30 results about "Automatic differentiation" patented technology

A digital twin modeling method and device for motors based on multi-field coupled PINN (Partial Differential Equations) belongs to the field of intelligent manufacturing and advanced control technology for motors. Its key features include: acquiring stator current, voltage, rotor position, and vibration signals of the motor to construct a full-domain multimodal sensing layer; constructing a multi-field coupled PINN model based on automatic differentiation; and embedding the multi-field partial differential equations (PDEs) inside the motor as prior knowledge into the network structure of the PINN model. By utilizing PINN technology, the physical laws of PDEs are integrated into the neural network, ensuring both millisecond-level inference speed and physical consistency in areas lacking sensor data, achieving both high fidelity and real-time performance. It enables precise virtual sensing of unmeasurable rotor permanent magnet temperature and local magnetic flux density saturation within the motor, achieving full-domain observability. Through edge-cloud collaboration, it can automatically evolve as the physical entities of the motor age (e.g., resistance changes, magnet demagnetization), maintaining model accuracy throughout its entire lifecycle.

The application discloses a steel bridge deck roller compacted asphalt concrete construction simulation method based on BIM, and relates to the technical field of simulation. The method first analyzes the geometric boundary of the BIM model and constructs an adaptive space-time sampling point set; then constructs a deep neural network introducing Fourier feature mapping, maps low-dimensional space-time coordinates into high-dimensional features to capture high-frequency physical changes; then establishes a composite loss function containing heat conduction, thermal coupling and boundary condition residual error, and uses automatic differentiation technology and an adaptive weight algorithm based on gradient statistics for unsupervised training; finally, the trained lightweight model is integrated into the BIM rendering engine. The application solves the problems of long time consumption of traditional finite element simulation calculation, inability of real-time interaction, and poor generalization ability of pure data-driven AI due to lack of internal measured data, and realizes real-time, high-precision dynamic simulation of the temperature field and stress field of the steel bridge deck pavement.

This invention discloses a method and system for solving the performance of a gas turbine engine by integrating differentiable operators and static computation graphs. The method includes: constructing differentiable component operators: reconstructing the input-output mapping relationship of the core components of the gas turbine engine into a globally differentiable surrogate model, which possesses analytical gradient solving capability; constructing the overall engine static computation graph: utilizing a programming framework supporting automatic differentiation, establishing a general graph construction mechanism decoupled from the engine topology, which automatically instantiates the corresponding static computation graph based on the input engine topology description; and reverse-mode automatic differentiation solving: calculating the residual vector through forward propagation, and using reverse-mode automatic differentiation technology to propagate the gradient flow backward along the static computation graph, analytically obtaining the Jacobian matrix required for Newton iteration, and completing the solution of the engine's overall performance residual equation. This invention achieves efficient, accurate, and stable simulation of gas turbine engine performance.

The application discloses a kind of based on deep learning's full-waveband microwave radiation transmission simulation method and system, method includes: based on global reanalysis data and surface type product, obtain atmospheric profile and surface parameter, adopt near neighbor search and de-redundancy, construct representative "surface parameter-atmospheric profile" database;Using high-precision microwave radiation transmission mode, simulate and calculate the atmospheric top layer brightness temperature covering full-waveband microwave, construct training sample pair;Calculate the correlation coefficient between all output channels, and select the representative channel subset with reduced number from threshold value;Build a fully connected neural network, with the brightness temperature of the representative channel subset as the intermediate feature layer, and use a phased composite loss function strategy for training;Surface parameters and atmospheric profile are input into the trained model to obtain full-waveband brightness temperature prediction results, and / or using automatic differentiation mechanism, the trained model is used as a differentiable operator to calculate the Jacobian matrix of the output brightness temperature relative to the input parameters.

The application discloses a kind of energy storage system prediction methods based on physical information neural network, first, the operating target of energy storage system is combined with security constraint, and the augmented state model containing electricity, thermology and attenuation mechanism is established, and prediction target is constructed;Second, PINN prediction network is constructed, and physical residual is calculated using automatic differentiation, to form a joint loss function consisting of data consistency loss, physical constraint loss and boundary penalty term;Then, BMS / EMS operating data is collected within the preset time interval, and the optimal model is solved based on the joint loss to update the network parameters;Finally, the trained model is deployed in the online rolling prediction link, and the predicted future SOC, temperature and available charge and discharge power are output, which are used as the scheduling input of energy management system.The method of the present application reduces the dependence on complex and high-precision mechanism modeling while ensuring that the prediction results meet the physical laws, achieving an integrated design from data and physical constraints to energy storage prediction and scheduling support.

A method for full-field inversion of corrected Mohr-Coulomb constitutive parameters based on automatic differentiation of computational graphs is presented, relating to the field of numerical simulation in geotechnical engineering. This method constructs a smoothed, differentiable corrected Mohr-Coulomb computational graph. After inputting initial data of the soil and rock mass to be analyzed, a forward finite element simulation is performed to obtain the simulated displacement field. The loss function is calculated, and its change is checked to determine convergence. If convergence is not achieved, the gradient is automatically calculated via backpropagation, and the parameters are updated using the L-BFGS optimization algorithm, with TV regularization constraints used to maintain the smoothness of the parameter field. The forward finite element simulation is repeated, and this process is iterated until convergence is achieved. The inverted parameter field is then output. By utilizing automatic differentiation technology, the accurate gradient of the full-field parameters can be obtained with a single backpropagation, introducing the automatic differentiation technology of deep learning frameworks into the field of geotechnical engineering, realizing the transformation from manual gradient derivation to automatic machine differentiation.

The application discloses a dimension reduction modeling method and device for ammonia synthesis reaction-heat transfer coupling process, comprising: constructing a forward propagation fully connected neural network with time coordinate t and axial space coordinate z as input; in the forward propagation process of the fully connected neural network, the mechanism expression of the ammonia synthesis reaction rate is explicitly embedded as a structural prior; based on the output of the fully connected neural network, the time and space derivatives of the ammonia mass fraction and the gas temperature are calculated through automatic differentiation, and a trend type penalty term loss, a mass conservation PDE residual loss, an energy conservation PDE residual loss and a boundary residual loss are constructed; the trend type penalty term loss is used to constrain the monotone increasing trend of the ammonia mass fraction and the temperature along the axial direction of the bed, and the penalty is generated only when the spatial derivative is negative; the total loss function is composed of the loss terms by weighting according to the preset weight, and the network parameters are optimized through back propagation, so that the dimension reduction solution and fast prediction of the reaction-heat transfer coupling partial differential equation are realized.

This disclosure relates to training and fine-tuning of neural networks on a neural processing unit. A core on a neural processing unit can perform matrix multiplication (MatMul) on tensors of different dimensions. A neural network can be trained through forward and backward operations, both of which can be offloaded to the core. For the forward operation, the core can perform a layer by performing MatMul on an input tensor and a weight tensor and generate an output tensor. A loss can be computed. For the backward operation, the core can compute a weight gradient of the loss by performing MatMul on a gradient of the output tensor and the input tensor, and compute an input gradient of the loss by performing MatMul on the gradient of the output tensor and the weight tensor. The gradient of the output tensor can be computed by an automatic differentiation module. The weight tensor can be updated based on the input gradient and the weight gradient.

The application discloses a cell migration behavior prediction method based on physical information width learning, which utilizes cell scratch experiment data to construct a base function matrix, combines a Fisher-KPP reaction diffusion equation to construct a nonlinear least square problem containing physical mechanism constraints, and adopts an enhanced nonlinear least square disturbance algorithm to iteratively solve output weights, so as to realize prediction of cell density space-time evolution. The application converts traditional deep network iterative training into a nonlinear least square problem, greatly reduces a search space through physical information initialization and linearization approximation, and realizes order-of-magnitude improvement of training speed. The application constructs a width learning architecture containing feature nodes and enhanced nodes, effectively avoids the gradient vanishing problem in deep learning, combines analytical derivative calculation, eliminates cumulative error of automatic differentiation on high-order derivatives, and can more accurately capture space-time evolution characteristics in cell migration.

This invention relates to the fields of computational materials science and materials informatics, and discloses a method for predicting and screening material properties based on a deep learning force field model. The method includes: inputting material structure data into a first-principles calculation module to generate an initial dataset; removing outlier samples through multiple physical plausibility checks; inputting the screened, consistent dataset into a force field model based on an isovariant graph neural network for training; updating model parameters by constructing a joint loss function using an automatic differentiation mechanism and physical conservation constraints; inputting candidate materials into the trained model to obtain energy and forces; outputting mechanically stable configurations after structural relaxation; calculating the multi-dimensional physical properties of these configurations; and weighting and ranking them using a multi-index comprehensive scoring function. This invention solves the problems of easy force divergence in traditional force fields and the high cost of first-principles calculations, achieving low-cost, high-throughput screening of unknown materials while ensuring physical consistency and accuracy.

This invention belongs to the field of embodied intelligent robot technology and discloses a method for optimizing the operational parameters of embodied intelligent robots based on digital twins. Addressing the problem of discrepancies and disconnects between simulation and reality in terms of physical parameters, this invention constructs a three-dimensional twin representation and uses photometric error gradients to generate virtual force injection into a differentiable engine to achieve state synchronization. The simulation is reconstructed into a differentiable computational graph, and physical parameters are identified through automatic differentiation via backpropagation. Then, using the aligned parameters, a nested dual-loop control architecture is employed. The outer loop outputs the target pose, while the inner loop estimates the contact force based on inverse dynamics residual inversion, executing sensorless compliant impedance control. This invention effectively improves parameter identification accuracy and achieves high-frequency closed-loop execution, making it widely applicable to tasks such as flexible assembly.

This application provides a method and apparatus for modeling and predicting the wake field of a wind turbine, addressing the difficulty in balancing accuracy, efficiency, and physical consistency in existing wake prediction methods. The method includes: constructing a deep neural network model, outputting predicted values ​​of corresponding flow field physical quantities; calculating the residuals of the physical equations based on the hydrodynamic control equations followed by the wind turbine wake, using the output of the deep neural network and automatic differentiation techniques, and constructing a composite loss function including data fitting loss terms, physical constraint loss terms, boundary condition loss terms, and initial condition loss terms; training the deep neural network model using a training dataset by minimizing the composite loss function to obtain a wind turbine wake field prediction model; and using the wind turbine wake field prediction model to simulate and predict the wake field of a target wind turbine.

The present application relates to the technical field of nanophotonics and computational optical imaging, and particularly relates to a computational spectral metasurface reverse design method based on an iterative optimization algorithm, comprising: S1, physical structure parameterization, defining the geometric parameters, arrangement period and material refractive index of the nanostructure in the unit, initializing the physical structure parameters and the corresponding electromagnetic response operator; S2, constructing a differentiable simulation model, using a rigorous coupled wave analysis algorithm based on an automatic differentiation framework to calculate the complex amplitude transmission spectrum data under the current parameters, and keeping the electromagnetic response gradient flow continuous in the calculation graph. In the present application, through end-to-end soft and hard collaborative optimization of the metasurface physical structure parameters and the neural network reconstruction algorithm weight, the deep coupling and collaborative evolution of the hardware coding characteristics and the software decoding logic are realized, the performance mismatch problem caused by the traditional step-by-step design is fundamentally solved, and the restoration accuracy and recognition sensitivity of the system to complex spectral characteristics are significantly improved.

The application discloses a face expression real-time closed-loop compensation method for a bionic human head, constructs a forward model through self-supervised learning and constrains a reverse model, only uses the reverse model to output a reference angle vector during operation, simultaneously extracts a Jacobian matrix in real time by using automatic differentiation, directly analyzes instantaneous expression residuals of visual feedback into actuator angle compensation amounts, generates driving instructions after superposition, realizes real-time closed-loop compensation of the actuator based on the Jacobian matrix, and triggers collaborative reconstruction of a mapping model when the residuals continuously do not decrease, and adapts to long-term physical drift. The application reduces calculation overhead by single model operation, and compensation calculation is only once matrix multiplication, so that response is rapid, and real-time accuracy and long-term stable control of face expression of the bionic human head can be ensured.

This invention provides a method and system for predicting and reversing the operating conditions of a waste incinerator, relating to the field of intelligent control technology. The method utilizes L... 25 (5) 5 Orthogonal experiments can significantly reduce the number of simulation case groups. Combined with a two-dimensional planar CFD model, it can quickly generate high-quality datasets, significantly reducing data acquisition costs and cycles. It adopts sliding window MLP prediction technology, which has high prediction accuracy and lightweight deployment. The latency of forward inference and backward optimization is extremely low. It can automatically differentiate gradient-based backward optimization based on the forward model, upgrading from "post-event adjustment" to "proactive prediction + optimal control". Multi-parameter collaborative optimization has low error and fast response, fully meeting the needs of industrial real-time control.

The application discloses a hydrological model deep learning rate setting method for large-scale multi-sub-basins, establishes a data warehouse supporting high-concurrency reading, constructs a structured index mapping table to realize direct addressing of data, reconstructs continuous hydrological physical processes into a vectorized whole-process differentiable computation graph, adapts parallel computing characteristics of a graphics processing unit by using instant compilation and operator fusion technology, constructs a double-parallel high-dimensional parameter tensor containing a basin dimension and a parameter group dimension, realizes concurrent forward evolution of a mass of basin-parameter groups by using multi-path random initialization and tensor broadcasting technology, calculates a gradient by using a reverse mode automatic differentiation mechanism, updates parameters by using a deterministic optimization algorithm, constructs a performance evaluation matrix based on a training set, filters out an optimal parameter set, and exports the optimal parameter set after verification by using a verification set. The application breaks through the efficiency bottleneck of traditional serial computation, considers convergence speed and global optimization accuracy, and realizes full-process automation of large-scale basin parameter rate setting.

A path planning method based on analytical gradients of a neural network includes: acquiring sample datasets of a rocket under multiple typical flight conditions; training the neural network with flight control variables as input and aerodynamic coefficients as output until convergence is met, resulting in a surrogate model for fitting the rocket's aerodynamic characteristics; collecting the rocket's current state vector and control vector, inputting the control vector into the surrogate model to obtain the current aerodynamic coefficients, and constructing the rocket's flight state equation; based on the flight state equation, using an automatic differentiation method, analytically calculating the gradient with respect to the control vector to form a gradient sensitivity matrix; using the gradient sensitivity matrix to determine the descent direction; and iteratively updating the control vector using a numerical optimization algorithm until the termination condition is met, outputting the optimal flight path. The optimal flight path obtained by this method more closely matches the rocket's actual aerodynamic environment, exhibiting higher terminal accuracy.

This invention belongs to the field of semiconductor manufacturing technology and discloses an intelligent cause-tracing method, system, device, and storage medium for WAT parameter anomalies. The method includes: collecting WAT test-related metadata from multiple systems; performing unified identification and time-series alignment; evaluating data reliability based on test link status; detecting anomalies and generating abnormal events according to reliability constraints; aggregating multi-source evidence to construct an event-level heterogeneous correlation graph; and using a pre-trained graph neural network model for graph reasoning to output root cause candidates, confidence levels, and traceable reasoning paths. Through these methods, automatic differentiation between test contamination and process anomalies is achieved, improving the accuracy, interpretability, and processing efficiency of cause tracing.

This invention relates to the field of optical technology, providing an automated design method and apparatus for anti-diffraction pupil filters based on process constraints. The aforementioned automated design method for anti-diffraction pupil filters based on process constraints includes: constructing a mixed spectral parameter space based on the microstructure of the pupil edge, mapping the mixed spectral parameters to a spatial domain edge radial offset sequence; constructing a closed vector graphic based on smooth edge contours, mapping the distance field to a two-dimensional pupil transmittance distribution function; constructing a full-link optical simulation model based on the two-dimensional pupil transmittance distribution function and Fourier scalar diffraction theory; calculating the gradient of the composite loss function with respect to the mixed spectral parameters based on an automatic differential framework, and iteratively updating the parameters using a numerical optimization algorithm. This automated design method for anti-diffraction pupil filters based on process constraints successfully solves the common problem in computational optical design of the disconnect between theoretical models and physical fabrication.

The application discloses a kind of robot process online regulation and control methods suitable for material uncertainty, and it is related to intelligent control technical field.The application collects the process parameter sequence of the component to be processed, initial topography data and post-processing topography data;Through the differentiable topography evolution agent model, the derivable predicted topography data is output;A two-parameter inversion optimization model is constructed, the actual and predicted topography difference is quantified by mean square error, and the minimum is taken as the goal, the material parameter prior L2 norm constraint and process parameter boundary penalty constraint are embedded;Based on the reverse automatic differentiation algorithm, the gradient vector is solved, and the optimal material parameter is obtained by the random gradient descent algorithm with momentum iterative optimization, and then the future time step process parameter sequence is optimized, and the optimal solution of material parameter and process parameter is output after constraint verification and convergence judgment.The application realizes the accurate online regulation and control of robot process under material uncertainty, and has wide application value.

The present application relates to a kind of phase correction full waveform inversion methods based on continuous wavelet transform, specifically is using continuous wavelet transform to transform time domain seismic signal to wavelet domain, then in wavelet domain to observed data is carried out phase correction, and corresponding wavelet domain objective function is constructed, relies on automatic differentiation technique to solve gradient, realizes a kind of method of velocity modeling.First, to seismic wave data is carried out continuous wavelet transform, extracts the wavelet domain phase information of observed data and simulated data, by introducing phase correction factor ε, constructs exponential phase difference, and establishes the objective function of phase correction inversion in wavelet domain;Second, the loss of objective function is transmitted back under the automatic differentiation framework, to avoid with source derivation, to obtain update gradient;Finally, using Adam optimizer updates velocity model.The present application can be reduced to within half cycle by phase correction so that observed data and simulated data phase difference, to effectively avoid period jump phenomenon, finally obtain high-precision inversion result.

The application discloses a lithium ion battery health state estimation method based on physical information DeepONet. The method extracts multi-dimensional health characteristics from charge and discharge cycle data and constructs paired samples; a decoupled DeepONet architecture is constructed, a Branch-Trunk double branch structure is adopted to explicitly separate the working condition characteristics and the cycle index, and a sine activation function is adopted to ensure infinite differentiability; a physical discovery module is constructed, the partial derivatives of the prediction value with respect to time and characteristics are calculated through automatic differentiation, and the residual constraint and soft monotonicity constraint of ordinary differential equations are used to force the prediction trajectory to conform to the degradation dynamics; a two-stage training strategy is adopted, the optimal candidate model is selected through integrated initialization pre-training in the first stage, and physical regularization fine-tuning is performed on the dynamic weighted composite loss function in the second stage. The application alleviates the problems of feature confusion, gradient conflict and initialization sensitivity, and realizes high-precision and high-physical-consistency health state estimation.

The application discloses a kind of based on physical information neural network numerical control energy efficiency adaptive planning method, it is related to intelligent manufacturing and numerical control processing technical field, in view of the defect that existing conservative feeding strategy leads to machine tool standby energy consumption is high, and pure data driven model generalization is poor, the application first constructs the instantaneous mechanical model containing to be identified physical parameter;Second, using small sample cutting data hybrid training physical information neural network, network weight and physical parameter are updated jointly, realize the pre-adaptive identification of material attribute and high-fidelity prediction of cutting force;Trajectory discretization and geometric boolean operation are carried out to numerical control program, and instantaneous cutting feature is extracted;Finally, in combination with network forward inference and machine tool physical constraint, rely on automatic differentiation mechanism and hot start gradient descent algorithm, reversely solve global optimal variable feed speed sequence.The application realizes the significant energy saving and efficiency improvement of complex curved surface machining under the premise of not increasing additional sensor hardware.

The application discloses a face expression real-time closed-loop compensation method for a bionic human head, constructs a forward model through self-supervised learning and constrains a reverse model, only uses the reverse model to output a reference angle vector during operation, simultaneously extracts a Jacobian matrix in real time by using automatic differentiation, directly analyzes instantaneous expression residuals of visual feedback into actuator angle compensation amounts, generates driving instructions after superposition, realizes real-time closed-loop compensation of the actuator based on the Jacobian matrix, and triggers collaborative reconstruction of a mapping model when the residuals continuously do not decrease, and adapts to long-term physical drift. The application reduces calculation overhead by single model operation, and compensation calculation is only once matrix multiplication, so that response is rapid, and real-time accuracy and long-term stable control of face expression of the bionic human head can be ensured.

This invention discloses a method for performance evaluation and optimization of automatic differential operators. Under specified hardware and a deep learning framework, it evaluates multiple candidate automatic differential operators and constructs performance profiles. Combining the target scenario, it sets data batch anchor points, compares and analyzes to determine the optimal operator at each anchor point, and uses this operator to train a physical information neural network, gradually increasing the network complexity. It then obtains the maximum runnable network structure under hardware constraints and the upper bound of the data batch without memory overflow, generating corresponding recommended configurations. This invention achieves reproducible operator performance profiles under unified evaluation, quantitatively assesses computational efficiency and resource consumption under different data and network scales, and provides a quantifiable optimization basis for the training configuration of physical information neural networks.

The application discloses a kind of acoustic vector multiplexing structure parameter generation method and system based on physical coupling constraint, first, obtain the target tensor set containing the first and second vector components of sound velocity field, after formatting, input physical driving neural network, output the binary structure parameter matrix of metasurface unit;Differentiable acoustic physics layer is embedded into network, based on wave equation and linear Euler equation, the physical mapping of parameter and vector field is established, to generate the prediction field data containing coupling characteristics;Construct a multi-channel joint loss function, while constraining the component difference of prediction field and target field and the physical coupling relationship;Gradient is calculated using automatic differentiation and backpropagation, iteratively update network weights to convergence, output structure parameters that meet physical coupling constraints;The application solves the difficulty of vector field decoupling and multiplexing crosstalk caused by physical strong coupling in the prior art, and realizes efficient generation of high-fidelity vector holographic multiplexing acoustic structure parameters.

This invention discloses a joint estimation method for power system state and parameters based on Newton-Raphson guided machine learning, relating to the fields of smart grid optimization and power system analysis. The method includes: addressing the problems of low iteration efficiency and machine learning physical mapping distortion in traditional state estimation and parameter identification, embedding the fixed-step Newton-Raphson method as a differentiable module into the machine learning training framework to construct an end-to-end model for parameter and state co-optimization; establishing a PV / PQ node hybrid input-output mapping mechanism that conforms to actual measurement scenarios, using the admittance matrix as an optimizable parameter, and achieving gradient backpropagation under physical constraints through automatic differentiation; designing a node-level state estimation loss function, combined with physical regularization methods for conductance / susceptance, to ensure the feasibility and stability of parameter estimation; and balancing computational efficiency and accuracy by dynamically adjusting the number of Newton iteration steps, ultimately achieving high-precision state prediction and grid parameter reconstruction. Verified on the IEEE-118 node system, this method converges within three Newton iterations, with a state estimation error below 1.2e⁻³, an admittance reconstruction error reduced by 41% compared to the baseline, and strong generalization ability under multiple operating conditions. This invention can be applied to the construction of digital twins for power grids, online security assessment, and fault diagnosis, significantly improving the efficiency and accuracy of power system analysis.

The application discloses a turbine damping blade dynamics design optimization method based on a convolutional neural network, comprising the following steps: obtaining a blade vibration reduction model, selecting model parameters and aerodynamic parameters to construct a characteristic space, calculating frequency response characteristics by using an arc length continuation method based on a multi-harmonic balance method, constructing a series of convolutional neural network models, normalizing characteristic parameters, frequency response characteristics and dynamic stress fields, and realizing vibration reduction performance optimization by combining automatic differentiation algorithms with characteristic prediction. The turbine damping blade dynamics design optimization method based on the convolutional neural network can realize accurate and rapid prediction of nonlinear response of the turbine damping blade, save calculation cost, improve solving efficiency, and obtain a friction damping structure scheme with the best vibration reduction effect in a given parameter working condition range, thereby providing theoretical support for dynamics design and optimization of turbine damping blades in engineering.