197 results about "Channel network" patented technology

The invention discloses a heterogeneous computing multi-target adaptive task scheduling method based on deep reinforcement learning, and the method comprises the following steps: S1, constructing a multi-dimensional dynamic perception model of a heterogeneous computing environment, and collecting and computing node performance indexes, task feature parameters and network states in real time; s2, defining a reward function as a multi-target weighted combination, fusing task completion time, energy consumption, resource utilization rate and cost, and dynamically adjusting the weight by a fuzzy comprehensive evaluation algorithm; s3, establishing a dual-channel deep reinforcement learning network architecture based on an attention mechanism; s4, establishing an adaptive exploration mechanism, combining an epsilon-greedy strategy and entropy regularization, and balancing exploration and utilization; the method has the beneficial effects that dynamic balance of multiple indexes such as task completion time, energy consumption and resource utilization rate is realized through combination of deep reinforcement learning and multi-objective optimization, a dual-channel network and a cross attention mechanism are adopted, and a task time sequence characteristic and a topological dependency relationship are modeled at the same time, so that a scheduling strategy is more accurate.

The invention relates to the technical field of hydrogen production, in particular to a bipolar plate runner structure-mass transfer efficiency parameter simulation optimization method. Comprising the steps that a flow channel structure is designed, specifically, a bipolar plate flow channel is constructed through a fractal tree-shaped network topological structure, the flow channel is composed of multiple stages of branch channels, the section size of each stage of branch channel is decreased progressively according to a self-similarity proportion, and a three-dimensional flow channel network with fractal dimensions is formed; multi-physics field coupling modeling: establishing a multi-physics field coupling model including fluid flow, electrochemical reaction and heat and mass transfer based on computational fluid mechanics and a finite element method, wherein the model considers a turbulence effect in a flow channel, interface impedance of a porous medium diffusion layer and reaction kinetic parameters at the same time; parameter sensitivity analysis: screening key parameters influencing mass transfer efficiency through orthogonal test design, including flow channel fractal dimension, branch angle, porosity and surface wettability parameters, and constructing a parameter response curved surface; the flow channel structure of the bipolar plate can be accurately optimized, and the mass transfer efficiency is effectively improved.

The invention provides a multi-sensor dynamic calibration method in rotary geosteering drilling, and relates to the technical field of petroleum engineering and drilling, and the method comprises the following steps: collecting original sensing data in real time through a multi-source sensor array deployed on a rotary geosteering drilling tool, wavelet packet decomposition is combined with an improved sliding window algorithm to carry out online denoising processing on original data, and a three-dimensional feature matrix containing environmental interference factors is established; and based on the three-dimensional feature matrix, constructing a dynamic calibration model based on depth time sequence association, respectively processing sensor body feature flow and environment interference feature flow by using a dual-channel LSTM network, performing dynamic weight distribution and fusion on dual-channel features by means of a gating attention mechanism, and outputting a dynamic deviation compensation coefficient of each sensor. The parameter set is generated and reconstructed through the three-dimensional feature matrix, the two-channel network and cooperative calibration, the drilling efficiency and precision are improved, errors are reduced, and intelligent development is promoted.

The invention discloses an intelligent drill bit wear prediction system and method based on multi-source perception, and relates to the technical field of crossing of artificial intelligence and intelligent manufacturing, and the method comprises the steps: synchronously collecting vision, vibration, temperature, pressure and engineering parameter data through a multi-source sensor, and carrying out the time-space alignment and preprocessing; extracting each modal feature by a multi-channel network, generating high-dimensional joint representation through cross-modal attention fusion, inputting a time sequence model formed by a gating circulation unit and an adaptive residual block, and outputting a wear loss prediction sequence and a wear type probability; the system comprises a data acquisition module, a preprocessing module, a feature extraction module, a fusion module, a prediction module and a risk decision module, parameter instructions are optimized, and early warning is realized when a threshold value is triggered. Through multi-source cooperative sensing, dynamic feature fusion and closed-loop decision control, the prediction precision, robustness and real-time performance are remarkably improved, the drilling safety is guaranteed, and the operation and maintenance cost is reduced.

The invention relates to the technical field of image analysis, in particular to a cushion foaming forming quality detection method and system based on machine vision, and the method comprises the steps: collecting image data, carrying out the iterative screening of the image data, and selecting a detection sample; the method comprises the following steps: collecting a three-dimensional point cloud of a detection sample, calculating the flatness of a seat cushion by using the three-dimensional point cloud, carrying out modeling by using polarized light reflection to obtain surface cell uniformity, and carrying out feature extraction on the three-dimensional point cloud through a three-dimensional mapping model to generate a three-dimensional semantic model; extracting mechanical characteristics by using the time sequence pressure map, and generating a cushion digital model; variational self-coding is carried out on the cushion digital model, and an abnormal feature map is generated; performing information extraction and pixel-by-pixel multiplication on the abnormal feature map by using a learnable convolutional layer to generate an abnormal enhanced map; and inputting the abnormal enhancement graph and the seat cushion digital model into a multi-channel network model for final evaluation. According to the method, the image data is analyzed, and the performance of the cushion is subjected to index quantification, so that the deep detection of the quality of the cushion is realized.

The utility model discloses an electrolytic bath bipolar plate with micropore flow channels and an electrolytic bath, a micropore flow channel network is arranged on the side surface of an anode of a plate body, and the micropore flow channel network comprises a plurality of groups of micropore flow channel units which are arranged at intervals along the width direction of the plate body; each micropore flow channel unit comprises a main flow channel arranged in the length direction of the plate body and a plurality of first-stage branch flow channels arranged in the length direction of the main flow channel at intervals, a plurality of short branch flow channels are arranged on the two sides of each first-stage branch flow channel at intervals, and a jet hole is formed in the top of each short branch flow channel. The jet flow holes are formed in the direction perpendicular to the side surface of the anode, the length of the short branch runner is smaller than that of the first-stage branch runner, and the short branch runner is perpendicular to the first-stage branch runner. And the microporous flow channel network adopts a structure of combining tree fractal and T-shaped nodes, so that the mass transfer uniformity of water flow in the microporous flow channels can be improved, and the water-membrane contact efficiency is improved, thereby ensuring the hydrogen production efficiency of the bipolar plate.

The embodiment of the invention provides a method, a system and equipment for predicting the corrosion rate of a water wall tube of a boiler and a storage medium. The method comprises the following steps: acquiring corrosion related parameters of the water wall tube, including electrochemical noise, surface strain, medium components and temperature and humidity; performing space-time alignment and fusion processing on the corrosion related parameters to obtain a corrosion state feature matrix; inputting the corrosion state characteristic matrix into a pre-established physical characteristic extraction model, and dynamically predicting the corrosion rate of the water wall tube by adopting a space-time dual-channel network; wherein the physical feature extraction model is pre-established through a physical information neural network, and the space-time dual-channel network comprises a long short-term memory network, a graph convolutional network and a sliding window. According to the embodiment of the invention, various factors influencing corrosion are comprehensively considered, the physical information neural network is introduced to carry out deep autonomous learning, finally, a reliable corrosion rate prediction condition is provided, and dynamic, lasting, intelligent and flexible corrosion state monitoring and prediction of the boiler water wall tube are realized.

The invention provides an optical and acoustic image registration method for underwater structure crack detection, and the method comprises the steps: collecting a multi-mode image through the vision field and time synchronization, building a one-to-one correspondence relation according to the pose and time metadata, and improving the image quality through the denoising, distortion correction and other means; optical and acoustic multi-scale features are respectively extracted through a dual-channel network, and cross-modal semantic alignment is realized in combination with a shared weight and a channel attention mechanism; generating a high-resolution scale offset field by using local cross-correlation and micro-upsampling, and introducing manifold regularization constraint to ensure that a vector field is smooth and continuous; according to the method, the local scale field and the affine parameters are combined, non-rigid space mapping is achieved through thin-plate spline interpolation, feedback iterative optimization based on edge structure consistency is assisted, the registration precision of a key structure is enhanced, the automatic registration effect of the underwater multi-modal image is improved, and the method has high robustness and practical application value.

The invention relates to a design method and system for a monolithic integrated microfluid cooling channel of a power device, and the method comprises the steps: obtaining the size structure data, hot spot density data and heat source data of the power device, and carrying out the analysis of the microfluid cooling requirements, and obtaining the channel design requirements and heat source distribution information; performing microfluid channel network construction and surface insulation sealing treatment on a preset wide bandgap semiconductor substrate according to the channel design requirements to obtain a microfluid three-dimensional channel structure; performing power device active region layout according to the microfluid three-dimensional channel structure to obtain a device integrated structure; performing cooling simulation evaluation on the device integration structure according to the heat source distribution information to obtain a cooling evaluation result; and performing iterative verification optimization on the device integration structure according to the cooling evaluation result to obtain a target design scheme. The heat dissipation performance of the power device can be effectively improved, and the size of an external heat dissipation system is reduced.

This disclosure provides methods, devices, and techniques to indicate operations by extremely high throughput (EHT) devices on an operating bandwidth, including devices in a basic service set (BSS) supporting the use of a 320 MHz channel. In some aspects, the supported functionality may include extensions to flexibility and support rules, structures, and signaling using legacy fields, frames, and features. In addition, the supported functionality may include channel sensing and reporting, such as per-channel network allocation vectors (NAVs) for the sub-channels of the operating bandwidth. A device may identify an operating mode for an operating bandwidth and determine a value for a bandwidth query report (BQR) or a target wake time (TWT) element. The device may check multiple NAVs for sub-channels of the operating bandwidth. The operating bandwidth may span concurrent operations on traditional Wi-Fi frequency bands including the 2.4 and 5 GHz bands as well as the 6 GHz band.

The invention discloses a cutting fluid drainage nozzle of a high-measurement slicing machine, which belongs to the technical field of cutting processing equipment and comprises a base, a drainage panel, an adjustable composite damping support module, a piezoelectric adjusting strip, a microcirculation flow channel network, a vibration sensing system and a control system. Comprising a high-molecular polymer damping material layer and a high-strength light alloy restraint layer, a piezoelectric ceramic actuator is embedded, passive damping energy absorption and active vibration counteracting are achieved, a piezoelectric adjusting strip is fixed in a drainage groove, and the dual functions of local vibration suppression and flow control are achieved; vibration sensing monitors the vibration state in real time through a plurality of piezoelectric acceleration sensors, the control system achieves working condition self-adaption and frequency division cooperative control through a neural network algorithm, cooperative work is achieved through multiple vibration reduction mechanisms, the vibration suppression effect is improved, the line breakage rate is reduced, meanwhile, the functions of fault diagnosis and predictive maintenance are achieved, and the reliability of the system is improved. And the cutting precision and efficiency and the equipment intelligence level are improved.

The invention discloses a micro-channel cooling method and system for semiconductor packaging, and belongs to the technical field of semiconductor packaging detection, and the method specifically comprises the steps: constructing a graded micro-channel network in a semiconductor packaging substrate, the graded micro-channel network is composed of a first-stage main channel and a second-stage branch channel with a smaller equivalent hydraulic diameter, and the first-stage main channel and the second-stage branch channel are communicated through a conical transition section; the pump pumps a cooling working medium into the second-stage branch channel through the first-stage main channel, and flow deviation caused by manufacturing tolerance is compensated through diameter difference. During boiling cooling, a cooling working medium absorbs heat in the second-stage branch channel to form a bubble core, and a periodic expansion or contraction structure in the channel regulates and controls bubble growth; a cooling working medium carrying bubbles enters an inertia damping structure comprising a sudden expansion cavity and a vortex generator for pressure stabilization and then enters an external condenser for phase change, the condensed liquid cooling working medium is injected into a primary main channel again through a pump, and efficient cooling circulation is achieved.

The invention discloses a small hole heat conduction and gas migration channel construction system and method based on seabed horizontal directional drilling, and relates to the technical field of ocean engineering and energy exploitation. A sea surface operation module is used for controlling all modules of the system to work cooperatively and supplying energy to the system; the seabed drilling guide module is used for realizing layer-entering drilling and channel pre-construction; the signal detection module is used for monitoring the spatial position and attitude parameters of the seabed drilling guide module in a seabed bottom layer in real time; the small hole perforation flow guide module is used for constructing a heat conduction and gas migration channel network after the drill bit enters a target hydrate layer; after the heat conduction and gas migration channel network is formed, the seabed drilling guide module continues to drill out of the seabed bottom layer, and a channel formed by the seabed drilling guide module is used for dragging a pipeline into a reservoir. A reusable mining channel suitable for the deep sea environment is constructed, the gasification efficiency is effectively improved, and stratum disturbance and energy consumption are reduced.

The invention discloses a hydrological sequence missing data complementation and trend prediction system and method, and belongs to the technical field of water conservancy information. The system comprises a data acquisition and preprocessing module, a spatial-temporal feature fusion module, a physical constraint completion module, an uncertainty prediction module, a credible evidence storage and verification module and a visual tracing module which are connected in sequence. The method comprises the following steps: collecting and preprocessing multi-source hydrological data; combining the spatial-temporal features through a dual-channel network to generate joint representation; generating and checking a completion sequence based on joint representation and coupling physical constraints such as a unit line method and a Manning formula; performing uncertainty trend prediction by using an integrated predictor; the key data of the whole process are linked and stored, and are automatically verified through an intelligent contract; and full-life-cycle visual tracing is provided. According to the method, the complementation robustness and accuracy in a high-missing-rate scene are effectively improved, the physical consistency of a complementation result is ensured, and credible evidence storage and transparent decision of the whole process are realized.

The invention relates to a traffic classification method based on a parallel multi-channel fusion graph neural network. The method comprises the following steps: S1, modeling an encrypted network traffic classification problem; s2, preprocessing the original data stream file to obtain a head part and a load part of each data packet; s3, modeling the preprocessed data packet byte sequence into a graph to obtain a flow graph of a head part and a load part; s4, inputting the flow diagrams of the head part and the load part into a parallel dual-channel graph neural network module to obtain a packet-level head vector representation and a packet-level load vector representation; s5, splicing the head vector representation and the load vector representation to obtain a complete data packet vector representation, and sending the three vector representations into a parallel three-channel network to obtain a flow level vector representation; and S6, performing feature fusion on the three vector representations, and obtaining a flow classification result through a final classification layer. Compared with the prior art, the method has the advantages of higher precision and the like.

The invention discloses a seismic strong background noise removal method based on a two-channel network, and relates to the field of geophysical exploration data processing, and the method comprises the steps: selecting a noiseless signal and background noise when the noiseless signal and the background noise are not excited from a single-channel seismic record after a seismic source is excited, carrying out the superposition, generating noisy data, and carrying out the normalization processing; constructing a dual-channel constraint denoising network composed of a global constraint sub-network and a denoising sub-network; optimizing network parameters through the training sample set; and performing global constraint denoising and slice local denoising on the noisy seismic data by using the trained model, and finally splicing to obtain a complete denoising result. According to the method, the splicing effect problem caused by traditional blocking processing is effectively solved, the strong background noise removal effect is remarkably improved, and seamless high-quality denoising with the complete edge is achieved.

The invention discloses a micro-fluidic chip-based trace tumor nucleic acid low-damage enrichment method and system, and the method comprises the steps: constructing a virtual flow channel network digital twin model of a multi-stage micro-fluidic chip, the virtual flow channel network digital twin model comprises enrichment units connected in series, each enrichment unit is composed of a tapered convergence section, a periodic bending section and an aptamer modified amplification section; and executing real-time simulation of fluid dynamics based on the digital twinborn model, generating a dynamic pressure control instruction set, and adjusting the flow velocity distribution of the biological sample in the flow channel through a negative pressure system driven by a fuzzy PID (Proportion Integration Differentiation) controller. According to the invention, through digital twin dynamic reconstruction flow channel and multi-modal fusion regulation, shear stress accumulation is inhibited to maintain nucleic acid integrity, impurity adsorption is reduced based on dynamic balance of probe density and binding saturation, and multi-field collaborative optimization is realized.

The invention discloses a three-phase wireless electric energy and signal reverse cooperative transmission system and a parameter design method thereof. The system comprises a direct-current power supply, a three-phase high-frequency inverter circuit, a three-phase primary side compensation capacitor, a three-phase secondary side compensation capacitor, a three-phase primary and secondary side magnetic coupling coil, a signal transmission channel network, a signal modulation-demodulation circuit network, an encoder, a decoder, a three-phase rectification filter circuit, a filter capacitor and a load resistor. The signal transmission channel networks respectively comprise three signal injection networks and three signal separation transmission channel networks; each signal injection network comprises a secondary side signal injection transformer, a secondary side injection signal resonant capacitor and a discharge resistor; and each signal separation transmission channel network comprises a primary side signal pickup transformer, an extraction signal resonant capacitor, a wave blocking network inductor, a wave blocking network capacitor and a sampling resistor. According to the invention, three-phase wireless energy and information parallel reverse transmission is realized, and the signal transmission rate is effectively improved while stable and efficient electric energy transmission is ensured.

The invention provides an InSAR (Interferometric Synthetic Aperture Radar) and GNSS (Global Navigation Satellite System) fusion settlement monitoring method based on a double-branch neural network, which comprises the following steps of: converting low-time-resolution InSAR data into a high-resolution time sequence matched with GNSS through an interpolation algorithm, and constructing an input matrix with aligned dimensions; a heterogeneous dual-channel network is adopted to respectively extract InSAR spatial features and GNSS time sequence features, after key information is extracted through dimension reduction processing, fusion data is reconstructed, a data splitting strategy is utilized to train a model and optimize parameters, and finally accurate modeling of a multi-source data nonlinear relation is realized. According to the method, the limitation of traditional linear fusion is broken through, a heterogeneous neural network is designed, multi-source data fusion is realized by using the high spatial resolution feature of the InSAR image and the high temporal resolution feature of the GNSS, the problem of insufficient modeling of a non-linear relationship of heterogeneous data in a traditional method is solved, and the space-time precision and robustness of settlement monitoring are improved.

The invention provides an intelligent audio power amplifier system and a working method thereof. The working method comprises the steps that a dynamic noise suppression module enables an input signal after reverse noise injection to enter a distortion compensation module; a distortion compensation signal is synthesized by using a distortion compensation module; the life cycle parameter self-calibration module calibrates parameters and inputs the parameters to the dynamic noise suppression module and the distortion compensation module; the dynamic noise suppression module is based on an intelligent learning method and adopts a frequency band channel network, a time sequence channel network and a fusion layer; the life cycle parameter self-calibration module provides the temperature and the working time for the dynamic noise suppression module to optimize the time domain channel network and provides the temperature and the working time for the distortion compensation module to update the predistortion coefficient of the dynamic nonlinear distortion model; the output of the dynamic noise suppression module is provided for the distortion compensation module; and the output of the distortion compensation module is fed back to the dynamic noise suppression module. The problems of high-frequency noise interference, nonlinear distortion degradation and insufficient long-term stability are solved at the same time.

The invention discloses a wavelength division multiplexer device and method compatible with channel monitoring, and relates to the technology of wavelength division multiplexers. According to the integrated wavelength division demultiplexer device with the wavelength division demultiplexing end adopting the waveguide deep etching Bragg grating, the functions of multi-channel network information transmission and network channel state monitoring can be realized at the same time; the wavelength division demultiplexer comprises a wavelength division demultiplexer common end input waveguide, a wavelength division demultiplexer, a wavelength division demultiplexer output waveguide and a Bragg grating which is written in the wavelength division demultiplexer output waveguide area by adopting a phase mask method. The device is high in integration degree and simple in structure. The channel monitoring method provided by the invention is realized based on the integrated wavelength division demultiplexer device, does not need to additionally distribute wavelengths for monitoring channels and independently distribute optical fibers for returning monitoring signals, is low in implementation difficulty compared with a software means for monitoring link node information, and can be applied to a complex optical communication network.

The invention discloses a series arc fault detection method based on DSCAN and feature fusion. The method comprises the following steps: collecting source domain data and target domain data; preprocessing data of a source domain and a target domain to obtain multi-dimensional features such as a grayscale image after GASF conversion, an IMF component after VMD adaptive decomposition, a frequency domain component obtained by DT-CWT and the like; constructing a four-channel feature fusion module, inputting the original current signals and the converted multi-dimensional features into corresponding channel networks, extracting signal features, and splicing the signal features; and constructing a DSCAN network to learn domain invariant features of a source domain and a target domain, and detecting a series arc fault through a model. Through multi-dimensional feature fusion and cross-domain transfer learning, arc fault features in a complex load scene are effectively extracted, the robustness of the model to different load environments is enhanced, the problem that a traditional method is insufficient in detection precision under mixed loads is solved, efficient and accurate detection of series arc faults is achieved, and the fault detection efficiency is improved. And a reliable technical scheme is provided for electrical safety monitoring.

This paper discloses a multi-scale parallel dual-channel method for identifying and grading pneumoconiosis lesions. The specific implementation process is as follows: the collected original X-ray pneumoconiosis image data is staged; a multi-scale attention mechanism is used to perform feature grouping and parallel convolution operations to enhance the model's receptive field and generate a corresponding attention weight map; the feature map extracted by parallel convolution is combined with a detail encoder with a residual structure to capture detailed texture feature information; to further grasp the invisible features in the feature map, a parallel dual-channel network module is constructed to effectively embed implicit local and global information and collaboratively capture pixel-level semantic information in high-frequency components; finally, transfer learning and multi-scale convolution are used to fuse the different channel information in the feature map and input it into the detail encoder and Transformer layer stack respectively to identify the pneumoconiosis lesion grade. Experimental results show that this method has higher recognition accuracy.

The invention discloses an employee performance evaluation method based on behavior data driving, and relates to the technical field of human resource management, and the method comprises the following steps: constructing a multi-channel network fluctuation prediction matrix, generating an expected behavior arrival time scale based on the transmission state of a task switching node, and taking the expected behavior arrival time scale as a baseline for behavior sequence gap recognition; and under the guidance of an expected behavior arrival time scale, performing sliding window scanning on the received behavior sequence, generating a gap identifier when detecting that a time slot exceeds a threshold value, and injecting a confidence coefficient weight descending label. According to the method, a behavior gap is accurately identified through network fluctuation prediction and sliding window scanning, behavior reconstruction is realized in combination with self-supervised generation and residual reasoning, and dynamic evaluation correction is completed by using confidence projection and an intelligent regulation engine, so that the data authenticity is guaranteed from the source, and the objectivity and robustness of performance evaluation and the resource allocation precision are improved.

The invention discloses a dense goods allocation warehouse-in and warehouse-out method and system based on an MAP map and a medium, and relates to the technical field of intelligent warehousing, and the method comprises the following steps: obtaining MAP map data, extracting channel information of a dense goods allocation area based on the MAP map data, and obtaining a channel network graph; constructing an AGV initial path based on the channel network diagram, and performing path oscillation evaluation on the AGV initial path to obtain an oscillation conflict point; building an obstacle avoidance area in the channel network diagram based on the channel position of the oscillation conflict point; discretizing the AGV initial path in the obstacle avoidance area by using an ORCA algorithm to obtain an optimized moving path; the AGV is controlled to execute the warehouse-in and warehouse-out operation of the dense goods allocation based on the optimized moving path, and the problem that in the dense goods allocation environment, path oscillation and collision are likely to happen when the AGV executes the warehouse-in and warehouse-out operation is solved.

The invention discloses a multi-modal signal identification method based on an additive attention mechanism, and the method comprises the following steps: inputting data, and carrying out the preprocessing operation of a data set; designing a multi-channel network structure; carrying out multi-modal fusion processing; an LSTM layer and an attention mechanism are added to extract sequence time sequence features, data are fused through multiple channels and multiple modules, further feature extraction of the data is achieved through a bidirectional Lstm layer and an attention mechanism module, and the characterization capacity of signals is enhanced; outputting the classification; compared with verification results, the method has the technical characteristics that key features can be enhanced, so that the method can better adapt to signal conversion in a complex environment, the recognition accuracy is improved, and the robustness of a system under noise and interference conditions is enhanced.

The invention discloses a station hub passenger flow load and flow composition prediction method and system, and the method comprises the steps: generating OD demands from an exit to a destination of each vehicle based on the flow of each exit and a traffic mode sharing rate predicted by a comprehensive hub prediction model; obtaining channel basic data in the hub, constructing a directed graph representing a channel network, and calculating a channel capacity attribute of each edge; for each OD pair in the OD demand, calculating corresponding OD trajectory data in combination with the directed graph; calculating the load rate of the channel according to the load attribute of the edge in the OD track data and the channel capacity attribute of the edge; and calculating the flow composition of the passenger flow according to the flow information of the nodes in the OD trajectory data. According to the method, the whole dynamic flow process from arrival to departure of the passenger flow in the station and traffic mode conversion can be predicted, so that the passenger flow distribution and dispersion can be better performed during congestion according to the prediction result, the passenger flow load pressure in the station in the peak period is reduced, and the occurrence probability of safety accidents is reduced.

The invention discloses a method for quickly constructing a transient hydraulic model of a canal network. The method comprises the following steps: firstly, numbering open channel natural sections in a channel network, respectively determining hydraulic parameters of different open channel natural sections, and uniformly summarizing section numbers of different open channel natural sections and node numbers corresponding to the section numbers; determining a topological structure form and a boundary change condition of the channel network; then according to the topological form of the channel network, respectively determining a corresponding structure topological incidence matrix, an open channel water depth incidence matrix, a water depth incidence matrix in a branch point continuity equation and a flow incidence matrix in the branch point continuity equation; and finally, constructing a transient hydraulic mathematical model of the canal network, giving boundary change conditions, and simulating the hydraulic transition process state of the canal network based on the transient model. The method provided by the invention is high in applicability, can theoretically simulate the transient transition process change of any topological structure channel network, and has a great application prospect.

This invention belongs to the field of artificial intelligence and system response calculation technology, and provides a method and system for calculating system response based on a generalizable physical information neural network. The technical solution involves establishing a family of differential equations consisting of multiple governing equations based on the physical characteristics of the system to be analyzed, and obtaining the corresponding dataset; constructing a neural network model consisting of a main network and multiple channel networks, where the main network learns the basis functions representing the system response and shares them with all channel networks; each channel network processes a specific differential equation task within the task family and learns the basis function coefficients corresponding to that task; inputting the dataset into the neural network model, training the model by minimizing a loss function containing multiple constraints, and obtaining the optimal solution for the system response under different stimuli based on the trained model. This method can solve complex differential equations more accurately and improve the model's generalization performance on unseen tasks.

The present invention discloses a multi-channel neural network model training method, device, and computer storage medium. The multi-channel neural network model training method includes: when training a target model, inputting training data into an N-channel neural network model for learning. The model adopts a multi-channel network structure, including a residual network channel and N-1 densely connected network channels with different connection structures. The convolutional layers in these N-1 DenseNet channels are densely connected according to the corresponding number of interval layers. In this way, the ResNet channel can well reuse model parameters, improving the efficiency of the model. At the same time, the N-1 DenseNet channels can utilize the intermediate results of the previous layers, greatly improving the computational efficiency and performance of the model.