playback region information. This method distributes video files to CDN nodes in the target copyright playback region, thereby ensuring that only videos allowed to be played in overseas countries and regions are distributed to the corresponding overseas country / region origin server. This minimizes the bandwidth costs incurred in distributing video files overseas and the storage costs of files on overseas origin servers while guaranteeing the user's viewing experience.