526 results about "Root mean square" patented technology

The invention discloses a self-adaptive threshold SAMP reconstruction method for a power quality disturbance signal. According to the method, a compression observation value is obtained by constructing a random Gaussian observation matrix, and sparse representation is carried out on an original signal by using discrete Fourier transform. In the iterative reconstruction process, the spectrum amplitude difference is introduced for the first time to serve as an adaptive termination basis, and automatic adaptation of different noise levels and different disturbance characteristics is achieved in combination with a dynamic threshold update function. According to the method, the problems of traditional SAMP sparseness overestimation and redundant iteration are effectively avoided, and the calculation load is remarkably reduced. Compared with an OMP method, an original SAMP method and the like, the method has the advantages that the number of iterations can be reduced by 30%-60%, the reconstruction signal-to-noise ratio is increased by 2-5 dB, the root-mean-square error is reduced by 10%-25%, higher robustness and real-time performance are achieved in power quality disturbance signal reconstruction, and the method is quite suitable for scenes such as compressed sampling, edge calculation and high-speed signal reconstruction in a power quality monitoring system.

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.

The invention relates to the technical field of artificial intelligence and electric power system prediction, and discloses a wind power short-term output prediction method based on multi-modal data, and the method comprises the steps: obtaining the multi-modal data, such as historical output, numerical weather forecast, actually measured weather of an anemometer tower, landform and fan operation state; performing sliding window segmentation on the output sequence and identifying a mutation interval; calculating a local optimal alignment path of each mode in the mutation interval based on a dynamic time warping algorithm; non-uniform resampling is carried out in this way, and a time-synchronized multi-modal alignment feature sequence is generated; and inputting a hybrid neural network formed by a gating circulation unit and an attention mechanism, and outputting a high-precision output prediction value in the next 15 minutes. The system comprises corresponding function modules. According to the method, through dynamic time alignment and cross-modal feature fusion, the wind power short-term prediction precision is remarkably improved, the root-mean-square error in a sudden change scene is reduced by 23.7%, and reliable support is provided for power grid dispatching.

The invention discloses a power distribution network transient characteristic prediction method based on supervised learning, and relates to the technical field of power distribution network state prediction, and the method comprises the steps: collecting historical operation data through a power distribution network monitoring system, carrying out the data preprocessing, and obtaining standardized multi-dimensional time series data; carrying out transient feature extraction, constructing a high-dimensional feature set, and carrying out feature dimension reduction according to a transient event tag to generate a feature subset; inputting the feature subset into a mixed supervised learning model of a gradient boosting decision tree GBDT and a long short-term memory network LSTM for joint training to obtain a transient feature prediction result; and calculating a root-mean-square error according to the transient characteristic prediction result and the real-time monitoring observation value of the power distribution network, and dynamically adjusting hyper-parameters of the supervised learning model based on a Bayesian optimization algorithm. According to the method, the detection accuracy can be improved, the calculation complexity can be reduced, and the discrimination capability and the time sequence prediction capability of the model are considered.

The invention discloses a key parameter long-time-sequence prediction method for a complex process industry, and the method comprises the steps: collecting multivariable sensor data in the process industry, and constructing a high-dimensional long-time-sequence prediction data set; constructing a PatchConvRNN prediction model by combining time slice embedding, dimension decoupling convolution, depth separable convolution and a recurrent neural network based on a sequence-to-sequence normal form; a point value-statistical mixed loss function is adopted, the point prediction precision, the sequence mean value and the standard deviation consistency are optimized at the same time, a prediction model is trained in combination with an optimization algorithm, and network model parameters are adjusted; and comprehensively evaluating the prediction model through a root-mean-square error, an average absolute percentage error and a standard deviation average absolute error. According to the method, high-precision prediction and fluctuation maintenance of the key time sequence variables under the complex working condition of the industrial process are achieved, and powerful support is provided for quality control and predictive maintenance of the production process.

The invention relates to the technical field of attitude determination of a photoelectric pod of an unmanned aerial vehicle, in particular to an MEMS-IMU dual-mode correction attitude determination method for the photoelectric pod of the unmanned aerial vehicle. Comprising the following steps: establishing a damping system second-order transfer function model, and outputting an attenuation coefficient matrix; based on MEMS-IMU three-axis motion parameters, distinguishing a large maneuver turning state and a linear flight state; respectively compensating a roll angle, a pitch angle and a course angle of the MEMS-IMU by using the main inertial navigation according to different states; converting the corrected attitude angle into a quaternion, and optimizing and outputting a fused quaternion through a gradient descent method; taking the fused quaternion as an input, constructing a seven-dimensional state space model, and outputting an error compensation amount through Kalman filtering; dynamically adjusting the main inertial navigation weight according to the accelerated speed root mean square; and outputting a final attitude angle in combination with the error compensation amount and the weight, and verifying the final attitude angle through GNSS (Global Navigation Satellite System) position inversion. The method has the advantages that cross interference is avoided; therefore, the method is suitable for the photoelectric pod scene of the unmanned aerial vehicle.

The invention provides a traffic flow prediction method combining a dynamic graph neural network and a Mama mechanism. The future short-term traffic flow is predicted by using historical traffic data. According to the method, firstly, normalization preprocessing is carried out on traffic state data collected by multiple sensors, a training sample is generated by adopting a sliding window, and a traffic flow value in the next one hour is predicted according to data in the past 24 hours. On the basis of the model structure, a time modeling module composed of multiple layers of MambaBlocks is constructed and used for capturing historical time sequence dependence; constructing a spatial modeling module of dynamic graph convolution, and combining a static adjacency matrix of the road network with a learnable adaptive adjacency structure to extract spatial association; and finally, the outputs of the modules are fused, and a prediction result is obtained through a prediction output module. In the training process, a Huber loss function is used as an optimization target, and evaluation indexes such as a mean absolute error (MAE), a root mean square error (RMSE) and a mean absolute percentage error (MAPE) are used for evaluating the performance of the model. According to the method, the traffic space-time dynamic characteristics are effectively mined, and the long-range dependence modeling capability and the prediction precision are improved.

The invention discloses a face gear planetary transmission system dynamic uniform load test platform and test method, and belongs to the field of gear transmission, and the test platform comprises a rack, a drive motor, a step-up gear box, a face gear planetary transmission system, a magnetic powder brake, a torque sensor, a slip ring, a stress strain test system and the like. The face gear planetary transmission system comprises an input face gear, a planet wheel, a planet carrier, a fixed face gear and the like. According to the dynamic uniform load test, stress values at the tooth root of the face gear and the tooth root of the planetary gear are measured mainly through strain gauges, and the corresponding relation between the tooth root stress and the meshing force at the gear meshing point is obtained through finite element simulation. And calculating the dynamic uniform load coefficient according to the dynamic meshing force root-mean-square values and the average values obtained at the tooth roots of the face gear and the planetary gear. By testing and analyzing the dynamic uniform load coefficient, the load distribution characteristic of the face gear planetary transmission system can be accurately revealed, a scientific basis is provided for structural parameter optimization, meshing performance improvement and reliability design of the face gear planetary transmission system, and then accurate and high-performance design of the transmission system is achieved.

The invention discloses a construction method of a multi-underwater robot cooperative control system, which relates to the field of underwater robots, and establishes a multi-parameter coupled data acquisition and preprocessing mechanism by using a Doppler current profiler ADCP, an inertial measurement unit IMU and controller software. A combined algorithm of sliding median filtering and Kalman filtering is adopted, and data disturbance caused by environment noise and attitude drift is effectively removed; and through abnormal point triple standard deviation determination and double adjacent point interpolation completion, stable correction of sampling data is realized. A standardized data set obtained after normalization processing unifies parameter scales of different physical magnitudes, so that the local average flow velocity, the flow velocity fluctuation root mean square and the average acceleration can be directly compared in a feature space. Therefore, according to the method, the high-consistency expression of the water flow disturbance characteristics in time and space is realized, so that the synchronous measurement precision of multiple underwater robots in a non-uniform flow field is remarkably improved.

The invention provides a vibration nonlinear signal energy analysis method and system based on variational mode decomposition, and relates to the technical field of signal processing, and the method comprises the following steps: obtaining a monitoring signal of a vibratory roller, determining a corresponding state based on root-mean-square data of the monitoring signal, and determining the state of the monitoring signal; performing segmentation processing on the monitoring signal based on the state corresponding to the monitoring signal to obtain a signal segmentation result; performing variational mode decomposition processing on the signal segmentation result to obtain at least two second mode components; performing Hilbert transformation processing on the second modal component, and combining the instantaneous frequency and the instantaneous amplitude of each modal component obtained through transformation to obtain frequency spectrum information of the monitoring signal; and performing marginal spectrum calculation and integration on the frequency spectrum information to obtain the total energy of the monitoring signal. According to the method, interference signals can be efficiently identified and eliminated in the vibration signals, so that the precision and robustness of compaction quality evaluation are improved.

The invention relates to the field of parameter identification, and provides a photovoltaic model parameter identification method and system based on an improved artificial bee colony algorithm, and the method specifically comprises the steps: obtaining a to-be-identified photovoltaic model, and constructing an optimization objective function which is based on a root-mean-square error; initializing an artificial bee colony algorithm and performing terrain complexity evaluation to obtain a terrain type, the terrain complexity evaluation being based on the population fitness difference matrix; and according to the terrain type, adaptive search is carried out to obtain a final identification result, and the adaptive search is based on a smooth terrain processing mechanism and a rugged terrain processing mechanism. According to the method, the defect of a traditional artificial bee colony algorithm facing a complex terrain area is avoided, and the accuracy and efficiency of photovoltaic model parameter identification are improved.

The invention provides a multi-objective structure optimization method and system suitable for a wheel excavator. Comprising the steps that a three-dimensional model which has the same actual size as an experimental test prototype of the wheel excavator and comprises a cab, an upper frame and other related parts is established; and comparing frequency domain response curve results obtained by experimental tests and verifying the accuracy of vibration performance prediction of the finite element model. Sensitivity analysis is carried out, and parts finally used for follow-up multi-target optimization are screened out according to the influence degree on the performance of the multiple targets; and performing test design according to the screened parts to obtain sample points, and importing the sample points into a finite element model for simulation to obtain a sample response data set. And fitting the agent model according to the sample data, optimizing by using a genetic algorithm to obtain an optimal solution, and sorting according to the magnitude of an acceleration root mean square value to select an optimal scheme. And importing the optimal scheme obtained according to the steps into a finite element model, and comparing the optimal scheme with an original scheme to verify the rationality of a new optimization scheme.

The invention relates to the technical field of cutter residual life prediction, and discloses a cutter residual life prediction method based on line angle attention and contrast drive aggregation, which comprises the following steps of: extracting six types of statistical characteristics including a mean value, a standard deviation, a median, an absolute maximum value, a root mean square and skewness by using original data of a multi-channel sensor in a cutter cutting process; a dual feature dimension reduction strategy of Pearson's correlation coefficient and grey correlation analysis is adopted, key features strongly related to the wear state are screened, and standardization processing is carried out; and constructing a deep learning architecture fusing line angle attention and contrast driving feature aggregation. According to the method, the model has higher recognition capability on the characteristic mode of numerical jump but consistent trend in the tool wear process, the problem that a traditional attention mechanism is prone to losing key time sequence association in the nonlinear degradation process is solved, and the modeling precision of non-stationary sensor data under the complex cutting working condition is remarkably improved.

The invention discloses a time sequence reconstruction and machine learning coupled chlorophyll monitoring data breakpoint restoration method, and belongs to the technical field of water quality monitoring. The invention discloses a chlorophyll monitoring data breakpoint restoration method based on coupling of time sequence reconstruction and machine learning, and the method comprises the following steps: S1, collecting water quality monitoring data, and cleaning the monitoring data to obtain preprocessed data; s2, performing time sequence reconstruction on the preprocessed data to obtain a weekly average 1 data set; s3, respectively constructing a radial basis function neural network model and a back propagation neural network model by taking the chlorophyll concentration as a response variable and the conventional water quality parameter as a predictive variable; s4, performing performance evaluation on each model by taking a root mean square error, an average absolute percentage error, goodness of fit and relative error distribution statistics as evaluation indexes, and screening out an optimal model; and S5, applying the conventional water quality parameters in the breakpoint interval of the chlorophyll monitoring data in the water body to the optimal model, and outputting the restored chlorophyll concentration value to complete the dynamic restoration of the breakpoint.

The invention discloses an impedance-invariant vector synthesis phase shifter, an array element and a phased array, and relates to the technical field of phased array design. According to the impedance-invariant vector synthesis phase shifter provided by the invention, two capacitive compensation branches are arranged in each Gilbert unit of a variable gain amplifier, and the capacitive compensation branches are connected to a circuit only when bits are turned off, so that the output impedance of the unit is kept unchanged when the bits are turned on and turned off, and the output impedance of the unit is not changed when the bits are turned off. The phase shifting precision of the phase shifter is effectively improved, the root mean square phase error of the phase shifter is reduced to be within the acceptable range of the engineering field, phase shifting is directly carried out according to the most ideal condition based on IQ orthogonal signals, complicated calibration work is avoided, and the calibration problem of the vector synthesis phase shifter is solved. The invention further provides an array element which adopts the impedance-invariant vector synthesis phase shifter, the continuous time linear equalizer and the current mode attenuator and has the advantage of being convenient to deploy.

The invention discloses a lake water quality prediction method and system based on a hybrid neural network, and a computer readable storage medium, and belongs to the field of environmental science engineering and deep learning. The method comprises the following steps: screening original water quality data, removing abnormal values, performing linear interpolation, dividing a training set and a test set, decomposing a sequence by using VMD, optimizing VMD parameters by using PSO, reconstructing a new sequence with noise removed, and finally performing prediction by using LSTM-KAN. Through verification of total phosphorus concentration data of four sections of the Dian Lake, comparison with LSTM, VMD-LSTM, VMD-LSTM-KAN and LSTM-KAN models is carried out, and a correlation coefficient (), a mean absolute error (MAE) and a root-mean-square error (RMSE) are selected to evaluate precision. The result shows that the PVLK model has the best performance in single-step and multi-step prediction, the total phosphorus concentration prediction of each section can be kept at 0.75 in 10-step prediction with the step length of 4 hours, the applicability to time sequence data containing abnormal values and high sampling frequency is good, and efficient prediction of lake water quality is effectively promoted.

The invention discloses a grasp point pose calculation method based on a geometric clustering algorithm without deep learning and a GPU (Graphics Processing Unit) and application of the grasp point pose calculation method. The grabbing point pose calculation method comprises the steps that 100, three-dimensional point cloud data of a to-be-grabbed object are obtained through an industrial-grade structured light three-dimensional camera, a geometric clustering algorithm is adopted for screening the three-dimensional point cloud data, and a unique target point cloud cluster conforming to the target form and size is extracted; step 200, performing rough matching on the target point cloud cluster and the template point cloud through the FPFH feature vector and RANSAC to obtain a rough matching result; step 300, performing fine matching on the target point cloud cluster and the template point cloud through the rough matching result and the GICP to obtain a fine registration result; step 400, judging whether a precise registration result meets a preset requirement or not according to the interior point root-mean-square error and the overlap ratio, if so, entering the next step, and otherwise, terminating the process; and 500, according to the fine registration result and coordinate transformation, the grabbing point pose of the to-be-grabbed object is obtained through calculation.

The invention relates to a method for detecting the INDF content in TMR of lactating cattle based on a near infrared spectrum. The method comprises the following steps: constructing a near infrared spectrum model capable of being used for detecting the INDF content; wherein a sample set for constructing the near infrared spectrum model comprises near infrared spectrum data corresponding to a lactating cattle TMR sample and an INDF actual value; the method comprises the following steps: extracting a wavelength point with a non-zero coefficient in near infrared spectrum data, and taking the wavelength point as a characteristic variable; constructing a near infrared spectrum model based on the characteristic variables and the corresponding INDF actual values by adopting a partial least squares regression algorithm; model parameters are optimized through a cross validation method, an optimal model parameter combination is determined by combining a cross validation root mean square error RMSECV and an external validation root mean square error RMSEP, and a near infrared spectrum model capable of being used for INDF content detection is obtained; and inputting near infrared spectrum data of a to-be-detected lactating cattle TMR sample into the near infrared spectrum model to obtain an INDF content detection result.

The invention provides a flood inundation situation intelligent perception and lightweight analysis model construction method, and relates to the technical field of hydrological monitoring and flood control and disaster mitigation. Unmanned aerial vehicle LiDAR point cloud data and satellite multispectral image data are acquired, a differential manifold topographic representation model is constructed, the earth surface is regarded as a Riemannian manifold, and complex topographic features are accurately described; a lightweight submerging calculation model is constructed based on a manifold diffusion theory, and a Laplace-Beltrami operator and a multi-scale solving strategy are adopted to realize rapid and accurate prediction of a flood submerging range and water depth; through a multi-terminal early warning information pushing mechanism, full-process intelligent support is provided, the simulation time is shortened to be within 2 hours from traditional 24 hours, the submerging range prediction goodness of fit reaches 89.7%, the water depth prediction root-mean-square error is controlled to be 0.35 m, and the flow velocity prediction error in a gradient sudden change area is reduced by 12%.

The invention relates to the field of ecological hydrology, and discloses an arid and semi-arid region plant ecological water consumption calculation method, which comprises the following steps: acquiring meteorological, soil and vegetation phenological data; constructing a root system depth dynamic evolution model, and updating the maximum depth of the root system day by day; determining the depth weight of the effective moisture extraction layer based on the dynamic root system distribution; calculating the weighted average effective soil water content; and outputting daily-scale ecological water consumption in combination with the corrected transpiration scale model. By coupling meteorological driving, soil moisture stress and a dynamic regulation and control mechanism of a plant growth stage on a root system, the water consumption inversion precision is remarkably improved, the root-mean-square error is reduced by more than 25% through actual measurement verification, and a high-precision quantification tool is provided for ecological water demand evaluation and water resource management.

The invention relates to a fractional order lithium battery model parameter identification method based on an improved chaos evolutionary algorithm, and belongs to the technical field of lithium battery model parameter identification. The method comprises the following steps: representing a lithium battery second-order fractional order equivalent circuit model parameter identification problem as a parameter optimization problem; wherein the parameter optimization problem comprises known parameters, to-be-identified parameters and a target function, the known parameters comprise a function relationship between the OCV and the SOC and equivalent internal resistance, and the target function is the sum of root-mean-square errors of the estimated terminal voltage and the actually measured voltage; an improved Latin hypercube sampling strategy is introduced, Givens transformation and Householder transformation are introduced, and a restart strategy based on reverse learning and element rearrangement is introduced to improve an original chaotic evolutionary algorithm; and performing parameter optimization on the parameter optimization problem based on the improved chaos evolutionary algorithm to obtain an optimal parameter combination, thereby completing parameter identification of the to-be-identified parameter. The method aims at solving the technical problems that in the prior art, parameter identification precision is poor, and local optimum is prone to occurring.

The present application belongs to the technical field of aircraft structure fatigue strength, and particularly relates to a method for converting a buffeting frequency spectrum of an aircraft structure into a fatigue load spectrum, comprising the following steps: Step 1: processing buffeting response load data of the aircraft structure to obtain a first-order bending modal characteristic frequency f1; Step 2: performing narrowband filtering on the buffeting response load data, and calculating a root mean square value σ of the filtered narrowband buffeting response load data; Step 3: performing rainflow counting on the filtered narrowband buffeting response load data to count the amplitude distribution of the narrowband buffeting response load data; fitting the amplitude distribution of the narrowband buffeting response load data with a Rayleigh distribution; and equivalently distributing the amplitude distribution fitted with the Rayleigh distribution to three typical load amplitudes; and Step 4: establishing a periodic step load spectrum according to the frequency of the three equivalent typical load amplitudes.

A glass article and associated brush polishing method are described. The glass article includes a polished edge extending between a first major surface and a second major surface. The polished edge exhibits at least one of: (a) an Ra surface roughness greater than or equal to 1 nm and less than or equal to 20 nm; (b) a root mean square surface roughness greater than or equal to 1 nm and less than or equal to 30 nm; and (c) a peak valley surface roughness greater than or equal to 10 nm and less than or equal to 50 nm. The glass article is strengthened prior to brush polishing such that a majority of the polished edge is not under compressive stress. The brush polishing process is performed such that the first compressive stress layer extends to the polished edge.

This invention relates to a method for measuring the dielectric constant of fibers. The method includes the following steps: First, composite material and pure matrix samples are prepared, and the reflection loss curve of the composite material and the complex dielectric constant of the matrix are measured in the 1-18 GHz frequency band. Second, a realistic three-dimensional microstructure model of the composite material is reconstructed using micro-CT scanning. Then, an electromagnetic simulation model is established based on this model, and an initial complex dielectric constant is assigned to the fiber for simulation to obtain the simulated S11 parameter. Finally, by iteratively adjusting the complex dielectric constant of the fiber, the root mean square error between the simulated S11 curve and the measured reflection loss curve is minimized, thereby retrieving the actual complex dielectric constant of the fiber. This invention avoids theoretical model errors, directly correlates the macroscopic properties and microscopic parameters of materials, and has the advantages of high accuracy and wide applicability, providing a reliable parameter acquisition method for the design and optimization of the electromagnetic properties of composite materials.

The invention discloses a lithium ion battery health state prediction system and method based on multi-mode sensor fusion, and belongs to the technical field of lithium ion battery health management and intelligent monitoring crossing. Comprising a multi-modal data acquisition module, a data preprocessing module, a multi-modal feature extraction module, a physical constraint modeling module and a multi-time-scale hybrid neural network module, data dimension shortages are supplemented, the prediction precision and robustness are improved, voltage, current, temperature and strain data are synchronously acquired through a multi-modal sensor, and the prediction accuracy and robustness are improved. The defect that the traditional technology lacks mechanical dimension information is overcome, and the SOH prediction root-mean-square error is reduced to be within 1% by combining multi-modal feature extraction and multi-time scale modeling, which is far better than that of the traditional method; meanwhile, the multi-modal fusion design ensures that when a single sensor fails, the prediction function can be maintained through other modal data, the system robustness is outstanding, physical mechanism constraints are fused, and generalization and interpretability are enhanced.

The invention relates to a Beidou No.3 precise point positioning method based on double-frequency combination and analysis center product optimization, and belongs to the technical field of precise point positioning, and the method comprises the following steps: S1, constructing three ionosphere-free combinations B1C-B2a, B1C-B3I and B1I-B3I by using four frequency point signals B1C, B2a, B1I and B3I; s2, through comparing a root mean square value RMS of the positioning deviation, the final positioning precision and the convergence time, respectively comparing and evaluating precision point positioning PPP performance of BDS-3 in different frequency combinations under precision products issued by a plurality of analysis centers; and S3, finding out the ionosphere-free combination and analysis center product with the optimal PPP calculation performance.

The application provides a doubly-fed fan parameter data identification method and system, which comprises the following steps: obtaining measured data of a doubly-fed fan by setting faults with different voltage amplitudes; constructing a doubly-fed fan model according to a doubly-fed fan control system, wherein the control parameters in the model are identification targets; performing rough estimation on the control parameters by using a recursive least square algorithm, and updating the control parameters by using random training samples in the measured data one by one, and stopping the updating when the control parameter adjustment amount is lower than a first threshold value, and then switching to a genetic algorithm; performing fine estimation on the control parameters by using the genetic algorithm, verifying the accuracy of the control parameters by using samples in a verification sample library in the measured data after each iteration, and stopping the iteration when the root mean square error between the output of the doubly-fed fan model and the verification sample is lower than a second threshold value, and taking the corresponding control parameters as optimal results, and taking the optimal results as the estimated values of the fan parameters.

The invention relates to the technical field of biomedical signal processing and medical electronics, in particular to a blood pressure measuring method for non-invasive exercise blood pressure monitoring. The method aims to solve the technical problem of low blood pressure measurement accuracy caused by broadband noise interference and failure of a fixed parameter model in a motion state in the prior art. According to the technical scheme, the method comprises the steps that a cardiac cycle reference signal and a Korotkoff sound signal are synchronously collected, a sampling window is calculated in a self-adaptive mode by utilizing a cycle triggering feature point gating mechanism and combining a heart rate, and time domain precise interception of the signals is achieved; carrying out root mean square calculation and Savitzky-Golay filtering on the intercepted signal, and keeping waveform characteristics while denoising; constructing a multi-dimensional feature regression model containing a peak value, skewness, kurtosis and dynamic template similarity so as to accurately judge systolic pressure and diastolic pressure; an automatic recharging mechanism is triggered by monitoring signal quality, and strong interference is physically avoided. The method has the advantages of being high in anti-noise capacity, high in feature extraction robustness and accurate and reliable in measurement result.

A neural processing unit for performing inference operations of a large-scale language model based on an artificial neural network is disclosed. The neural processing unit according to the present disclosure includes a processing element core configured to perform an attention mechanism-based operation based on input data in vector format to output an operation result, a special function unit comprising a plurality of arithmetic circuits including at least one vector-dedicated arithmetic circuit that exclusively performs vector operations and at least one mixed arithmetic circuit capable of performing both vector and scalar operations, and configured to perform a special function operation on the operation result, and a controller configured to, upon receiving an RMS normalization operation execution command, activate at least one of the plurality of arithmetic circuits to control the special function unit to perform an operation of converting at least one of the operation result or the input data into a normalized vector whose magnitude is adjusted based on a root mean square (RMS), wherein the operation result may include an attention score for the input data.

An optical device is described. At least a portion of the optical device includes lithium niobate and is fabricated utilizing ultraviolet lithography. In some aspects the at least the portion of the optical device is fabricated using deep ultraviolet lithography. In some aspects, the short range root mean square surface roughness of a sidewall of the at least the portion of the optical device is less than ten nanometers. In some aspects, the at least the portion of the optical device has a loss of not more than 2 dB / cm.