20 results about "Linear approximation" patented technology

This specification pertains to the field of refining and chemical production processing technology, and particularly relates to a method and apparatus for optimizing refining and chemical production plans for a target secondary processing unit. The method includes: obtaining parameters of the secondary processing unit under a refining and chemical production optimization scenario; constructing a nonlinear constraint equation for product output based on the parameters of the secondary processing unit; performing an equivalent transformation on the nonlinear constraint equation to obtain an equivalent transformation equation; extracting the nonlinear constraint terms and performing a Taylor expansion; transforming the Taylor expansion formula into a linear approximation equation based on the proportion of the feed material to the output material of the blending tank, the feed material quantity to the blending tank, and the feed property values ​​of the blending tank, according to the property balance constraints of the blending tank; and then substituting these into the nonlinear constraint equation for product output, thereby transforming the nonlinear constraint equation for product output into a linear approximation equation containing only material information variables. This method effectively ensures model convergence and shortens the solution time.

The invention relates to a computer-implemented method for creating a model for detecting an anomaly or an operating state or a change in an operating state in a technical system (1), comprising the following steps: - Providing (S1) time series of operational variables, each indicating a temporal progression of an operational variable, as training data; - Performing (S2) a piecewise linear approximation on the time series of the operating variables to obtain segments of concatenated linear functions for each of the operating variables, obtaining change points (T) as time points between the segments of all operating variables, and summarizing the change points (T) of all operating variables to define state segments between two temporally successive state segments that define time intervals of the time series of the operating variables; - Determining (S3) characteristics of the time-dependent behavior of the operating variables for each of the state segments in order to form a state vector for each (S4); - Training (S6) a classification model based on the state vectors; - Implement (S7) the classification model in a control unit of the technical system (1).

The application provides a trusted federated learning system and method based on a blockchain and zero-knowledge proof, comprising a disturbance module.In the disturbance module, a non-interactive zero-knowledge protocol is applied, so that the noise added to the local model can be verified. The trainer and the verifier independently and respectively use linear approximation to construct the same circuit containing specific Gaussian noise, and the trainer uses the circuit and a public generator and model parameters to construct a proof. The verifier uses their circuit to verify the received proof; a verification module. The verification module constructs a sharded blockchain consensus process. Each verifier first verifies the accuracy of the local model gradient update using a public data set. The verification result is used as a voting summary to form a consensus in the network. For block data containing local model selection information, a Byzantine fault tolerance protocol is introduced to ensure data credibility. The application can be proved to achieve differential privacy and can analyze the upper bound of the influence of noise on the model performance.

A control model generation device includes: an observation value acquisition unit to acquire a plurality of observation values that are outputs of a control target having non-linear characteristics; a state space model estimation unit to estimate a state space model that expresses a linear approximate curve related to the plurality of observation values acquired; and an upper bound model estimation unit to calculate an estimation error, and estimate an upper bound model that expresses an upper bound of the estimation error, the estimation error being an error between each of the observation values acquired and the linear approximate curve expressed by the state space model estimated. Furthermore, the control model generation device includes a control model generation unit to generate a control model that expresses an equation of motion of the control target using the state space model estimated and the upper bound model estimated.

The present application relates to a new power system technology, in particular to a user load curve similarity measurement method based on piecewise linear approximation; by piecewise linear approximation to the user daily load curve electricity consumption data, the piecewise breakpoints of each two users are sorted in order, and the daily load curve is re-segmented; the slope, electricity consumption average and end time of each segment after re-segmentation are re-characterized; the difference between the slope and the total electricity consumption in different segments is calculated to evaluate the difference between the shape and the value of the daily load curve of different users; the weighted average of the shape difference and the value difference is calculated to calculate the total difference of the daily load curve; the total difference matrix of all users is constructed to calculate the daily load curve similarity between any two users; the present application overcomes the sensitivity of the existing user load curve similarity measurement method to extreme value and noise.

The application is suitable for the technical field of steam trap pump, in particular to a double-floating-ball steam trap pump and a method for determining the volume and buoyancy parameters of the floating ball, the method accurately captures the nonlinear mapping of the buoyancy requirement and working condition from the test data through multiple rounds of iterative feature extraction, reduces the reference buoyancy prediction error; the accurate cubic function formula of the spherical floating ball immersion volume is used to eliminate the huge geometric error caused by linear approximation; the total buoyancy requirement is simultaneously transmitted to the individual floating ball margin term, the total drainage constraint and the Lagrange optimization target to form a complete quantitative design link, and the optimal allocation of the double floating ball volume is realized through the analytical solution; a differentiated attenuation model is established for floating ball A and floating ball B respectively to compensate for the effects of corrosion and erosion in long-term operation. The above technical means can solve the problems of large volume and buoyancy parameter error and inadaptation to the double-floating-ball steam trap pump obtained by the existing method.

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, an input tensor is accessed as input to a softmax operation of a machine learning model. A first intermediate tensor is generated based on the input tensor using a non-uniform piecewise linear approximation (PWLA) of an exponent operation, and a second intermediate tensor is generated based on the first intermediate tensor using a normalization operation. A third intermediate tensor is generated based on the second intermediate tensor using an inverse operation. An output tensor is generated as output of the softmax operation based on the third intermediate tensor using a bitwise shift operation, and an output of the machine learning model is generated based on the output tensor.

The application discloses a two-dimensional galvanometer visual axis pointing coupling correction method, and mainly solves the problem that the visual axis pointing error calculated by the prior art is in the order of milliradians, and it is difficult to meet the high-precision correction requirement.The method comprises the following steps: step 1, establishing a visual axis pointing model for a two-dimensional galvanometer to be corrected; step 2, acquiring a linear model between the rotation angle of the two-dimensional galvanometer and the visual axis pointing angle of a two-stage stable pointing system based on the visual axis pointing model by using a linear approximation method; acquiring a pointing error based on the visual axis pointing model and the linear model; performing nonlinear correction on the pointing error by using a Volterra series to acquire a nonlinear correction relationship; step 3, acquiring a corrected pointing error by using the nonlinear correction relationship; and step 4, acquiring the control angle of the two-dimensional galvanometer based on the corrected pointing error.The application can reduce the visual axis pointing error from the order of milliradians to the order of microradians, the pointing calculation precision is improved by about 75 times, and the visual axis pointing calculation precision is significantly improved.

This invention belongs to the field of bridge engineering technology, and specifically relates to a multi-objective intelligent decision-making method for cable-stayed bridge cable forces based on equilibrium degree. This invention includes a method for constructing a small-rate-increase influence matrix for cable-stayed bridges and a multi-objective intelligent decision-making algorithm for cable forces based on a high-low equilibrium degree strategy. The method for constructing the small-rate-increase influence matrix adjusts the cable forces at the same small rate of increase based on the initial cable forces, and normalizes the resulting effect increments to form a small-rate-increase influence matrix, resulting in a smaller linear approximation error when calculating the structural response of cable-stayed bridges. The multi-objective intelligent decision-making algorithm for cable forces based on a high-low equilibrium degree strategy is a strategy of multi-objective particle swarm optimization that simultaneously retains high-equilibrium-degree and low-equilibrium-degree solutions during the iterative process to guide the optimization direction. It can efficiently and reliably select Pareto non-dominated solutions that achieve the preset standard of optimization degree on multiple objectives, thus improving decision-making efficiency.

The application provides a kind of fixed area beam forming physical layer security transmission system and method, it is related to wireless communication physical layer security field, it aims at solving the technical defects that traditional directional angle beam forming and rotating angle beam forming are easily intercepted signal by intelligent eavesdropping equipment due to existing fixed main lobe transmission track.The system includes baseband modulation module, frequency adjustment module and phased array module, by dividing phased array into multiple multi-frequency phased subarrays, transmitting multiple rotating sub-beams, effectively superimposing to form beam main lobe in the target receiving position small area, mutually canceling to form side lobe outside the target area;Combining frequency offset increment probability switching realizes side lobe randomization, for unknown eavesdropping position, crowd search algorithm is used to optimize frequency offset increment, for known eavesdropping position, block coordinate sinking linear approximation algorithm is used to optimize frequency offset increment.Eliminate the long distance transmission track of traditional beam main lobe, significantly improve the physical layer security performance of wireless communication system.

The application discloses an intelligent complex natural gas pipeline network pipeline application verification method under an open transportation mode, which comprises the following steps: obtaining pipeline network and pipeline application basic data, and establishing a topological structure; introducing node, pipeline and booster station physical characteristics and boundary constraints, and constructing a pipeline application verification non-convex MINLP model; solving the model by using a two-stage algorithm, judging the application technical feasibility according to the model, and outputting a scheduling scheme or an infeasible explanation. The application comprehensively considers factors such as pipeline hydrodynamics, compressor station discrete state switching and non-convex feasible region boundary, innovatively proposes a two-stage solving algorithm, combines one-dimensional nonlinear function piecewise linear approximation, high-dimensional nonlinear function space grid approximation and compressor non-convex feasible region convex relaxation technology, effectively solves large-scale non-convex mixed integer problem solving difficulties and low calculation efficiency and other problems, and provides theoretical support for intelligent operation scheduling under the open transportation mode of the natural gas pipeline network.

The invention relates to a computer-implemented method for creating a model to identify an anomaly, an operating state or an operating state change in a technical system, comprising the following steps: - providing time series of operating variables as training data, the time series each illustrating a course of change over time of an operating variable; - performing piecewise linear approximation on the time series of operating variables to obtain a piecewise arrangement of linear functions for each operating variable, wherein change points are obtained as time points between the pieces for all operating variables, wherein the change points for all operating variables are pooled to each define a state piece between two change points defining time segments of the time series of operating variables in time succession; - determining features of the course of change over time of the operating variables for each state piece to each form a state vector; - training a classification model based on the state vectors; - implementing the classification model in a control device of the technical system.

The application discloses a kind of magnetic suspension device control method of full habitat unmanned aerial vehicle, it is related to unmanned aerial vehicle technical field.The method includes: based on magnetic circuit Kirchhoff law constructs equivalent magnetic circuit and solves main air gap magnetic flux, calculates electromagnetic suspension force using virtual displacement principle, and torque is solved by geometric analysis and cross multiplication operation.Suspension gap is deduced using Hall sensor, and equation set is constructed to solve floater space position and total suspension force.Through establishing floater dynamics equation, control voltage is generated using virtual magnetic flux cascade control structure, and temperature and inductance change are corrected in real time.Taylor expansion is used to linearly approximate nonlinear electromagnetic force, PID parameters are dynamically adjusted, and finally PWM signal is generated to drive coil.The application is contactless suspension and drive, discards traditional mechanical slip ring and connecting rod, significantly reduces weight, improves energy efficiency and response speed, and has excellent waterproof and dustproof ability.

This invention discloses a method, system, and storage medium for optimal differential and linear feature search in SPN-type cryptography. The invention constructs a lightweight Super S-box differential distribution table (DDTSS) and a linear approximation table (LATSS), respectively. Then, the DDTSS and LATSS are sorted by upper bounds of the differential and linear probabilities, respectively. Following an improved branch-and-bound algorithm, optimal differential and linear features are searched on the DDTSS and LATSS under given pruning conditions. This invention uses stricter pruning conditions to optimize the branch-and-bound search strategy, enabling the acquisition of precise security bounds against differential and linear analysis in SPN-type lightweight cryptography within a shorter time. This invention can be used for security evaluation of SPN-type lightweight cryptography and to assist in the design of lightweight cryptographic algorithms, possessing high practical value and application prospects.

A control model generation device configured to include: an observation value acquisition unit (1) for acquiring a plurality of observation values ​​that are outputs of a control target having nonlinear properties; a state-space model estimating unit (2) for estimating a state-space model expressing a linear approximation curve with respect to the plurality of observation values ​​acquired by the observation value acquisition unit (1); and an upper bound model estimating unit (3) for calculating a determination error and estimating an upper bound model expressing an upper bound of the determination error, wherein the determination error is an error between each of the observation values ​​acquired by the observation value acquisition unit (1) and the linear approximation curve expressed by the state-space model estimated by the state-space model estimating unit (2).Furthermore, the control model generation device includes a control model generation unit (4) for generating a control model that expresses an equation of motion of the controlled target using the state space model estimated by the state space model estimating unit (2) and the upper bound model estimated by the upper bound model estimating unit (3).

This disclosure provides a computer system and method based on an axial strain field reconstruction model, including: an acquisition module, a preprocessing module, a separation module, a processing module, and an output module; the preprocessing module interacts with the acquisition module to acquire the time-series phase signal in the interference spectrum; the separation module constructs a complex phase signal using the phase difference signal and analyzes the complex phase signal using a two-dimensional Gaussian window function to separate the effective phase difference signal and phase noise; the processing module separates the effective phase difference signal within a preset local analysis window region. y - z The surface is linearly approximated to construct a linear mapping model between the instantaneous frequency of the maximum spectral energy and the phase gradient. The output module is used to obtain the axial strain inside the material based on the relationship between the instantaneous frequency of the maximum spectral energy and the phase gradient. Accurate strain reconstruction under high noise conditions can be achieved without numerical differentiation, which has broad application prospects.

The application discloses a method and device for solving a crude oil scheduling model. The method comprises the following steps: constructing an original mixed integer nonlinear programming (MINLP) model of original crude oil scheduling, and identifying key nonlinear terms in the original MINLP model; the original MINLP model comprises continuous variables, binary variables and key nonlinear terms; the key nonlinear terms comprise nonlinear mapping of physical properties of crude oil components after mixing and proportions of each component; according to the identified key nonlinear terms, the key nonlinear terms are designed for piecewise linear approximation; according to the design result of the piecewise linear approximation, an approximate MINLP model is constructed, and a solution of the approximate MINLP model is used as an initial solution of the original MINLP model; and according to the initial solution, the original MINLP model is solved to obtain a final solution of the original MINLP model. The method and device significantly reduce the nonlinear calculation burden, effectively improve the solving speed, and realize a large proportion of solving time shortening on the premise of ensuring feasibility and solution quality.

This invention belongs to the field of bridge engineering technology, and specifically relates to a multi-objective intelligent decision-making method for cable-stayed bridge cable forces based on equilibrium degree. This invention includes a method for constructing a small-rate-increase influence matrix for cable-stayed bridges and a multi-objective intelligent decision-making algorithm for cable forces based on a high-low equilibrium degree strategy. The method for constructing the small-rate-increase influence matrix adjusts the cable forces at the same small rate of increase based on the initial cable forces, and normalizes the resulting effect increments to form a small-rate-increase influence matrix, resulting in a smaller linear approximation error when calculating the structural response of cable-stayed bridges. The multi-objective intelligent decision-making algorithm for cable forces based on a high-low equilibrium degree strategy is a strategy of multi-objective particle swarm optimization that simultaneously retains high-equilibrium-degree and low-equilibrium-degree solutions during the iterative process to guide the optimization direction. It can efficiently and reliably select Pareto non-dominated solutions that achieve the preset standard of optimization degree on multiple objectives, thus improving decision-making efficiency.

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, an input tensor is accessed as input to a softmax operation of a machine learning model. A first intermediate tensor is generated based on the input tensor using a non-uniform piecewise linear approximation (PWLA) of an exponent operation, and a second intermediate tensor is generated based on the first intermediate tensor using a normalization operation. A third intermediate tensor is generated based on the second intermediate tensor using an inverse operation. An output tensor is generated as output of the softmax operation based on the third intermediate tensor using a bitwise shift operation, and an output of the machine learning model is generated based on the output tensor.

A method and system for dual target learning index in a key-value store is disclosed. The key-value store is a log-structured merge tree based storage. A sorted string table is divided into a plurality of data blocks, wherein a sum of values of keys in each data block is less than or equal to a maximum block size value. Piecewise linear approximation functions are generated for the plurality of data blocks, wherein a maximum interpolation error of the piecewise linear approximation functions is less than or equal to a model error value. The maximum block size value and the model error value are provided by a reinforcement learning agent. Upon receiving a lookup query request to find a target key, the system initiates a discovery operation to discover a data block from the plurality of data blocks and generates a search range of the discovered data block.