124 results about "Sequence prediction" patented technology

The present application relates to a kind of based on multi-source industrial data's car body welding quality intelligent perception, prediction and decision-making method and system, belong to the technical field of intelligent manufacturing and industrial data analysis.The present application obtains multi-source high-frequency synchronous original signal by sensor array and robot bus;Perform space-time alignment and multidimensional feature mining;Quality perception model is constructed based on multi-path parallel network and cross-modal attention;Quality evolution trend and electrode cap life are evaluated using time series prediction model;Integrate expert rules and reinforcement learning to make process compensation decision.The present application realizes the accurate perception of welding quality, forward warning and closed-loop real-time compensation, significantly improves the perception accuracy, production robustness and prolongs the service life of electrode.

The application discloses an urban waterlogging intelligent prediction method and system based on a physically constrained enhanced video generative adversarial network. The method comprises the following steps: constructing a meteorological geographic data set containing static terrain data and dynamic rainfall data; generating waterlogging simulation data based on a physical numerical simulation model, and matching the waterlogging simulation data with the meteorological geographic data set into a comprehensive data set; constructing a physically constrained enhanced video generative adversarial network model, wherein the model comprises a generator and a discriminator; constructing a multi-angle loss function system that fuses an adversarial loss, a time sequence consistency loss, a data supervision loss and a water quantity balance constraint physical loss; and alternately optimizing the generator and the discriminator until the model converges, and outputting a spatiotemporally continuous waterlogging inundation depth prediction result. The application explicitly embeds a physical hydrological law into a deep learning network, and eliminates the physical logic contradiction of a pure data-driven model in continuous spatiotemporal sequence prediction through water quantity balance constraint.

This invention discloses a method for predicting standard cell transistor placement sequences based on a Transformer model, an electronic device, and a storage medium. Step 1: Construct training samples based on standard cell netlist information and corresponding existing placement results; train the placement model using these training samples to obtain a trained placement model. Step 2: Obtain the standard cell netlist information to be placed and the current placement MOS sequence, and construct the feature encoding of the current state. Step 3: Input the feature encoding of the current state and the decision space into the placement model to obtain the next standard cell transistor to be placed, and place it. Step 4: Repeat steps 2-3 until all standard cell transistors have been placed. This invention improves the efficiency, accuracy, and versatility of placement, while reducing design time and cost.

The application provides a new energy electric drive system intelligent torque distribution method based on AI prediction control and related devices; the method acquires vehicle state and sensing data in real time, outputs future speed sequence of the vehicle in the prediction time domain through an environment sensing traffic sequence prediction network; outputs each wheel adhesion coefficient estimation value through a road adhesion coefficient online identification network; constructs a feedforward-feedback dual-channel model prediction controller, solves a feedforward torque distribution sequence based on the future speed sequence in a feedforward layer, and generates a feedback torque correction amount in a feedback layer based on a state deviation; dynamically generates a stability and economy weight vector according to an adhesion margin through a multi-objective dynamic weight distribution network; corrects a total optimization target with the dynamic weight and solves each motor torque instruction; and utilizes an online experience playback buffer to perform incremental online learning on the network during steady-state cruising. The application realizes forward-looking prediction, online identification, dynamic trade-off and continuous evolution, and improves the distribution performance.

This application discloses a digital twin control method for facility poultry farming environments based on deep spatiotemporal data fusion, relating to the field of poultry farming technology. It proposes a spatiotemporal sequence prediction model for environmental parameters using a bidirectional long short-term memory network (VMD-Attention-BiLSTM) that integrates variational mode decomposition and attention mechanisms, effectively overcoming the system's large time lag. A digital twin decision engine based on nonlinear model predictive control is constructed, and the optimal control sequence for energy consumption and environmental quality is solved under multiple constraints by establishing a joint state equation for thermodynamics and gas diffusion. Specifically, for complex winter conditions, a ventilation-heating collaborative game strategy based on real-time heat loss compensation is designed. Adaptive sliding mode control with radial basis function neural network compensation is introduced into the underlying execution unit, significantly enhancing the system's anti-disturbance capability. This achieves precise environmental control under complex conditions, improving poultry welfare and energy efficiency.

An air interface apparatus (210, 314) configured to utilize artificial intelligence (212) to perform air interface algorithms (316) is provided. The air interface apparatus is further configured to (i) obtain channel state information, CSI, including information on channel properties, at a global time (t) corresponding with internal sampling index (ℓ), (ii) obtain environment state information, ESI, including information on environmental properties, at the sampling time (ℓ), (iii) tokenize the CSI and ESI together into a state information token (x(t)), (iv) predict a future sequence of state information tokens (formula(I)) for a future internal sampling index (ℓ + Ρ) based on the sequence of past state information tokens until the global time t ({x(t)}), (v) generate a set of air interface representation inputs (formula (II)) for the future internal sampling index (ℓ + Ρ) by performing representation learning on the channel and environmental properties and the predicted future sequence of state information tokens and then to perform the air interface algorithm(s) utilizing artificial intelligence air interface algorithms (316) based on the air interface representation inputs (formula (II)).

This application relates to a pipeline inspection method, apparatus, computer equipment, storage medium, and computer program product. The method includes: acquiring a multi-channel raw signal sequence collected within a target pipeline; performing median filtering on the raw signal sequence based on a preset sliding window to obtain an intermediate observation sequence; performing dynamic prediction filtering on the intermediate observation sequence to obtain a reference filtered signal; wherein the dynamic prediction filtering is used to predict candidate sequences based on the intermediate observation sequence; and fusing the intermediate observation sequence and the candidate sequences; identifying abnormal sequences in the raw signal sequence based on the reference filtered signal; correcting the abnormal sequences to obtain an initial corrected sequence; and performing secondary filtering on the initial corrected sequence to obtain a target filtered signal. This method can effectively improve the signal-to-noise ratio and anomaly identification accuracy of pipeline inspection data.

The application discloses a kind of based on CATIA's assembly sequence generation method, system, electronic equipment and storage medium, is related to industrial parts assembly technical field, the assembly sequence generation method based on CATIA includes: S10, the geometry and topological parameter of industrial parts is extracted, and assembly parameterization constraint is defined based on the assembly relationship between industrial parts;S20, sample data containing the geometry and topological parameter, the assembly parameterization constraint and corresponding artificial assembly operation sequence are collected, and the sample data is preprocessed and labeled, to construct the assembly dataset for model training;S30, assembly sequence prediction model is trained based on assembly dataset and loss function;S40, the assembly sequence prediction model predicts corresponding assembly operation sequence according to the parameterization information of input to-be-assembled industrial parts, and the assembly of industrial parts is automatically completed through CATIA platform.The beneficial effects of the application: improve assembly efficiency and assembly accuracy.

A data-driven method for monitoring machine tool spindle cutting torque is characterized by: synchronously monitoring servo signals such as the output torque or three-phase current of the machine tool spindle motor, spindle motion status, and spindle cutting torque; conducting cutting experiments to obtain a spindle servo monitoring signal-cutting torque dataset; and then training a data-driven time series single-step prediction model using a sequence-sequence supervised learning approach. In practical applications, the corresponding spindle servo monitoring signals are input into the prediction model for single-step real-time prediction or sequence-sequence prediction, thereby achieving online monitoring of the spindle cutting torque. Compared to traditional spindle cutting torque monitoring methods based on benchtop / rotary cutting force gauges, this invention is lower in cost and does not intrude on the machine tool workspace, possessing significant potential for application in actual production processes.

This invention discloses a multi-feature fusion time series prediction method, system, device, and medium for power load peak prediction, belonging to the field of power system load prediction technology. It includes: acquiring and preprocessing raw power load time series data; extracting and constructing a load peak time series for feature construction to obtain a prediction feature set; constructing supervised learning samples to transform the time series prediction problem into a regression problem; and training the supervised learning samples using a machine learning model to obtain a load peak prediction model for predicting and evaluating future load peaks. The beneficial effects of this invention are as follows: Through a feature engineering method of multi-feature fusion, this invention can effectively capture the historical dependence, short-term volatility, and long-term periodicity in power load time series, thereby significantly improving the accuracy of load peak prediction. It has a predictive advantage, especially for power systems with large load fluctuations, in environments with large-scale integration of new energy sources.

The application discloses a method, device, computer device and storage medium for training a model as a basis for a future time series prediction task. The method comprises: dividing past time series into a plurality of sequence blocks at a channel level, and linearly embedding the plurality of sequence blocks; adding a position embedding to the linear embedding of the plurality of sequence blocks; inputting the embedded plurality of sequence blocks into an encoder; and outputting a predicted future time series based on the encoder. In this way, more accurate prediction of future time series can be obtained under the condition of a limited number of past time series and zero-sample training, thereby providing greater inspiration for the field of time series prediction basis modeling.

This invention discloses a high-speed communication link transmitter modeling method based on a Transformer decoder. This method parameterizes the input signal and establishes a deep learning model with an encoder-decoder architecture, including a non-sequential encoder and a Transformer sequential decoder, to predict the output signal of the transmitter given the input signal. The encoder processes the input signal parameters, generates a context vector, and then passes it to the decoder. The decoder uses the context vector and the transmitter output signal sequence to generate a class probability distribution for each point in the sequence. The model is trained using a random masking strategy and employs non-autoregressive decoding and filtering techniques for inference, achieving parallel output sequence prediction. The final output is obtained through a single signal filtering step. Compared with traditional simulation techniques, this method significantly improves simulation speed while maintaining a very small error.

The present application relates to the field of aero-engine simulation technology, more particularly to a semi-physical simulation method and system based on a prediction model.The method comprises: collecting engine test data; identifying a transfer matrix of a state space of the prediction model according to the engine test data, the transfer matrix being used to describe the relationship between an input sequence and an output sequence; calculating matrix coefficients of the prediction model based on the transfer matrix; calculating a predicted value of the output sequence corresponding to the input sequence based on the matrix coefficients and current engine state quantities; and adjusting and updating the operating state of an engine physical system in real time based on the predicted value of the output sequence, so that the engine meets the expected performance.The present application provides high-precision prediction and control for the engine control system by constructing a real-time updated prediction model in combination with actual engine test data.The robustness and adaptability of the control system under fault conditions are enhanced by a fault injection simulation method which simulates various fault types.

This invention relates to the field of text detection, specifically to a text detection method, system, and storage medium in low-light environments. The method includes: acquiring a low-light text image to be detected, and performing adaptive image quality preprocessing on the low-light text image to obtain an enhanced text image; inputting the enhanced text image into a pre-trained target detection network to obtain the location coordinates of the text region in the text image; cropping the text region image based on the location coordinates of the text region, and inputting it into a pre-trained pixel segmentation network to obtain the text information contour in the text region image; extracting text information based on the text information contour, and inputting it into a pre-trained text recognition network for feature extraction and sequence prediction to form a target text sequence.

This invention discloses a drilling intelligent decision-making system and method integrating physical constraints, transfer learning, and online updates. The system includes a data access and processing module; a physical perception sequence prediction module, which uses a selective state-space model to process preprocessed drilling parameters and introduces physical consistency constraints during model training; a transfer learning adaptation module, which uses a weight decomposition low-rank adaptation method to decompose the weight matrix of the selective state-space model into an amplitude vector and a direction matrix; an online drift detection and update module, which is used to implement an adaptation strategy during mixed testing and correct the parameters of the selective state-space model; and a multi-objective safety optimization module, which is used to construct a dynamic safety operation envelope and generate drilling parameter recommendation instructions within the dynamic safety operation envelope using a multi-objective optimization algorithm. This invention effectively solves the problems of poor model interpretability and insufficient safety of optimization schemes in existing technologies.

This invention discloses a visual-language navigation method with cross-modal alignment in dynamic occlusion environments. The method utilizes visual sensors, inertial measurement units, and LiDAR to collect multimodal data, and performs preprocessing and time synchronization. It perceives dynamic occlusions using a model composed of convolutional neural networks and long short-term memory networks, and predicts their future changes using a spatiotemporal sequence prediction algorithm. It extracts and fuses visual and semantic features using a dual-branch convolutional neural network and a Transformer based on a dynamic attention mechanism. Based on occlusion prediction, it extracts potential occlusion region features in advance from the temporal dimension, and repairs occluded images in the spatial dimension using generative adversarial networks and geometric constraints. It optimizes cross-modal feature alignment through an attention mechanism. Finally, it plans the path using a hybrid reinforcement learning algorithm based on deep Q-networks, spatial and fast exploratory random trees, and dynamically adjusts the path according to real-time occlusion. This invention improves the accuracy, adaptability, and reliability of visual-language navigation in dynamic occlusion environments.

This invention belongs to the technical field of cross-e-commerce platform information processing, specifically involving an AI-based method and system for joint prediction and recommendation of cross-platform e-commerce user behavior. The method collects user behavior data from different e-commerce platforms and converts the raw behavior into unified event records. Based on anonymized identity mapping information, cross-platform aggregation is performed, and behavioral fragment units are generated. Event density, category overlap, cross-platform jump time interval, search click consistency, time difference between adding to cart and placing an order, and platform switching frequency are extracted to generate migration trigger tags. The behavioral fragment units and migration trigger tags are then input into a dual-channel sequence prediction network, which outputs click probability, add-to-cart probability, and purchase probability, forming a recommended content list. This list is incrementally updated based on actual response behavior. This scheme can uniformly represent and jointly predict continuous behavior across platforms.

The present application relates to a kind of chaos sequence prediction network training and use method, interference system and medium, method includes: obtaining the derivative of several chaos sequence data of pre-set chaos system;Approximate calculation of the derivative of several chaos sequence data is based on pre-set numerical difference equation, determine the trend data of chaos sequence;Obtain initial KANs, the network parameter of initial KANs is trained based on the trend data and the chaos sequence data, obtain initial prediction network;Obtain the prediction fixed point of initial prediction network, the training state of initial prediction network is evaluated based on the prediction fixed point, when the training state is training completion, the initial prediction network is output as chaos sequence prediction network.On the basis of high prediction accuracy, training speed is improved, and then the real-time, accurate prediction of chaos sequence in intelligent interference system is realized.

This invention presents a simulation integration method for external deployment of deep learning models based on the TorchScript and FMU interface, belonging to the field of vehicle dynamics simulation and artificial intelligence technology. The method includes the following steps: Step 1, converting the PyTorch model to a TorchScript model; Step 2, loading the TorchScript model encapsulation DLL using a C++ interface; Step 3, encapsulating the DLL into an FMU using the FMI standard; Step 4, the simulation software calls the FMU in real time for vehicle simulation. Under the FMI 2.0 standard interface framework, this invention deploys a Transformer-based tire longitudinal force sequence prediction model in FMU form to the whole vehicle simulation environment, ensuring stable and reliable predictions through training and deployment. The method described in this invention achieves efficient and stable direct integration between the PyTorch model and vehicle simulation software, avoiding compatibility issues that may occur when using ONNX as an intermediate conversion step. The model deployment steps are simpler, and the deployment difficulty is significantly reduced. The combination of TorchScript and C++ interfaces ensures the efficiency and stability of real-time model calls, effectively improving the overall efficiency and reliability of the simulation.

The application relates to the technical field of software-defined wide area network and discloses a multi-link global routing method and system based on dynamic policy linkage, which is applied to an SD-WAN controller and used for solving the technical limitation problem that in a traditional SD-WAN routing method, decision basis is limited to a current state and policy rules are statically fixed. The method generates a predicted value and a confidence evaluation of future performance of a link through a time series prediction model; based on the confidence evaluation, historical, current and future state information is dynamically fused to form a link quality trend value; and the confidence evaluation is used to dynamically modulate the validity degree of a multi-dimensional service policy weight in evaluation; finally, path selection probability distribution is obtained through iterative optimization, and an optimal path is selected according to a service type for forwarding, so that routing decision is upgraded from static rule response to dynamic optimization based on risk perception, and the foresight, intelligence, stability and adaptive robustness of network routing decision in a complex dynamic environment are significantly improved.

The application discloses a kind of lncRNA subcellular localization prediction method based on lightGBM, comprising, first, the nucleotide of the front section multiple of known lncRNA sequence is intercepted as sequence sample one;Then by based on single strand multiclass position-specific three nucleotide bias and reverse complementary kmer, sequence sample one is respectively characterized coding, the combination of two kinds of feature coding is vector;Use lightGMB as learning algorithm;Using 5-fold cross validation optimization reverse complementary kmer and LightGBM's hyperparameter;Most intercept unknown lncRNA sequence front end multiple bases as sequence sample two, and its single strand multiclass position-specific three nucleotide bias and the combination of optimized reverse complementary kmer feature coding is input in trained lightGBM, will obtain its localization subcellular type, the patent of the application can be according to long-chain non-coding RNA sequence prediction in cytoplasm, nucleus, ribosome, cytosol, exosome five subcellular positions, the application realizes simply, and prediction precision is high.

The application relates to the technical field of marine monitoring and disaster early warning, and discloses a tsunami wave detection method and system based on a TCN-informer algorithm, which comprises the following steps: S1, acquiring original pressure time series data of the seabed, and performing standardization, resampling and segmentation processing on the original pressure time series data to obtain a model input sample sequence; and S2, constructing a TCN-Informer hybrid model in series connection of a TCN feature extractor and an Informer sequence predictor, training the TCN-Informer hybrid model by using historical pressure data without a tsunami event, and enabling the model to learn the change rule of the seabed pressure under normal conditions and the like; the method realizes accurate detection of a tsunami wave signal from original or simply pretreated seabed pressure data by fusing the advantages of the TCN and Informer models.

The application provides a dynamic risk early warning method and system, the method comprises the following steps: collecting multi-mode heterogeneous data and preprocessing the multi-mode heterogeneous data; determining a risk event vector corresponding to the same event based on the preprocessed multi-mode heterogeneous data; embedding and constraining the first type of language data in the risk event vector using a knowledge graph to generate target embedding features aligned with the risk features of the second type of language data; determining the attention weights of the multi-dimensional risk features in the risk event vector and the embedding features; determining the comprehensive risk score of the risk event by combining the multi-dimensional risk features and the corresponding attention weights; inputting the comprehensive risk score into a pre-constructed sequence prediction model to obtain the comprehensive risk score trend of the corresponding data in the future T time windows; determining the risk escalation probability by combining the comprehensive risk score trend; and performing corresponding early warning based on the comprehensive risk score trend and the risk escalation probability.

The application discloses a consumer industry enterprise closing risk early warning method and device and a storage medium, relates to the technical field of business data analysis, and comprises the following steps: multi-source data acquisition; cleaning, missing value filling, duplicate removal and standardization are performed on the collected data; a demand curve is fitted by using a least square method, a demand price elasticity coefficient is calculated, the sensitivity of commodity demand to price is judged, and market stability is evaluated; and an ARMA time series model is adopted to perform stationarity processing and modeling on historical sales data, and to predict short-term and medium-term sales trends. The application adopts demand elasticity analysis and ARMA time series prediction to perform data analysis, and converts qualitative market feelings into quantitative indexes; in the risk evaluation stage, the analytic hierarchy process is adopted to determine weights and the fuzzy comprehensive evaluation method, subjective judgment deviation is effectively reduced, and risk grade determination is more objective and accurate.

The present application relates to the technical field of intelligent operation and maintenance, in particular to a rail transit low-voltage device adaptive testing and predictive diagnosis method and system. By applying electrical excitation and mechanical vibration stress to the device to be tested and collecting multi-dimensional time series data, then using a waveform analysis model to extract deep features to identify early faults, using a sequence prediction model to analyze the performance degradation trend, and based on this, dynamically generating an adaptive optimal test path through a reinforcement learning model. Finally, with the help of a domain knowledge graph, correlation reasoning is performed to generate a comprehensive diagnostic report containing fault root causes and maintenance recommendations. The present application deeply integrates artificial intelligence into the testing process, realizes the transition from passive measurement to active cognition, significantly improves the depth, efficiency and intelligence level of testing, and provides forward-looking and accurate protection for the reliability of rail transit low-voltage devices.

This application relates to the technical field of CAD modeling, and more particularly to a parametric CAD modeling method, apparatus, device, and storage medium for fused views. It includes: acquiring multi-view images after alignment processing of the modeling target; extracting features from each view image to obtain view feature representations corresponding to each viewpoint, and performing pooling processing on all view feature representations to obtain visual feature labels; inputting the visual feature labels as keys into a preset sequence prediction model, whereby the sequence prediction model, during autoregressive generation, uses the currently generated modeling label sequence as a query to progressively predict the next modeling label until a complete modeling label sequence is generated as the prediction model sequence; and performing deserialization processing on the prediction model sequence to obtain the CAD model corresponding to the modeling target. This application can be adapted to common multi-view engineering drawing input scenarios, improving the convenience of CAD modeling applications.

This invention discloses a method and system for predicting pressure anomalies in gas pipelines based on time-series data mining, specifically relating to the field of gas transmission and distribution safety monitoring. This invention collects multi-source time-series pressure data from gas pipelines; performs data cleaning, missing value compensation, and multi-scale normalization preprocessing; constructs a multi-scale feature extraction module integrating wavelet transform and sliding window statistics to extract short-term fluctuation features, long-term trend features, and time-frequency domain features; employs an improved attention mechanism time-series prediction model for pressure sequence prediction; dynamically calculates adaptive thresholds based on prediction residuals, and combines this with continuous sliding window detection to determine anomalies; this invention effectively improves the accuracy of anomaly prediction and early warning capabilities, reduces false alarm rates, and adapts to dynamic changes in pipeline operating conditions.

The embodiment of the present specification provides a kind of control method, device, system, equipment and storage medium of production system, the method is applied to predictive control system, the predictive control system includes linear model and nonlinear model, the method comprises: at each sampling time, the following steps are iteratively executed until the target function value that meets the preset condition is solved: determine input sequence;The first output sequence that the linear model is predicted to the input sequence determined this time and the second output sequence that the nonlinear model is predicted to the input sequence determined this time are fused, and then substituted into the preset target function to obtain the target function value by calculation;Using the input sequence corresponding to the target function value that meets the preset condition solved, control instruction is generated to control the production system.

This invention discloses a muscle stimulation modulation system and method based on real-time dynamic optimization, comprising: a real-time signal acquisition module for continuously acquiring multimodal physiological signals of the target limb, including at least electromyography (EMG) signals, joint kinematic angle signals, and neural electrical signals; a state assessment and prediction module for calculating real-time human state indicators based on the multimodal physiological signals, fusing historical state data, and generating future state trend prediction information through a time series prediction model; and a dynamic parameter optimization module for solving the optimal combination of stimulation parameters for the neuromuscular stimulator in real-time using an embedded optimization algorithm, based on the real-time human state indicators and state trend prediction information, with minimizing the objective function as the core. The stimulation parameter combination includes stimulation pulse intensity, frequency sequence, and waveform duty cycle. This invention enables better muscle stimulation modulation.

The present application relates to artificial intelligence, deep learning, financial technology, time series prediction technology field, specifically refers to a kind of financial time series prediction method and system of frequency domain enhancement and morphology embedding, comprising: first, obtain stock market daily line level opening price, highest price, lowest price, closing price, transaction volume multidimensional K line data and do moving average smoothing;Again, reorder column dimension, extract K line morphology features by multilayer one-dimensional convolution and project into high-dimensional vector;In the training phase, the data is enhanced in the frequency domain, and only the high-frequency component is disturbed;The features are input into the iTransformer encoder to extract the time series features;Finally, the real prediction value is restored by adaptive inverse normalization.The present application realizes high-precision prediction of stock price by morphology embedding, frequency domain fine enhancement, adaptive inverse normalization and iTransformer modeling, eliminates the numerical cliff effect, improves the trend following and inflection point capturing ability, and enhances the overall robustness of the algorithm.