103 results about "Orthogonal decomposition" patented technology

The invention relates to the technical field of refrigeration house energy-saving control and digital twinning, in particular to a refrigeration house fan energy-saving operation method and system, and the method achieves the whole-field temperature deduction of a sensor-free area by constructing a CFD reference model and simulating and predicting the three-dimensional transient airflow and temperature distribution in a refrigeration house. Model parameters are calibrated through measured data, and the order of the model is reduced by adopting intrinsic orthogonal decomposition and Galerkin projection, so that the calculation amount is remarkably reduced, and online optimization is supported. And in combination with a data assimilation algorithm, the prediction result of the reduced-order model is continuously corrected, and the adaptability to environment change and system aging is improved. The optimization control stage takes minimization of the total energy consumption of a fan as a target, solves an optimal fan control sequence under the hard constraint that the whole-field temperature does not exceed the cargo safety upper limit and does not exceed the cargo safety lower limit, and adopts a rolling time domain mode for execution to realize collaborative optimization of safety and energy conservation.

The invention relates to a trajectory prediction model robustness enhancement method based on dynamic subspace projection decomposition. Comprising the following steps: firstly, extracting hidden layer semantic features containing historical tracks and map topology through a multi-modal feature encoder; secondly, constructing a dynamic routing mechanism based on scene self-adaption, and calculating projection weights of input features on a plurality of expert subspaces; then, executing truncation projection operation based on orthogonal decomposition, retaining core semantics located in a low-dimensional space, and filtering out adversarial disturbance located in an orthogonal complementary space; and finally, introducing a feature consistency constraint training mechanism, taking the reconstructed features of the clean sample as anchor points, and compulsively aligning the purified features of the confrontation sample with the anchor points. Compared with the prior art, the method has the advantages that the robustness of the model in white box gradient attack, black box query attack and physical semantic deception scenes is remarkably improved through feature purification of a physical level and structured consistency constraint, and the prediction reliability of the automatic driving system is ensured.

The invention relates to the technical field of face verification, and discloses a face identity verification data processing method based on dynamic feature extraction, which comprises the following steps: acquiring a continuous face video frame sequence and calculating a full-pixel instantaneous velocity vector to generate an original dense optical flow field, selecting rigid region anchor points to calculate a rigid affine transformation matrix and construct a theoretical rigid motion field, performing differential stripping on the theoretical rigid motion field from the original dense optical flow field, and extracting a non-rigid micro-motion residual field; mapping the non-rigid micro-motion residual field to a facial muscle topological grid to generate a time sequence feature tensor, and calculating a geodesic line distance between a covariance matrix of the time sequence feature tensor and a reference dynamic feature in a Riemannian manifold space; when the geodesic distance is smaller than a threshold value, verification is passed, through a rigid-non-rigid orthogonal decomposition mechanism, the special viscoelastic micro-motion and cooperation law of biological soft tissue is captured by utilizing a residual field, and the high-simulation mask is effectively defended.

The invention discloses a monitoring equipment fault rapid positioning method based on causal diagram reasoning, and the method comprises the following steps: building a thermal diffusion model composed of equipment nodes and links through collecting the topological structure and operation data of a monitoring system; normalizing and fusing the operation deviation information into a node abnormal energy value, and establishing an abnormal energy propagation equation; when an alarm is triggered, collecting an energy change sequence of each node to generate a space-time energy distribution matrix, and reversely solving a thermal diffusion equation to calculate a disturbance source item; performing clustering backtracking on disturbance source items, determining fault root cause nodes, and comprehensively evaluating confidence; and when multi-source abnormity exists, thermal field orthogonal decomposition is executed, and model parameters are adaptively updated according to field results. According to the method, abnormal propagation of monitoring equipment is modeled into a thermal diffusion process, and reverse solution and adaptive calibration are combined, so that rapid, explainable and high-precision positioning of fault root causes in a complex monitoring network is realized.

The invention discloses a flow field prediction-oriented full-process automatic joint reduced-order modeling method, which comprises the following steps of: specifying a target physical field parameter space, and randomly generating a sample space according to a Latin hypercube sampling method; constructing a full-process automatic simulation tool chain, driving target physical field numerical calculation and generating a training data set; carrying out singular value decomposition-based intrinsic orthogonal decomposition on the output physical field data, and only retaining first r main feature components to construct a reduced-order data set; constructing a multi-input multi-output full-connection feedforward neural network, and modeling and training a nonlinear mapping relation between input parameters and reduced-order features; new working condition parameters are input, reduced-order features are predicted through the trained neural network, distribution of a target physical field is reconstructed according to a singular value decomposition reduction matrix, and more flexible and reliable technical support is provided for reducing the training cost of a reduced-order model and improving simulation efficiency.

The invention discloses a mountainous area wind field downscaling optimization method based on high-precision simulation, and relates to the technical field of meteorological numerical simulation, and the method comprises the steps: obtaining the terrain elevation data and meteorological monitoring data of a target mountainous area, recognizing slope abrupt change points, constructing a spiral sampling path, calculating a slope change value, determining the distance between sampling points, and generating target sampling data; performing orthogonal decomposition on the sampled data to obtain a frequency component, calculating a surface fluctuation coefficient, and constructing a boundary disturbance equation to obtain surface stress distribution; calculating an airflow motion state based on surface stress distribution, solving a vorticity equation to obtain vorticity characteristics of a leeward area, and calculating airflow motion correction parameters; and carrying out downscaling iterative operation on the corrected parameters and the meteorological monitoring data to generate target area wind field data with hectometer-magnitude spatial resolution. The simulation precision of the local wind field under the complex terrain condition is improved.

The invention relates to the technical field of digital twinning, physical modeling and multi-source data fusion, and provides a digital twinning dynamic construction method based on multi-source data fusion and physical simulation. The method comprises the following steps: acquiring a multi-source heterogeneous data stream from a preset sensor array, a numerical simulation result and a historical database, identifying a key feature mode of a dominant physical process in the multi-source heterogeneous data stream, acquiring a key feature mode time-varying physical field evolution rule corresponding to the key feature mode by using a time sliding window and a forgetting mechanism, extracting a low-dimensional sparse characteristic parameter set reflecting dynamic behaviors from a high-dimensional observation space, and constructing a reduced-order proxy model by adopting Gaussian process regression, a neural network proxy model or an intrinsic orthogonal decomposition combined interpolation technology; and receiving a corresponding real-time observation data stream to establish a full-closed-loop feedback link from model prediction, high-fidelity solution verification to observation data correction in combination with the reduced-order proxy model so as to complete the construction of the digital twin.

The invention discloses a summer rainfall sub-season prediction method and system fused with multi-scale deep learning, and the method comprises the steps: collecting multi-source weather forecast data and observation data, and carrying out the empirical orthogonal decomposition of the observation data, and obtaining a rainfall main mode and a mode sequence; performing multi-scale signal extraction on the observation data, and performing correlation analysis on the observation data and the modal sequence to obtain respective weight fields; constructing and training a deep learning model fusing a multi-pole attention mechanism and time sequence decomposition; inputting the forecast data into the trained model to carry out transfer learning, and optimizing the model; and substituting forecast data of preset time into the trained model to generate a high-quality summer rainfall sub-season forecast product. According to the method, the synergistic effect of sea, land and gas and the interaction of multi-scale signals are fully considered, the model is constructed based on an artificial intelligence method and a numerical model forecasting product, the sub-season forecasting skill of summer rainfall is effectively improved, and the method plays an important role in disaster prevention and reduction.

The invention discloses a truss type load-bearing structure time domain dynamics topological optimization method and system based on an intrinsic orthogonal decomposition method, and relates to the technical field of truss type load-bearing structure topological optimization. The method is provided for solving the problems of high calculation power and the like caused by large-scale space-time response iteration of original transient problems and intensive calculation of dual problem sensitivity analysis in the prior art. According to the method, a local geometric control strategy for the cross interference problem of rod piece units is introduced, and a transient dynamics full-order topological optimization model of the truss structure is established through an equal geometric stiffness diffusion method. Based on an intrinsic orthogonal decomposition method, a transient dynamics reduced-order model of a truss structure is constructed, a reduced-order model base is updated in real time through an incremental singular value decomposition method, and then efficient solving of a transient dynamics equation and an adjoint equation in sensitivity analysis is achieved. And updating an iterative design variable by adopting a penalty function method, and seeking the optimal layout design of the truss structure. The invention provides a time-domain dynamics topological optimization method of the truss type load-bearing structure based on an intrinsic orthogonal decomposition method, and the method can be used for efficiently solving the dynamics layout optimization problem of the truss structure under the action of a transient load.

The invention relates to the technical field of fluid mechanics modeling and aerodynamic acoustic analysis, in particular to a spectral orthogonal decomposition-based separated flow wall surface pressure pulsation dominant structure modeling method, which comprises the following steps of: acquiring a whole flow field and / or wall surface unsteady pressure data of a separated flow under a preset working condition; performing spectral orthogonal decomposition on the unsteady pressure data, performing frequency domain decoupling on multi-scale features, and extracting feature values and spatial dominant feature modes under each feature frequency; constructing a complex wave packet physical parameterized model, fitting the spatial dominant feature modals, and compressing the spatial dominant feature modals into sparse physical parameter vectors after nonlinear regression solution; and reconstructing a wall surface pressure pulsation dominant component based on the sparse physical parameter vector and the characteristic value, and carrying out at least one of flow mechanism analysis and pneumatic order reduction modeling according to the dominant component. Therefore, the problems of low reconstruction precision and the like caused by the fact that asymmetric evolution and variable acceleration convection of a large-scale structure in separation flow wall surface pressure pulsation cannot be accurately represented in a frequency domain in related technologies are solved.

The invention relates to the technical field of wind power plant power control, in particular to a wind power plant dynamic power distribution optimization method based on projection dimensionality reduction. The method comprises the following steps: collecting a high-dimensional original operation data stream of a wind power plant, and preprocessing to output a high-dimensional state vector; extracting a dominant mode matrix through an intrinsic orthogonal decomposition algorithm, and projecting a high-dimensional state vector to a low-dimensional space; future modal coefficient evolution is predicted through an autoregression prediction model, and a low-dimensional rolling optimization problem is constructed and solved through a sequential quadratic programming algorithm; reconstructing the optimal low-dimensional modal coefficient vector into a target power instruction of each fan, and issuing and executing the target power instruction; and updating the dominant mode matrix through an incremental singular value decomposition algorithm and correcting parameters of the autoregressive prediction model. According to the method, projection dimensionality reduction from high-dimensional data to a low-dimensional space is realized through intrinsic orthogonal decomposition, and the problem of low optimization solution efficiency caused by high data dimensionality of a traditional method is solved.

The invention relates to the technical field of circuit load state monitoring, and discloses a directional drilling pipeline surface coating damage characteristic analysis method based on data inversion, which comprises the following steps: injecting a dual-frequency composite excitation signal into a pipeline transmission loop; locking a current phase, collecting a magnetic flux gradient vector and performing orthogonal decomposition; according to the method, impedance distortion caused by medium heterogeneity is eliminated through the pilot frequency response difference, the problem of misjudgment caused by the fact that traditional detection depends on medium isomorphism hypothesis is solved, and the detection accuracy is improved. And accurate identification of transmission load leakage impedance abnormity is realized.

The invention belongs to the technical field of hydraulic engineering digital twinning, and provides an arch dam static response rapid calculation method which comprises the following steps: sampling and simulating based on a finite element model in an offline stage to obtain a displacement field snapshot matrix, extracting a dominant mode through intrinsic orthogonal decomposition to construct a reduced-order subspace, and calculating a mode coefficient; and training a radial basis function neural network to establish a nonlinear mapping relation by taking the working condition parameters as input and the modal coefficient as output, and finally, inputting the target working condition parameters into the trained network to predict the modal coefficient in an online stage, and quickly reconstructing a complete displacement field response through linear combination with the POD modal. The method achieves the quick and accurate prediction of the static physical field of the arch dam, and remarkably improves the calculation efficiency of the displacement field of the arch dam.

The invention discloses a micro-channel heat exchanger topological optimization design method, which comprises the following specific steps of: 1, executing key working condition parameter sampling in a design space, constructing a topological optimization mathematical model, carrying out iterative computation, and establishing a topological optimization database comprising a topological field and corresponding temperature field, flow field and pressure field data; 2, realizing unified order reduction of a topological field and each physical field by adopting intrinsic orthogonal decomposition, and converting a complex high-dimensional structure into low-dimensional modal representation; 3, constructing a multi-task learning neural network, taking the multi-target constraint parameters as input, taking the topological field modal coefficient and each physical field modal coefficient as output, and adopting a PyTorch deep learning framework to train a topological structure generation model; and 4, a model is generated based on the trained topological structure, the high-resolution topological structure can be predicted within second-level time, and the thermal-fluid performance of the high-resolution topological structure can be synchronously fed back. The method has a remarkable calculation speed-up ratio and good engineering mobility, and an efficient solution is provided for micro-channel heat dissipation design of a high-power electronic device and a new energy system.

The invention discloses an extra-high voltage transformer bushing transient temperature rise rapid prediction method, which comprises the steps of constructing a simulation model of an extra-high voltage transformer bushing, and simulating and extracting global node temperatures and corresponding working condition parameters of a bushing profile at multiple moments to form a data set; selecting global node temperature to perform orthogonal decomposition, and selecting first r-order modals to construct a modal matrix according to cumulative energy proportion truncation; projecting global node temperature in the data set to a modal space according to a modal matrix to obtain a modal coefficient vector as an output sequence, and performing sliding window on working condition parameters in the data set to form an input sequence; constructing a long-short-term memory network model, optimizing network hyper-parameters of the long-short-term memory network model, and training the model by adopting an input sequence and an output sequence to obtain a time sequence mapping model; and acquiring working condition parameters of the extra-high voltage transformer bushing in a to-be-measured time period, inputting the working condition parameters into the time sequence mapping model, predicting a modal vector at a future moment, and reconstructing transient temperature field distribution of the extra-high voltage transformer bushing in combination with inversion of a modal matrix.

The invention discloses a rapid prediction method for a particle thermal-mechanical coupling process, and the method comprises the steps: scanning energy carrier particles based on a microfocus computed tomography technology, and generating a three-dimensional digital model; importing the three-dimensional digital model into a multi-physics field simulation platform, and establishing a transient full-order model; based on the transient full-order model, generating a training data set and a test data set under different working conditions and corresponding temperature fields and stress fields; performing decomposition calculation on the training data set by using an intrinsic orthogonal decomposition POD reduced-order model to obtain a POD primary function and a spectral coefficient; establishing a mapping model from an input working condition to an output spectral coefficient based on a BP neural network; inputting the target working condition into the mapping model, and outputting a POD spectral coefficient; and on the basis of the calculation result and the POD spectral coefficient, reconstructing full-field physical quantity distribution at any moment under the target working condition through linear combination. The method solves the problems that traditional numerical simulation is low in calculation efficiency and large in resource consumption.

The application discloses a kind of offshore wind farm automatic power generation control flow field dynamic modeling methods, comprising the following steps: data sampling is carried out using CFD simulation, obtains data set;Data is collected and arranged;Eigen-orthogonal decomposition is used for order reduction processing;Offshore wind farm automatic power generation control flow field dynamic model based on yaw, variable pitch, speed regulation process and its disturbance process is constructed.The application can effectively overcome the shortcomings that the data set obtained by high-precision wind farm flow field dynamic simulation is large in structure and difficult to apply directly, and the dynamic reduced-order model of the wind farm flow field obtained can greatly compress the data structure, retain the main characteristics of the flow field, and has high wake prediction accuracy.

This invention relates to the field of engineering structure performance prediction and optimization technology, specifically disclosing an engineering full life cycle performance prediction and optimization system. The system collects time-series data of load, acceleration, and strain of engineering structures under extreme loads; constructs a physical consistency mapping operator based on real-time input data and finite element prior information, outputting a predicted response field; performs intrinsic orthogonal decomposition on the finite element prior information to extract the basis vector and integral point subset, trains the time-series evolution rule of the basis coefficients, and outputs low-dimensional basis coefficient prediction values; compares the real-time input data with the predicted values, and triggers online basis assimilation update when the deviation exceeds a threshold, assimilating the low-dimensional basis coefficients to output a corrected stiffness field; substitutes the corrected stiffness field into the time-series evolution rule to extrapolate the future response field, generating a smooth control force command for the active mass damper; this invention achieves real-time closed-loop optimization control of engineering structures under extreme loads.

The application relates to the technical field of industrial equipment state monitoring, and particularly discloses a fault prediction and health management system based on an industrial robot, which collects robot benchmark operation data and real-time operation data, and constructs a benchmark data set containing individual identity labels; multi-domain feature extraction and weighted fusion are performed on the data, a high-dimensional feature space is obtained through phase space reconstruction; the high-dimensional features are decomposed into mutually orthogonal individual attribute subspaces and degradation state subspaces by using an orthogonal subspace learning algorithm, and pure benchmark degradation features are obtained; a health index is constructed based on the benchmark degradation features, and a segmented continuous degradation model is established; real-time features are projected into the degradation state subspace to obtain real-time degradation features, which are input into the degradation model to invert the remaining service life and output graded early warning information; the application effectively suppresses false abnormal alarms by stripping individual difference interference through orthogonal decomposition of the feature space, and improves the cross-device generalization capability and prediction accuracy.

This application relates to a method, system, device, and medium for measuring pressure distortion in an aero-engine inlet. The method includes: acquiring typical flow field categories and corresponding intrinsic orthogonal decomposition modal basis functions (IOD) of the target aero-engine inlet; based on optimization objectives and constraints, solving for the measurement positions of the aerodynamic interfaces of the target aero-engine inlet using IOD to obtain a pressure sensor measurement position arrangement scheme; acquiring sparse pressure data collected by each pressure sensor deployed according to the pressure sensor measurement position arrangement scheme; determining the current typical flow field category based on the sparse pressure data, and reconstructing the sparse pressure data using Gappy orthogonal decomposition based on the IOD corresponding to the current typical flow field category to obtain the global total pressure distribution; and calculating the performance parameters of the inlet pressure distortion based on the global total pressure distribution. This method enables accurate evaluation of pressure distortion characteristics.

The application discloses a kind of bridge pier digital twin intelligent monitoring method and system based on agent model, it is related to bridge pier state monitoring technical field.The method includes the following steps: constructing the finite element model of the bridge pier to be monitored;The intrinsic orthogonal decomposition method is used to reduce dimension and extract features to input-output data set;The agent model is constructed and trained using the Gaussian process regression method, an adaptive particle swarm optimization and a nested cross-verification strategy are used to verify the agent model during the training process, and the optimal hyperparameters are obtained;The agent model is retrained using the optimal hyperparameters, and the final bridge pier digital twin intelligent monitoring agent model is obtained.The application aims to predict and evaluate the health status of the bridge pier under various complex external conditions such as horizontal stress, scour depth, vibration, inclination and settlement with high precision, speed and visualization.

The invention discloses a method applied to lathe CX coordinate conversion analysis, which comprises the following steps of: (a) capturing commands and feedback waveforms of C and X in a machining process by using a new-generation analysis platform tool, and loading a file by using a CX analysis conversion tool loading file button; (b) inputting shaft names to be converted, and respectively filling corresponding axial actual shaft names; (c) setting cutter length data H, (d) setting start and end of a time period, loading a processing file through a program, and preliminarily judging problem paragraphs needing to be processed; and (e) performing orthogonal decomposition on each time point to obtain XY components of the current time point. The direction of the problem can be preliminarily clarified, and a system control problem or a mechanism problem can be quickly judged; after conversion, an analysis platform is used for carrying out 2D processing and path analysis, so that problems can be conveniently found; and controlling variable adjustment parameters or data subjected to mechanism conversion to be loaded together, and searching a problem rule by using an overlapping graph analysis tool.

The invention relates to the technical field of high-speed micro pumps, and discloses a high-speed micro centrifugal pump impeller profile pre-compensation method based on fluid-solid coupling analysis, which comprises the following steps: constructing an impeller parameterized geometric model and extracting design variables; an optimized Latin hypercube sampling strategy is adopted to construct a fluid-solid coupling sample library, and an intrinsic orthogonal decomposition technology is combined with Gaussian process regression to establish a high-precision full-field deformation prediction model; establishing a multi-objective optimization function based on a full-field deformation prediction model, and performing optimization by using a non-dominated sorting genetic algorithm to obtain a target thermal state geometry considering both hydraulic performance and structural safety; and finally, carrying out reverse iteration solution on the target thermal-state geometry by utilizing a geometry-load mixed correction algorithm to obtain the cold-state manufacturing geometry in a static state. According to the method, the calculation cost is remarkably reduced through the agent model, the adverse effect of fluid-structure interaction deformation on the impeller performance is eliminated from the manufacturing end through the reverse pre-compensation strategy, and the actual operation efficiency and lift stability of the high-speed micro centrifugal pump are effectively improved.

The application discloses a pipeline vibration characteristic prediction method and system, and belongs to the technical field of natural gas station pipeline vibration prediction, wherein the method comprises pipeline three-dimensional model construction, pipeline vibration data obtained by fluid and structure dynamics method calculation, reduced order basis vector obtained based on intrinsic orthogonal decomposition method, reduced order coefficient obtained by using a neural network model, and vibration prediction realized by coupling the reduced order basis vector and the reduced order coefficient. The application can quickly predict the pipeline vibration change condition by combining the reduced order coefficient and the neural network model. The method can perceive the change of the overall dynamic characteristics of the pipeline network caused by the slight change of the key parameters, has very high detection sensitivity for early and local abnormalities which are difficult to be found by the traditional method, and realizes the accurate prediction of the pipeline vibration.

The invention discloses a submersible motor temperature field joint calculation method based on a finite volume method and an analytical method, and the method builds an efficient coupling temperature field calculation method through building a temperature dependence model of key parameters and integrating the temperature dependence model in fluid-solid coupling calculation. According to the method, through parameterization updating instead of complex multi-physical field direct coupling, the huge calculation burden of a traditional two-way coupling model is avoided while the temperature-loss interaction is accurately reflected, and the balance between the calculation precision and the calculation efficiency is achieved. According to the method, the calculation dimension can be reduced to be extremely low based on the temperature field reduced-order model constructed by the intrinsic orthogonal decomposition. Therefore, when temperature prediction is carried out aiming at different working conditions, a time-consuming full-order model does not need to be repeatedly operated, and only quick calculation needs to be carried out in the low-dimensional subspace, so that quick response to variable working conditions is realized, and the design and analysis efficiency is greatly improved.

The invention relates to the technical field of data processing, in particular to a code disc concentricity adjusting method for a DD motor, and the method comprises the steps: obtaining a rotor absolute angle, a pulse error and a rotor angular velocity of each sampling point, and calculating the credibility weight of each sampling point according to the rotor angular velocity; a weighted vector integral model is introduced, the mode length of an eccentric main vector and an eccentric phase angle are obtained through the combined operation of orthogonal decomposition and full-period accumulation, and the physical adjustment distance of the code disc is calculated based on the mode length of the eccentric main vector in combination with a geometric mapping coefficient and a friction compensation amount. According to the eccentric phase angle and the physical adjustment distance, visual guide information is generated, an operator is instructed to conduct concentricity adjustment on the coded disc, noise and deformation interference are effectively filtered out, pulse errors are converted into specific physical displacement instructions, and standardization and high efficiency of debugging operation are achieved.

The invention discloses an ocean container transport ship design system and method based on digital twinning, and belongs to the technical field of ship engineering and digital twinning. The method aims at solving the technical problem that traditional finite element analysis cannot provide real-time feedback due to massive working condition combinations in the design process of the ocean container transport ship. The system and the method comprise the following steps: constructing a standardized topological domain for unifying different ship type variants; a high-fidelity full-component stress tensor sample set under the working condition of a small number of anchor points is calculated in an off-line mode; performing orthogonal decomposition on the sample set, and extracting a group of orthogonal stress bases carrying dimensional information; in the design interaction stage, the system predicts a dimensionless modal coefficient through a pre-training model according to design parameters such as a hull geometric variant, a stowage scheme and a sea condition input by a designer, performs linear weighted stacking on the dimensionless modal coefficient and a stress base, reconstructs a full-field stress tensor of a hull structure in real time, and generates a visual evaluation result.

The application relates to the technical field of data security, in particular to an end user data leakage prevention method based on dynamic trust evaluation; the behavior change vector is subjected to orthogonal decomposition, is split into a drift component for representing a long-term trend and a mutation component for representing instantaneous fluctuation, and different processing strategies are designed for the two components; the mutation component is used for real-time detection of abnormal mutation events, and the drift component is used for baseline updating and trust value updating when a normal behavior drifts, so that the purpose of distinguishing between a normal behavior drift and malicious abnormal mutation is achieved; on the premise of maintaining abnormal detection sensitivity, the false positive rate is significantly reduced; each dimension of the behavior feature vector is divided according to time sensitivity, and different smoothing coefficients are allocated to high-sensitivity features and low-sensitivity features, so that the problem of feature dimension response mismatch caused by uniform smoothing coefficients is avoided, and the false alarm risk caused by improper feature extraction is further reduced.

The application discloses a kind of carbon fiber composite material performance optimization method and system based on multi-physical field reduced order model, and it is related to carbon fiber composite material design field.Method includes: generating sample point in design variable space and carrying out high-fidelity simulation, extract full-field response data to construct snapshot matrix;Eigenvalue orthogonal decomposition is carried out to snapshot matrix to extract leading mode and constitute reduced order base, construct deep neural network to establish the nonlinear mapping of design parameter to reduced order coordinate, train network using the total loss function of data fitting loss and physical residual loss weighted summation;Reduced order base and neural network are encapsulated as fast predictor;Predictor is integrated into optimization algorithm, and the process of recommending candidate point, predicting performance, updating historical data is carried out, and the optimal design scheme is output.The application shortens single analysis time from hour level to second level by reduced order technique, and physical information constraint ensures model reliability, to realize the efficient and high-precision optimization design of composite structure.

The application discloses a tunnel smoke spread real-time prediction method based on physical priori and timing network, comprising the following steps: acquiring multi-source timing data and preprocessing, and constructing multi-working condition data set; constructing a mixed physical priori containing smoke spread length priori and smoke layer height field shape priori; adopting intrinsic orthogonal decomposition to reduce the dimension of the smoke layer height field, and extracting spatial main mode and mode coefficient; constructing a multi-task prediction model based on a causal timing convolution network, and synchronously predicting the smoke backflow length and the mode coefficient; using a mixed loss function fusing data fitting loss, physical priori loss and smoothing constraint, and combining a progressive loading mechanism to train the model; using the trained model to output the prediction result in real time, and reconstructing the smoke layer spatial distribution. The application fuses physical priori and causal timing network, realizes real-time and high-precision prediction of the tunnel fire smoke backflow length and the smoke layer spatial distribution, and has the advantages of strong physical rationality, high generalization ability and stable prediction.