Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

52 results about "Multilayer perceptron neural network" patented technology

Hearth temperature field prediction method and equipment fusing actual measurement point data mechanism model

The invention belongs to the related technical field of coal-fired boiler detection, and discloses a hearth temperature field prediction method and equipment fused with an actual measurement point data mechanism model, and the method comprises the following steps: completing the installation of a radiation temperature measurement device based on a determined hearth radiation temperature measurement point, and obtaining the hearth radiation temperature data of each measurement point on line, inputting the current unit operation parameters and the hearth radiation temperature data into a hearth temperature field machine learning model, and outputting a hearth temperature field by the hearth temperature field machine learning model; wherein the hearth temperature field machine learning model based on the multilayer perceptron neural network is established, and the training data set is utilized to train the hearth temperature field machine learning model to obtain the hearth temperature field machine learning model. The method solves the problem that the prediction precision of an existing machine learning model on the hearth temperature field is low.
Owner:HUAZHONG UNIV OF SCI & TECH

Deep learning-based staged tensioning prestress long-term loss prediction method

The invention discloses a deep learning-based staged tension prestress long-term loss prediction method, which belongs to the field of deep learning, solves the problem of poor prediction result robustness of a traditional prestress measurement method, and comprises the following steps: obtaining bridge original data based on a standard drawing set and a bridge design specification; based on the construction environment conditions and the concrete age, concrete material attributes are calculated; based on bridge original data and concrete material attributes, finite element software is used for simulating prestressed reinforcement staged tensioning processes under different working conditions, and a bridge prestressed reinforcement staged tensioning loss data set is generated; and constructing a prestress loss prediction model based on the multi-layer perceptron neural network, processing the staged tensioning loss data set of the bridge prestressed reinforcement based on the prestress loss prediction model, and outputting a prediction result of long-term loss caused by shrinkage and creep of the prestressed reinforcement. According to the method, the stress loss of the prestressed reinforcement can be effectively predicted, and the prediction robustness and accuracy are improved.
Owner:JILIN JIANZHU UNIVERSITY

TR component gold wire bonding process parameter prediction method based on multilayer perceptron neural network

The invention discloses a TR assembly gold wire bonding process parameter prediction method based on a multilayer perceptron neural network, and belongs to the technical field of microwave device intelligent manufacturing. According to the method, an intelligent mapping model of gold wire bonding geometric parameters and radio frequency performance is constructed by fusing a multi-layer perceptron neural network and parameterized electromagnetic simulation. The method specifically comprises the following steps: generating 45 groups of samples in a process parameter space by adopting Latin hypercube sampling; obtaining an S parameter data set through batch processing electromagnetic simulation; box-Cox conversion and normalization preprocessing are carried out on the data; the method comprises the following steps: constructing an MLP neural network model of a 3-32-16-2 structure, and determining hyper-parameters by using Bayesian optimization; and after training is completed, rapid reverse mapping from target performance to process parameters is realized. According to the method, the number of traditional tests is reduced from more than 200 to 45, the predicted root-mean-square error of S21 is smaller than or equal to 0.12 dB, the determination coefficient is larger than or equal to 0.96, and the parameter backstepping time lt is obtained; according to the method, full-process automation from simulation, training, optimization to production and issuing is realized, and the development efficiency of the TR component is remarkably improved.
Owner:GUILIN UNIV OF ELECTRONIC TECH

Cable fault positioning method and system based on neural network learning

The invention relates to the field of electrical system fault detection, and provides a cable fault positioning method and system based on neural network learning. The method comprises the following steps: acquiring a cable traveling wave signal and preprocessing the cable traveling wave signal to obtain a preprocessed traveling wave signal; performing high-speed parallel acquisition on the preprocessed traveling wave signal to obtain multi-channel digital traveling wave data; performing multi-resolution decomposition on the multi-channel digital traveling wave data through a fast wavelet transform algorithm, and performing feature extraction on a wavelet coefficient obtained through decomposition to obtain a fault feature vector; and carrying out classification identification on the fault feature vector through a multilayer perceptron neural network, and carrying out fault position calculation on a classification result to obtain a fault positioning result. According to the invention, the cable fault positioning precision and reliability are improved.
Owner:SHANDONG KUANGWEI INFORMATION TECH CO LTD

Charging pile grouping method and system based on multilayer perceptron neural network

The invention relates to a charging pile grouping method and system based on a multilayer perceptron neural network, and the method specifically comprises the steps: calculating a boundary condition which needs to be satisfied by an individual charging pile error coefficient under a preset misjudgment rate and a missed judgment rate when a preset measurement guarantee scheme is adopted to determine an out-of-tolerance individual in a to-be-detected charging pile group; in combination with existing charging pile data, a pre-trained multi-layer perceptron neural network is adopted to obtain an incidence relation between the charging pile explicit tag and the error coefficient; analyzing the explicit labels of the to-be-grouped charging piles by using the obtained incidence relation, predicting error coefficients of the to-be-grouped charging piles, namely inputting the explicit labels of the to-be-grouped charging piles into a pre-trained multilayer perceptron neural network model, and calculating and outputting predicted values of the error coefficients of the charging piles; and grouping the to-be-grouped charging piles according to the boundary conditions of the error coefficients of the individual charging piles and the error coefficients of the to-be-grouped charging piles.
Owner:国网福建省电力有限公司营销服务中心 +3

Training method, verification method and device of waybill track verification model

The invention provides a waybill track verification model training method, verification method and device, and the method comprises the steps: carrying out the vectorization of the waybill data provided by a freight enterprise and the actual passing high-speed track data of a truck provided by a high-speed transaction platform, calculating a track vector included angle, calculating the difference value of auxiliary parameters, and carrying out the verification of the waybill track. And connecting the track vector included angle with the affiliated parameter difference value to obtain track tensor data. By training a neural network model, extracting local features in the trajectory index data by adopting a multilayer perceptron neural network with a residual error and a relatively strong local feature extraction capability; and finally, extracting global features in the trajectory index data by using a full connection layer with global perception capability to obtain a verification result, thereby realizing automatic matching verification judgment on the waybill data and the actual passing high-speed trajectory data of the truck. Under the condition of large data volume, data features can be accurately extracted, the analysis speed is kept, and a result can be obtained within milliseconds.
Owner:XINGYUNSHUJU (BEIJING) TECH CO LTD

Optical element pollution layer distribution prediction method based on neural network

The invention relates to the technical field of optical processing, in particular to an optical element pollution layer distribution prediction method based on a neural network. Comprising the following steps: carrying out optical element processing experiments under n groups of different processing parameters, collecting process parameter data, and establishing n groups of data sets; collecting pollution layer distribution data corresponding to the process parameter data; dividing the n groups of data sets into a training set and a test set in proportion, and training the multilayer perceptron neural network by using the training set to obtain a neural network prediction model; the method comprises the following steps: acquiring new process parameters in an optical element pollution layer distribution prediction task, inputting a new process parameter vector and copying three parts, respectively inputting the three parts of process parameter vectors into three sub-networks of a neural network prediction model, respectively outputting to obtain neural network prediction coordinate sub-matrixes # imgabs0 #, # imgabs1 # and # imgabs2 #, inputting the neural network prediction coordinate sub-matrixes into a reconstruction module for combination, and finally obtaining a prediction result of the optical element pollution layer distribution prediction task according to the neural network prediction coordinate sub-matrixes # imgabs0 #, # imgabs1 # and # imgabs2 #. And obtaining predicted pollution distribution point cloud data. The method has the advantages of no need of re-sample preparation analysis, low cost and rapid reaction.
Owner:CHANGCHUN INST OF OPTICS FINE MECHANICS & PHYSICS CHINESE ACAD OF SCI

Bridge prestressed reinforcement elastic compression loss prediction method and system and storage medium

The invention relates to a bridge prestressed reinforcement elastic compression loss prediction method and system and a storage medium. The method comprises the following steps: simulating a bridge model through finite element software; the sum of all the prestressed reinforcement units is calculated, and temperature loads applied to all the reinforcement units are calculated; according to a prestressed reinforcement tensioning sequence and a life-death unit method, temperature loads are applied to reinforcement units one by one, and calculation is carried out; the method comprises the following steps: constructing an original data set, preprocessing, and dividing to obtain a training set, a verification set and a test set; constructing an MLP multilayer perceptron neural network model as a prediction model of the elastic compression loss of the bridge prestressed reinforcement to the concrete, and training the prediction model; and utilizing the trained prediction model to predict concrete elastic compression prestress loss caused by tensioning other prestressed reinforcements by the prestressed reinforcements. According to the invention, the elastic compression loss prediction of the bridge prestressed reinforcement can be quickly and accurately realized.
Owner:JILIN JIANZHU UNIVERSITY

A Depression Monitoring and Intervention System Based on Internet Hospital and Heart Rate Variability

PendingCN122074982AAchieve deep optimizationSolve the problem of lack of personalized adjustment mechanismHealth-index calculationMedical automated diagnosisNerve networkMild depression
This invention relates to the field of depression monitoring and intervention technology, and discloses a depression monitoring and intervention system based on an internet hospital and heart rate variability (HRV). The system includes: a multimodal physiological signal acquisition module for simultaneously acquiring raw HRV signals, skin conductance fluctuations, and respiratory rate data; a signal processing module for preprocessing and feature extraction of the raw signals to obtain multidimensional feature vectors; and a depression level inference module that constructs a multilayer perceptron neural network, taking as input the multidimensional feature vector containing multidimensional HRV features, skin conductance fluctuations, and respiratory rate statistics, and outputting probability values ​​for no depression, mild depression, and moderate to severe depression levels. This invention establishes an intelligent management mechanism for internet hospitals for the first time, collecting multidimensional physiological data of patients with depression in real time through wearable monitoring units. Based on the dynamic trends of physiological data characteristics, it can serve as an auxiliary diagnostic tool for psychiatry, especially suitable for early screening and remote follow-up of depression, reducing the consumption of medical resources.
Owner:RENJI HOSPITAL AFFILIATED TO SHANGHAI JIAO TONG UNIV SCHOOL OF MEDICINE

Method and device for deviation alarm processing based on temperature measurement points

This invention discloses a deviation alarm processing method and device based on temperature measurement points. The method includes: obtaining temperature change data from each temperature sensor installed at a designated location on the compressor unit through a preset detection cycle; acquiring sensor judgment parameters corresponding to the current operating condition type based on the current operating condition type received before the compressor unit starts operation; determining whether the temperature change data is abnormal under the current operating condition type using the sensor judgment parameters and the temperature change data from each temperature sensor; issuing an early warning when the temperature change data is determined to be abnormal, and capturing compressor unit log data; inputting the compressor unit log data into an intelligent judgment model, and outputting the cause of the compressor unit abnormality; the intelligent judgment model is pre-trained on a multilayer perceptron neural network model using historical compressor unit log data and the cause of the compressor unit abnormality. This invention can reduce unnecessary shutdowns of the unit and stabilize production.
Owner:RICHFIT INFORMATION TECH +1

Water lubricated bearing performance calculation method and device based on physical information neural network

The invention provides a water-lubricated bearing performance calculation method and device based on a physical information neural network, and the method comprises the steps: inputting related physical parameters of a water-lubricated bearing sample into a multilayer perceptron neural network, and obtaining a dimensionless pressure field in the water-lubricated bearing sample, the related physical parameters comprise dimensionless circumferential coordinates x and dimensionless axial coordinates z; according to the dimensionless pressure field and the dimensionless membrane thickness of the water lubricated bearing sample, determining a Reynolds equation residual error, absolute pressure non-negative constraint and boundary condition constraint so as to obtain a total loss function, and training the multilayer perceptron neural network according to the total loss function; inputting the dimensionless circumferential coordinate x and the dimensionless axial coordinate z of the target water-lubricated bearing into the trained multilayer perceptron neural network to obtain a dimensionless pressure field in the target water-lubricated bearing; and calculating a performance index according to the dimensionless pressure field in the target water lubricated bearing. The efficiency and precision of bearing performance calculation are improved.
Owner:SANYA SCI & EDUCATION INNOVATION PARK WUHAN UNIV OF TECH

Method for predicting three-dimensional engine flow field based on multi-layer perceptron neural network

The present application relates to a kind of based on multilayer perceptron neural network prediction three-dimensional engine flow field method, first constructs multilayer perceptron neural network and carries out training to multilayer perceptron neural network, then target OpenFOAM file is preprocessed to obtain normalized input data, finally, the multilayer perceptron neural network of training completion is batched with normalized input data and obtains the prediction three-dimensional engine flow field corresponding to target OpenFOAM file.The method of the present application, point cloud data can be accurately converted into the format suitable for the training of multilayer perceptron neural network from the data preprocessing of OpenFOAM software, and the prediction of three-dimensional engine flow field is realized by using multilayer perceptron neural network.Compared with the traditional computational fluid dynamics simulation method for predicting three-dimensional engine flow field, the prediction efficiency is high and the reliability is high.
Owner:NORTHWESTERN POLYTECHNICAL UNIV

A Method and Apparatus for Calculating the Performance of Water-Lubricated Bearings Based on Physics Information Neural Networks

This invention provides a method and apparatus for calculating the performance of water-lubricated bearings based on a physical information neural network. The method includes: inputting relevant physical parameters of a water-lubricated bearing sample into a multilayer perceptron neural network to obtain a dimensionless pressure field within the water-lubricated bearing sample; the relevant physical parameters include dimensionless circumferential coordinates x and z. Based on the dimensionless pressure field and dimensionless film thickness of the water-lubricated bearing sample, the Reynolds equation residuals, absolute pressure non-negativity constraints, and boundary condition constraints are determined to obtain a total loss function; the multilayer perceptron neural network is trained based on the total loss function; the dimensionless circumferential coordinates x and z of a target water-lubricated bearing are input into the trained multilayer perceptron neural network to obtain the dimensionless pressure field within the target water-lubricated bearing; and performance indicators are calculated based on the dimensionless pressure field within the target water-lubricated bearing. This invention improves the efficiency and accuracy of bearing performance calculation.
Owner:SANYA SCI & EDUCATION INNOVATION PARK WUHAN UNIV OF TECH

Mechanical system constraint following control method based on historical state information neural network

The invention provides a mechanical system constraint following control method based on a historical state information neural network, and relates to the technical field of automatic control, and the method specifically comprises the following steps: building a constraint-containing mechanical system dynamic model, and designing a nominal control part based on a U-K theory; establishing a mechanical system lumped disturbance model, constructing a feature vector including the historical state of the mechanical system, and obtaining an offline training data set; using a multilayer perceptron neural network to learn inverse dynamic mapping in an off-line manner, and establishing a lumped disturbance estimation model; performing low-pass filtering smoothing processing on the original prediction of the multi-layer perceptron neural network, and forming a total control law in combination with nominal control force to be applied to a mechanical system; and performing simulation verification on the mechanical system. According to the technical scheme, the problems that in the prior art, robust control is high in conservative property, buffeting is likely to be caused by high gain, and characteristics are not matched and signals shake in a pure data driving method are solved.
Owner:SHANDONG UNIV OF SCI & TECH

Hybrid model-based wind power medium-and-long-term generating capacity prediction method and device

The invention discloses a hybrid model-based wind power medium-and-long-term generating capacity prediction method and device, and belongs to the technical field of wind power generation. The method comprises the following steps: acquiring historical actually measured wind power data and historical meteorological data as input, and obtaining a cleaned power generation time sequence; performing discrete wavelet decomposition on the cleaned power generation time sequence to obtain multiple layers of subsequences of different frequency bands; based on the sub-sequence and the historical meteorological data, selecting an optimal prediction feature subset from the data; constructing a plurality of multi-level perceptron neural network models running in parallel, and outputting corresponding prediction results; and performing wavelet recombination on all the output prediction results to obtain a final prediction value of the medium-and-long-term generating capacity of the wind power. According to the technical scheme, high-precision prediction of the medium-and-long-term generating capacity of wind power is achieved, abnormal data can be effectively removed, multi-scale features are extracted, multi-model collaborative learning is achieved, and the reliability and adaptability of a prediction result are improved.
Owner:中国船舶集团风电发展有限公司 +1

Pixel antenna multi-objective optimization method based on BPSO-MLP algorithm

The invention relates to a pixel antenna multi-objective optimization method based on a BPSO-MLP algorithm, and is applied to the technical field of wireless communication antennas. Global search is carried out in a high-dimensional discrete coding space of a pixel antenna through an improved binary particle swarm optimization algorithm, a dynamic temperature parameter is introduced to adjust the shape of a Sigmoid mapping function, and a probability-driven variation mechanism is adopted to effectively balance exploration and development capabilities; a performance prediction model is established through the designed multilayer perceptron neural network, and rapid and accurate prediction of key performance indexes such as bandwidth, gain and radiation pattern of a candidate antenna structure is realized; and submitting the predicted and screened high-potential solution to full-wave electromagnetic simulation through an intelligent optimization mechanism, feeding back a simulation result to an optimization algorithm, and dynamically updating a prediction model. According to the method, the problems of high-dimensional discrete search, high simulation cost and multi-performance index coordination in pixel antenna multi-target optimization are effectively solved, and an efficient solution is provided for reconfigurable antenna design of a new-generation communication system.
Owner:THE 54TH RESEARCH INSTITUTE OF CHINA ELECTRONICS TECHNOLOGY GROUP CORPORATION +2

A Multi-Image Encryption Method Based on 3D Face Key and Multiplexed Digital Holography

This invention discloses a multi-image encryption method based on 3D face keys and multiplexed digital holography. The encryption steps include: using 3D face high-order data acquisition technology, chaotic face structured light phase mask generation technology, grating modulation technology, multiplexed digital holographic coding technology, and watermark embedding technology based on discrete wavelet domain singular value decomposition, multiple plaintext information images can be encrypted into an amplitude-type holographic ciphertext and embedded into a host image; the decryption steps include: the user must first calculate the face similarity using a 3D face multilayer perceptron neural network. If the similarity exceeds a preset threshold, authentication is successful. Then, the final decryption result is generated using watermark image extraction technology based on discrete wavelet transform, image decryption technology based on the Cramer-Kroni relation, and spectral filtering technology; otherwise, authentication fails. This method has advantages such as a large key space, high sensitivity, strong robustness of the 3D face key, and low decryption crosstalk.
Owner:XIAN TECH UNIV

An Optimization Method for High Torque, Low Vibration Built-in Permanent Magnet Synchronous Motor

This invention relates to the field of permanent magnet synchronous motor optimization design, specifically to an optimization method for a high-torque, low-vibration built-in permanent magnet synchronous motor, comprising the following steps: S1 Calculate the electromagnetic field distribution of the motor using the finite element method, and determine the key noise sources of the motor based on the calculation results; S2 Define the optimization objectives and constraints, construct an optimization mathematical model, and use the random forest algorithm to perform sensitivity analysis on the design variables, dividing the design parameters into strongly sensitive, moderately sensitive, and weakly sensitive levels; S3 For the strongly sensitive parameters, construct a multilayer perceptron neural network surrogate model, and use a second-generation non-dominated sorting genetic algorithm for multi-objective optimization; S4 For the moderately sensitive parameters, use a response surface model surrogate model, and combine it with a multi-objective particle swarm optimization algorithm for optimization; S5 For the weakly sensitive parameters, use a parametric scanning method, and combine the weights of different objectives with a weighted summation method to comprehensively optimize the objective function; S6 Synthesize and verify the optimization results.
Owner:HUAQIAO UNIVERSITY

Attraction control system and method using multilayer perceptron neural networks and machine learning

A controller is used to provide a dynamic ride profile for a ride system. The controller is used to access a data set indicative of conditions in an environment of the ride system in response to an alert indicative of a hindrance for the ride system in the environment. The controller is used to determine a classification from a plurality of classifications for the environment, via a predictive engine trained using supervised machine learning, based on the data set and a ride profile of the ride system. The controller is used to determine one or more recommendations, via the predictive engine, based on the data set and the classification and visually or audibly presenting the classification with the one or more recommendations.
Owner:UNIVERSAL CITY STUDIOS LLC

Agricultural machine driver operation state sensing system based on digital twinning

The invention discloses an agricultural machinery driver operation state sensing system based on digital twinning, and belongs to the technical field of agricultural machinery intelligence and digital twinning. The invention aims to solve the problems of incomplete operation state sensing, inaccurate risk early warning and low man-machine ring cooperation efficiency caused by single sensing dimension, isolated data mode, static risk assessment and lack of closed-loop control in the existing agricultural machine state monitoring technology. According to the method, three types of state vectors of machinery, a driver and an environment are synchronously collected through a multi-source sensor, and after preprocessing, a fusion feature vector is generated by adopting a weighted multi-mode fusion model. Based on the vector, performing state classification by using a multilayer perceptron neural network, and describing the interaction between a machine and a driver state in combination with a coupling constraint dynamic model; meanwhile, a fatigue index and an environment risk index are calculated, and real-time accurate recognition of the operation state is achieved through a comprehensive discrimination function. And a discrimination result is fed back to the digital twin virtual model in real time, and online parameter optimization and closed-loop control are driven.
Owner:NORTHEAST AGRICULTURAL UNIVERSITY

A neural network-based intelligent early warning diagnosis method and system for solenoid valves

The application discloses an electromagnetic valve intelligent early warning diagnosis method and system based on a neural network, relates to the technical field of industrial equipment state monitoring, and comprises the following steps: synchronously collecting instruction, coil current and outlet pressure signals of the electromagnetic valve, extracting multi-dimensional time domain and state characteristics after pretreatment; a multilayer perceptron neural network is constructed and trained, the distribution mode of normal data in a characteristic space is learned to output an abnormal probability; a threshold determination result based on a physical rule is combined with the abnormal probability of the neural network to realize fine diagnosis of four levels of health states, i.e., normal, degradation, early warning and failure; when the state is abnormal, the fault types of an electromagnetic coil fault, a valve core sticking and an internal leakage of the electromagnetic valve are accurately identified through a preset rule base according to characteristic combinations such as current, pressure response and holding capacity. The application improves early warning capability and diagnosis precision of faults, and provides an effective intelligent solution for predictive maintenance of the electromagnetic valve.
Owner:YANSHAN UNIV

Programmable metasurface regulation and control method and device, electronic equipment and medium

The invention relates to the technical field of communication, and discloses a programmable metasurface regulation and control method and device, electronic equipment and a medium, and the method comprises the steps: receiving an analog radio frequency signal outputted by an antenna array, and carrying out the preprocessing; a covariance matrix is utilized to decompose and simulate a radio frequency signal, the two-dimensional arrival angles of a target signal and an interference signal are estimated, a minimum variance undistorted response algorithm is adopted to construct an intelligent beam forming model, and the optimal weight of each unit of the programmable metasurface is calculated; inputting the optimal weight into a lightweight multilayer perceptron neural network model, and mapping the optimal weight into a discrete phase control matrix; according to the method, the minimum variance undistorted response algorithm and the lightweight multi-layer perceptron neural network model are combined, the generalization ability is guaranteed, the parameter scale and the calculation complexity are greatly reduced, programmable metasurface phase regulation is carried out, and the performance of the sensor is improved. And even in a complex scene with multiple overlapped interference sources, the interference can be stably suppressed.
Owner:BEIJING JIAOTONG UNIV +1

Photovoltaic generating capacity prediction method based on AI

The invention discloses an AI-based photovoltaic generating capacity prediction method. The method comprises the following steps: S1, acquiring meteorological and photovoltaic system data of different regions; s2, inputting the processed meteorological and photovoltaic system data of different regions into an initial prediction model based on an MLP (Multilayer Perceptron Neural Network); and S3, establishing constraint conditions defined by combining the data with single-variable and cross-variable rules, and carrying out multi-dimensional judgment on the data rationality. And S4, sorting data, reports and verification results to form a final photovoltaic power generation capacity prediction result report. The method mainly solves key problems existing in an existing photovoltaic generating capacity prediction method, achieves remarkable improvement of prediction precision and reliability, and has important practical significance and wide application prospects. The method not only improves the prediction precision and reduces the prediction error rate, but also enhances the generalization ability of the model, so that the model can better adapt to different environment and climate changes.
Owner:BEIJING UNIV OF TECH

Cutting analogue simulation method and system based on dynamic characteristics of machine tool

The invention discloses a cutting simulation method and system based on dynamic characteristics of a machine tool, and the method comprises the steps: constructing a cutting power simulation model which comprises a workpiece geometric model and a tool geometric model, and constructing a dynamic characteristic equivalent model based on the cutting power simulation model, thereby achieving the cutting simulation of a workpiece; meanwhile, in the process of constructing the dynamic characteristic equivalent model, a multi-degree-of-freedom spring-damper multi-node space distribution subprogram is adopted to simulate the vibration characteristic in the cutting process, and modal parameters in the actual cutting process are calibrated through a multi-layer perceptron neural network algorithm; iterative simulation of dynamic coupling of the cutting force and the cutting mode is achieved, the problem that an existing cutting simulation method ignores real dynamic characteristics of a machine tool-tool-workpiece system is solved, and the cutting force, vibration response and temperature prediction precision are improved.
Owner:HANGZHOU DIANZI UNIV

Small sample yarn tension prediction method based on physical information neural network

The invention relates to a small sample yarn tension prediction method based on a physical information neural network. The method comprises the following steps: constructing a yarn tension theoretical model; testing the output tension of the yarn under the conditions of different initial tension, contact wrap angles and cylindrical guide roller radiuses in a preset process parameter range to form a small sample experiment data set; fitting specific values of power law parameters in the yarn tension theoretical model by using the small sample experiment data set; training a multi-layer perceptron neural network by using the small sample experiment data set and the power law parameter specific value, introducing the yarn tension theoretical model into a neural network training process as a physical constraint condition, and forming a physical information neural network prediction model; in the training process, the neural network parameters are iteratively updated by minimizing a joint loss function containing a data loss item and a physical constraint loss item. The technical problems of high experiment cost, low calculation efficiency and insufficient small sample prediction precision of the existing yarn tension prediction are solved.
Owner:NANJING FIBERGLASS RES & DESIGN INST CO LTD +2

An intelligent traditional Chinese medicine acupuncture point selection method and system based on meridian syndrome differentiation

The application provides an intelligent Chinese medicine acupuncture point selection method and system based on meridian syndrome differentiation. Through three-dimensional reconstruction model, high-precision modeling of patient body surface feature points is performed, combined with statistical shape model and re-projection error minimization method, individual body surface three-dimensional structure can be accurately restored, and accurate geometric reference is provided for subsequent acupoint mapping. The spatial mapping relationship between the body surface feature points and the meridian acupoint feature points is learned by using a multilayer perceptron neural network, which can effectively capture the nonlinear individual differences and realize intelligent deduction from the body surface morphology to the acupoint position. A virtual engine generates a large number of diversified virtual patient images, and a convolutional neural network is trained to directly detect the two-dimensional coordinates of the acupoints from the real images, thereby improving the generalization ability and robustness of the network in actual application. The acupoint coordinates obtained by weighted fusion of the geometric projection and the deep learning detection of the two methods are fully utilized to improve the acupoint selection accuracy.
Owner:REHABILITATION UNIVERSITY QINGDAO CENTRAL HOSPITAL

Universal thin-film material automatic defect identification and material performance mapping prediction system

The invention provides a universal thin-film material automatic defect identification and material performance mapping prediction system, and the system comprises an image collection unit which is used for obtaining high-resolution microscopic images of a material substrate and a thin film; the image preprocessing unit is configured to perform registration, effective area cutting and adaptive grid division on the image; the defect identification unit is used for realizing automatic defect detection based on an improved standardized cross-correlation template matching algorithm; the performance prediction unit adopts a multi-layer perceptron neural network model comprising at least two hidden layers; and the visual interaction unit provides a graphical interface to support artificial review and regional analysis. The universal thin-film material automatic defect identification and material performance mapping prediction system provided by the invention has the beneficial effects of low cost, high precision, high efficiency, high universality, strong adaptability and the like.
Owner:SHENZHEN UNIV

Electric vehicle regenerative braking distribution method based on neural network

The invention relates to the technical field of electric vehicle control, in particular to an electric vehicle regenerative braking distribution method based on a neural network, and the method comprises the following steps: S1, collecting the state information of a vehicle in real time; s2, the collected state information is preprocessed, and a multi-dimensional feature vector X is constructed; s3, inputting the multi-dimensional feature vector into a pre-trained multi-layer perceptron neural network model, and outputting a motor braking ratio alpha between 0 and 1 by the neural network model; and S4, calculating distribution values of motor braking torque and mechanical braking torque according to the motor braking proportion alpha and the total required braking torque of the vehicle, and issuing a corresponding control instruction. According to the method, the pre-trained neural network model is used for replacing complex online optimization calculation, the requirement for the calculation power of the vehicle-mounted controller is lowered, and therefore the real-time requirement for brake control is met.
Owner:CHONGQING VEHICLE TEST & RES INST CO LTD

Method and device for fast prediction of three-dimensional nozzle flow field and sensitivity parameter analysis

The application provides a three-dimensional nozzle flow field rapid prediction and sensitive parameter analysis method and device, wherein the method comprises the following steps: obtaining nozzle geometric shape parameters, performing feature extraction on the geometric shape parameters by using a self-encoder to obtain geometric feature parameters; reconstructing the nozzle geometric shape parameters through the geometric feature parameters, taking nozzle flow state parameters and flow control parameters as design variables, and obtaining nozzle flow field training data sets through CFD simulation; building a multilayer perceptron neural network, training the multilayer perceptron neural network through the training data sets, introducing a flow control equation into a loss function of the training, and obtaining a flow field prediction model; and realizing three-dimensional flow field rapid prediction of specified nozzle geometric feature parameters, flow state parameters and flow control parameters through the flow field prediction model to obtain flow field information of the specified nozzle. The application realizes high-precision and rapid prediction of three-dimensional flow field distribution and completes sensitive parameter analysis of the nozzle.
Owner:TSINGHUA UNIVERSITY +1

Electric fish signal detection method based on Mel-frequency cepstrum and multilayer perceptron

The invention discloses an electric fish signal detection method based on Mel-cepstrum and a multilayer perceptron, and belongs to the field of electric fish signal detection, and the method comprises the steps: firstly collecting an acoustic emission signal generated by underwater discharge, and carrying out the preprocessing of pre-emphasis, framing and windowing, so as to improve the signal-to-noise ratio; then extracting Mel-frequency cepstrum coefficient characteristics of the signals, and performing first-order and second-order difference calculation optimization and fusion to form a characteristic parameter set with static and dynamic characteristics; and finally, inputting the characteristic parameters and the detection voltage value into a multi-layer perceptron neural network for nonlinear fitting, and reconstructing high-precision electric signal intensity. According to the method, the acoustic emission signal characteristics are effectively utilized, and the sensitivity and the confidence coefficient of underwater electric fish signal detection are remarkably improved.
Owner:HUNAN UNIV OF SCI & TECH SANYA RES INST