16 results about "Linear variation" patented technology

The invention discloses a frequency response test method and system for an optical fiber F-P cavity filter. The method comprises the following steps: S1, executing the following operations through a signal generator: S11, enabling an approximate linear change region of a transmission peak of the filter to be located near the working wavelength of a narrow linewidth laser; s12, a driving signal x (t) changing periodically is added to the filter, and the center of an approximate linear change area of a rising edge / falling edge of a transmission peak of the filter is made to be close to the working wavelength of the narrow-linewidth laser; s13, enabling the working wavelength of the narrow linewidth laser to be always located in an approximate linear change region of the rising edge / falling edge of the transmission peak; and S2, when the driving mechanism of the filter works at the driving frequency f of the driving signal x (t), detecting the narrow linewidth laser passing through the filter by adopting a photoelectric detector PD, and converting the narrow linewidth laser into a response electric signal y (t). According to the invention, the frequency response of the optical fiber F-P cavity filter can be effectively tested.

A kind of contact force measurement method of load launch process, comprising: step one: install load ring assembly on the mouth piece of launch tube through shear sensor;Step two: adjust the concentricity of load ring assembly and launch tube, adjust the levelness of shear sensor;Step three: set stress measuring element on shear sensor;Step four: hang standard weight on load ring assembly, fit linear variation curve between the signal of stress measuring element and the total mass of standard weight;Step five: place launch tube in water and launch load, obtain the contact force corresponding to load ring assembly by linear variation curve between the signal data of stress measuring element and the total mass of standard weight, accurately measure the contact force between the mouth end of launch tube and load during launch process, provide test data support for the establishment of contact force mathematical model of carrier platform and load separation process, improve the accuracy and reliability of load launch trajectory design.

The embodiment of the invention provides an asset correlation determination method and a related device, and the method comprises the steps: determining a joint feature function between a variance factor and the logarithmic prices of two types of assets based on the variance factor and the logarithmic prices of the two types of assets through responding to the input of observation data; the variance factor obeys an overall equilibrium model of the interest rate; converting the joint feature function into a conditional density function about the observation data through Fourier inversion; updating a preset likelihood value based on the conditional density function to obtain an updated likelihood value; and determining the asset correlation based on the updated likelihood value. According to the embodiment of the invention, the dynamic relationship is recursive based on the data model, the method does not depend on any statistical approximation and linear hypothesis, is more in line with the actual market nonlinear change process, and improves the accuracy of asset correlation.

The invention relates to the technical field of oil exploration and development, and discloses a stratum collapse pressure earthquake prediction method and device based on rock mechanics nonlinear variation, and the method comprises the steps: carrying out the nonlinear fitting of compressive strength with rock cohesion and an internal friction angle through a triaxial mechanics experiment, and obtaining a corresponding fitting relation; then, through seismic inversion, a rock mechanical attribute body, an overlying stratum pressure body, a pore pressure body, a maximum horizontal principal stress body, a minimum horizontal principal stress body, and three-dimensional space rock cohesion distribution and internal friction angle distribution are obtained through calculation; and finally, according to the pore pressure body, the maximum horizontal principal stress body, the minimum horizontal principal stress body, the three-dimensional space rock cohesion distribution and the three-dimensional space rock internal friction angle distribution, predicting stratum three-dimensional collapse pressure body data in combination with a collapse pressure calculation model. By means of the method, the formation collapse pressure can be predicted more accurately, geological disasters caused by improper drilling operation can be reduced, and therefore the geological environment is protected, and potential influences on the earth surface and underground water resources are reduced.

This invention relates to the fields of wireless sensing and deep learning, specifically to a human pose estimation method based on a multimodal attention network. In video modality learning, a multi-resolution network (FCN) is designed to extract pose features. By applying a local self-attention network to the generated heatmap, the weights and attention levels of different regions are adaptively adjusted to capture more refined keypoint information. In CSI modality learning, a spatiotemporal attention network and a multimodal-guided linear spatial position variation layer are designed to facilitate self-learning of its spatiotemporal features. A teacher-student architecture is adopted in the overall multimodal learning network to allow the Wi-Fi signal-based pose estimation model to learn more from the deep learning capabilities of the video modality. By transferring the correct human pose estimation information from the video learning network to the Wi-Fi signal learning network, accurate human pose information can be estimated using only the CSI modality as input.

This invention discloses an LSTM-based thermal error prediction method for lead screws based on an attention mechanism, relating to the field of lead screw technology. The method includes: collecting multi-source data such as temperature field, thermal error, and time-series operating conditions during lead screw operation; preprocessing the collected multi-source data and filtering key thermal features; constructing an improved LSTM basic prediction model embedded with a dual attention mechanism; fusing the improved LSTM model with a lightweight tree model at the decision layer; inputting the preprocessed and filtered key thermal features into the fusion model; and outputting the lead screw thermal error prediction result. This invention achieves high-precision time-series prediction of lead screw thermal errors, solving the problems of insufficient mining of the correlation between thermal features and thermal errors, incomplete capture of time-series features, resulting in prediction results being greatly affected by irrelevant factors, insufficient accuracy, and difficulty in adapting to the nonlinear variation law of lead screw thermal errors, thus failing to meet the operating requirements of high-precision equipment.

The present application relates to a stranded air-blown micro cable, comprising an outer sheath (3), wherein the outer sheath (3) is internally provided with a plurality of optical units (2) stranded around a strength member (1) as the center; spirally distributed grooves (4) of designed dimensions are provided in a surface of the outer sheath (3); a plurality of grooves (4) are provided, and are evenly distributed on the circumference of the cross-section of the stranded air-blown micro cable; and one portion of the grooves (4) and another portion of the grooves (4) are symmetrically arranged using the diameter of the strength member (1) as the axis of symmetry. The stranded air-blown micro cable has a linearity factor β ≤0.02, and a dynamic friction coefficient less than 0.3. The overall unit density ρ of an optical cable is in the range of 0.0005-0.0015 g / mm3. By means of the above structure, the grooves (4) are symmetrically and uniformly arranged, such that an airflow generated by air blowing uniformly acts on the outside of the outer sheath (3) so as to increase an air blowing distance; moreover, since the linearity factor and the dynamic friction coefficient are less than design values, the combined effect of these two factors reduces linear variations during an air blowing process and improves an air blowing force. In addition, the dimensions of the grooves (4) and the overall unit density of the optical cable are specifically limited, thereby ultimately increasing an air blowing distance in routes having many bends.

This invention discloses a method for predicting and attributing the spatiotemporal variation of groundwater storage based on a Transformer-LSTM hybrid network. The method includes collecting and analyzing groundwater storage variation data and various potential factors influencing these variations in a target area; constructing a Transformer-LSTM hybrid deep learning model using high-dimensional driving factors as input and the spatiotemporal variation of groundwater storage as output; training and validating the Transformer-LSTM model; deeply interpreting the temporal evolution and spatial distribution characteristics of groundwater storage based on the prediction results of the Transformer-LSTM model; and quantifying the influence of potential driving factors on groundwater storage variation using the interpretable SHAP method to identify key driving elements. This method can effectively capture the strong spatiotemporal heterogeneity and complex nonlinear variation characteristics of groundwater storage at a large regional scale.

The invention discloses a method and a system for measuring a rolling angle of a magnetic suspension model based on a trapezoidal marker line. The system comprises a pose measurement sensor and a trapezoidal marker line. The trapezoidal marker line is a closed trapezoidal ring which is adhered to the mass center of the model in a surrounding manner, the width value, parallel to the axis of the model, of the trapezoidal marker line is monotonically and linearly changed, and the trapezoidal marker line is a trapezoid with unequal bottom edges in a geometric unfolding state. During measurement, the linear array CCD scans the surface of the suspension model and detects pixel serial numbers corresponding to two edges imaged by the trapezoidal marker line; according to the difference value of the pixel serial numbers corresponding to the two edges, the roll angle of the suspension model can be solved in combination with a preset geometrical relationship; the axial position of the suspension model can be calculated through the mean value of the two edge pixel serial numbers. On the premise that a hardware system is not changed, integrated measurement of the rolling posture and the axial position is achieved only by changing the marker line configuration and the resolving method, and the method has the advantages of being good in compatibility, low in cost and high in measurement real-time performance.

One aspect of the invention relates to a method for compensating the rate as a function of temperature of a watch (1) in which, in an internal volume V occupied by n moles of a gas with constant R, a waterproof case (2) contains a movement (3) with an oscillator (4), in which the pressure coefficient Cp and the humidity coefficient Ch of the movement (3) are determined, the optimal value Cto of the thermal coefficient Ct of the oscillator (4) is calculated so as to define its rate as a function of the relative linear variation of the temperature T in order to compensate for pressure and humidity deviations, and, for after-sales, the watch (1) is equipped with a compensation device (10) to change the pressure P and / or the constant R and / or the amount of gas and / or the temperature T in the case (2), or, in the factory, by modifying the thermal coefficient of the elastic return means of the oscillator (4) by modifying the oxide layer and / or applying or ablating a coating, and / or modifying the amount and / or the nature of the gas in the watch, and / or modifying the internal volume of the case (2).

This invention discloses a method for testing the oil content of an automotive air conditioning compressor, relating to the field of air conditioning compressor testing technology. The method includes acquiring the original oil content monitoring data sequence of the compressor collected by an intelligent sensor within a test cycle; correcting the standard Kalman filter prediction step by incorporating the nonlinear characteristics of oil content state transitions; and dynamically denoising the original data sequence using an improved Kalman filter algorithm to obtain a regular and reliable denoised oil content data sequence. The denoised data sequence is then time-series segmented to separate the oil content quantum sequences for the startup phase and steady-state operation phase within the test cycle. The average oil content value for each phase is calculated, thereby generating the final oil content test result for the compressor. This method adapts to the nonlinear variation characteristics of oil content, weakens interference components in the monitoring data, and achieves segmented quantification of oil content states according to operating conditions, aligning with the testing and evaluation needs of actual automotive air conditioning compressor operating conditions.

This invention discloses an epitaxial method for InGaAs layered linear gradient variation buffer layers based on molecular beam epitaxy, comprising the following steps: S1: Determine the total variation range of the specific In composition and the total thickness of the linear variation of In, and calculate the retreat composition of the retreat layer based on the variation rate of the In composition; S2: Design the InGaAs linear gradient buffer layer in layers and determine the thickness of the entire layer; S3: Determine the value of the In composition after actual epitaxy, and adjust the In composition and thickness in step S2 according to the value after actual epitaxy; S4: Determine the growth rate of each gradient layer of InGaAs; S5: Epitaxially construct the InGaAs layered linear gradient variation buffer layer based on the layer thickness and the growth rate of each gradient layer of InGaAs. This invention realizes the InGaAs layered linear gradient variation buffer layer. x Ga 1‑x As linear gradient buffer layer, the linear change of In composition in As reduces the influence of source furnace temperature changes on the uncertainty of In and Ga growth rate changes.

The invention discloses a tree body-double rope vibration frequency calculation method, and belongs to the technical field of branch mechanical property analysis. The method comprises the following steps: constructing an ideal variable-cross-section circular-truncated-cone-shaped branch model, regarding a branch as a cantilever beam, and defining the linear change of the radius of the cross section of the branch along the length direction; establishing a variable cross-section deflection curve differential equation based on the ideal model; carrying out integral solution on the variable cross-section deflection curve differential equation, and deducing a bending elastic modulus formula of the variable cross-section circular truncated cone-shaped branch in combination with boundary constraint conditions; establishing a tree body-double rope model; and deducing a tree body-double rope vibration frequency formula based on the tree body-double rope model and a bending elastic modulus formula. The method provides a theoretical basis for design of vibration type fruit picking equipment, vibration frequency parameters can be optimized, tree body damage is reduced, and picking efficiency is improved.

The application discloses a CORS ionosphere modeling method and equipment based on linear variation characteristics, first, a plurality of GNSS stations in linear distribution are selected, observation data is preprocessed and STEC is calculated, and then STEC is projected into VTEC through a single layer model; then, a VTEC dataset along a target direction at the same time is constructed, IPP distance normalization is performed on the ionosphere piercing point, a VTEC reference value is calculated by calling an international reference ionosphere model as a fitting weight, a dynamic order determination strategy is adopted to complete polynomial fitting through a weighted least square method, and a regional ionosphere model is obtained. The VTEC corresponding to a virtual reference station is obtained through interpolation, ionosphere delay is obtained, and the ionosphere delay is further applied in virtual observation value calculation, and linear regional high-precision network RTK calculation is realized. The application fully utilizes the physical characteristics that the ionosphere has variation regularity in the longitude / latitude direction, relaxes the distance requirement between stations, significantly improves the accuracy of the regional ionosphere model, and further improves the observation accuracy of the linear regional network RTK.

The invention provides a large model training method and system based on differential low-rank adaptation, and the method comprises the steps: connecting a low-rank adaptation module in a linear change layer of a pre-trained large model in parallel, freezing the original parameters of the pre-trained large model, setting the parameters in the low-rank adaptation module as trainable parameters, and carrying out the training of the pre-trained large model. Performing forward propagation on the input small-batch samples to obtain a loss function, calculating a gradient vector of each sample in the small-batch samples about the low-rank adaptation module based on the loss function, and further performing a differential privacy protection mechanism based on the gradient vector to update parameters in the low-rank adaptation module to obtain an optimized low-rank adaptation module; and performing iterative optimization based on the optimized low-rank adaptation module until a preset training iteration condition is met to obtain a target large model. According to the method, the memory ability of the model to high-sensitivity information is reduced, the method has good attack resistance and privacy protection strength, the conflict between privacy protection and model performance is solved, and the calculation efficiency and the parameter efficiency are optimized.

A plurality of waveguide display substrates, each waveguide display substrate having a cylindrical portion having a diameter and a planar surface, a curved portion opposite the planar surface defining a nonlinear change in thickness across the substrate and having a maximum height D with respect to the cylindrical portion, and a wedge portion between the cylindrical portion and the curved portion defining a linear change in thickness across the substrate and having a maximum height W with respect to the cylindrical portion. A target maximum height Dt of the curved portion is 10−7 to 10−6 times the diameter, D is between about 70% and about 130% of Dt, and W is less than about 30% of Dt.