31results about How to "Avoid invalid calculations" patented technology

The invention discloses an iterative propulsion disturbance domain updating method for aircraft dynamic aerodynamic characteristic simulation. Aiming at the problem that a large amount of invalid calculation exists in an existing aircraft dynamic aerodynamic characteristic numerical simulation method, three kinds of dynamic calculation domains of unsteady, convection and viscosity are adopted, anda solving thought that only non-convergent unsteady units are solved and the viscosity effect is only considered in a local area is provided, wherein the unsteady dynamic calculation domain only comprises grid units which play a leading role in the unsteady effect at each physical moment; the convection dynamic calculation domain only comprises grid units which are disturbed in each iteration step and are not converged yet, the viscosity dynamic calculation domain only comprises grid units which play a leading role in the viscosity effect in the convection dynamic calculation domain, and thethree types of dynamic calculation domains can be updated in a self-adaptive mode in each iteration step. By avoiding invalid calculation in unsteady numerical simulation and reducing the number of iteration steps required for inner iteration to reach a convergence state, the calculation efficiency of aircraft dynamic aerodynamic characteristic numerical simulation can be remarkably improved.

The invention provides a solid rocket engine performance progressive matching design method. The solid rocket engine performance progressive matching design method comprises the steps of acquiring task objectives of total indexes of a solid rocket engine; determining a time interval division number according to the working time of a preset thrust demand in the task target; obtaining an initial training sample set of the preset thrust demand matching degree based on an initial balance pressure method; according to the initial training sample set, operating the combustion surface and internal ballistic trajectory calculation model in the initial time interval to obtain N thrust curves, and constructing a thrust approximation model according to the N thrust curves; carrying out inaccurate sampling to obtain alternative solutions; substituting the alternative solution into a combustion surface calculation and internal trajectory calculation model, and calculating the performance of the internal trajectory within simulation time to obtain an alternative thrust curve; when the simulation time is not less than the working time, adding the alternative solution as a new sample point into asample set to obtain an updated sample set; and when a preset convergence condition is reached, outputting the alternative thrust curve as an optimal solution.

The invention discloses an unmanned obstacle avoidance path planning method and system, and belongs to the technical field of artificial intelligence unmanned driving. The system comprises a driving vehicle map server and a path planning platform; in the driving process of a driving vehicle, obstacles in front of a lane are detected in real time through detection equipment, detected obstacle information is transmitted to the map server through communication layer equipment, obstacle avoidance notification information is sent to the communication layer equipment, and the communication layer equipment sends a path planning request to the path planning platform; the path planning platform responds to path planning request information to obtain obstacle map data and lane map data, judges whether the vehicle can avoid an obstacle or not, and calculates vehicle steering, an obstacle avoidance end point and an obstacle avoidance passing point; and a reference path is calculated and determined, local adjustment is performed on the reference path to obtain an adjusted obstacle avoidance path, and then safety detection is performed on the obstacle avoidance path to determine a final obstacle avoidance path. The system can improve the path planning efficiency, and guarantees the vehicle driving safety.

The invention discloses an overlapped grid disturbance domain updating method for aircraft multi-body separation simulation. The method comprises the steps that a wall distance algorithm is simplified by employing an implicit assembly method based on the attribute characteristics of a wall distance marking unit, so that the grid assembly efficiency is improved; secondly, under the framework of a disturbance domain updating method, numerical simulation of the overlapped grids is realized, so that only units in a non-convergent disturbed region are subjected to interpolation operation, and the calculation amount of the interpolation operation of the overlapped grids is reduced; and thirdly, a novel overlapping grid information exchange method is established, and the mode that information is transmitted based on grid deformation and encryption in a large flow gradient area and information is transmitted based on flow gradient interpolation in a small flow gradient area is achieved, so that the accuracy of the overlapping area is improved, and the isolated point problem is avoided. According to the method, the calculation efficiency of numerical simulation of the multi-body separation problem of the aircraft can be remarkably improved, and the problems that the numerical precision is lost in a large-flow gradient grid overlapping region, and isolated points exist are solved.

The invention discloses a partition disturbance domain updating calculation method for steady viscous compressible streaming of an aircraft, and belongs to the field of computational fluid mechanics.The method comprises the following steps: reading a file; initializing a flow field; establishing an initial convection dynamic domain and an initial viscosity dynamic domain; allocating a storage space; assigning a conservation amount of the boundary virtual grid according to the boundary condition type; estimating a residual error; performing time integration; increasing the convection dynamic domain; reducing convection and viscosity dynamic domains; reallocating the storage space; increasing a viscous dynamic domain; repeating the above steps until a convergence condition is reached; and outputting a result. According to the invention, the convection dynamic domain and the viscosity dynamic domain which can be adaptively increased / decreased are adopted; only the disturbed units which are not converged yet are solved, the viscous effect is only considered in part of the disturbed units, the calculated amount of a single iterative step and the total iterative step number can be effectively reduced, and therefore the convergence rate of solving the steady viscous compressible streaming of the aircraft is increased.

The invention discloses a multi-grid disturbance domain updating acceleration method for aircraft streaming numerical simulation. Aiming at the problem that a large number of invalid calculations exist when an existing multi-grid method simulates aircraft streaming, two dynamic computational domains of convection and viscosity are adopted on a fine grid and a coarse grid, thus solving the flow control equation in each iterative step in a convection dynamic calculation domain only containing non-convergent disturbed units; and the viscous effect is only considered in the viscous dynamic calculation domain only containing the viscous effect leading unit, so that the solving thought of only solving the non-convergent disturbed unit and only considering the viscous effect in the local area isrealized. By effectively reducing the calculation amount of a single iterative step and the number of iterative steps required for achieving a convergence state, the calculation efficiency of simulating the aircraft streaming by the multi-grid method can be remarkably improved.

An OpenMP parallel disturbance domain updating method for aircraft aerodynamic characteristic simulation comprises the steps of data reading, static region decomposition, dynamic computational domain establishment, dynamic region decomposition, storage space allocation, boundary condition processing, residual estimation, time integration, in-block increase of a convective dynamic domain, in-block reduction of a dynamic computational domain, in-block increase of a viscous dynamic domain, in-block reduction of the viscous dynamic domain, in-block increase of the dynamic computational domain, judging of whether the solution reaches a convergence condition, outputting of a result and the like. The problems that a large amount of invalid calculation exists when an existing parallel numerical simulation method simulates the aerodynamic characteristics of an aircraft, and the calculation efficiency may be lost when an existing disturbance area updating method carries out large-scale parallelization can be solved.

The invention discloses a video recommendation method and device, a server and a storage medium, belonging to the technical field of communication. The video recommendation method comprises the steps of receiving a video recommendation request sent by a terminal where a login account is located; in response to the video recommendation request, acquiring historical video information of the login account, wherein the historical video information is used for recording information of a historical video list which is displayed on the terminal and subjected to playing operation; according to operation features and video features of the historical videos, screening a target video out from the historical video list, wherein the operation features are used for representing operations executed by the login account on the historical videos; screening out candidate videos from a video library at least based on the video features of the target video; and acquiring a recommended video of the login account from the candidate videos. By adopting the video recommendation method and device, the server and the storage medium provided by the invention, the problem of low accuracy of existing video recommendation is at least solved.

The invention provides a method and a device for performing convolution calculation on an image by using a convolutional neural network. The image is input into a convolutional neural network; and thefirst convolutional layer outputs a first feature map, the first feature map comprises a first control channel, and the first convolutional layer is a middle convolutional layer of the convolutionalneural network. The method comprises: taking an area with a characteristic value smaller than a preset threshold value in a first control channel as a first control area, wherein the characteristic value of any position of the first control channel is used for representing the probability that the corresponding position of a first characteristic graph belongs to target characteristics; determininga redundant area of the first feature map according to the first control area; and enabling other regions except the redundant region in the first feature map to enter the next convolution layer of the first convolution layer for convolution operation so as to obtain a second feature map. In the mode, the redundant area does not perform the next layer of convolution operation, invalid calculationcan be prevented, the calculation amount of the neural network is reduced, the convolution calculation efficiency is improved, and the power consumption of convolution calculation is reduced.

The invention relates to a fast generation method of water flow streamline based on an unstructured triangulation network, comprising the following steps: 1) pretreatment of an unstructured triangulation network topology structure to construct a 'point-unit' topology relation table; 2) setting a starting point for searching that water flow line and finding out the unit number of the starting point; 3) adopting the recursive search mode of variable search step size + water wave type expansion to trace the streamline upstream and downstream of the starting point; 4) sorting the water flow line point according to the order of the searching end point of the upstream water flow line, the starting point and the searching end point of the downstream water flow line. The invention adopts the recursive searching mode of water wave type expansion when generating the water streamline, minimizes the times of judging the relationship between the coordinate points and the grid unit when searching the water streamline, and remarkably improves the searching efficiency of the water streamline. At the same time, the method of dynamic searching step size is adopted to reduce the searching step size automatically in the dense area of the grid, which ensures the precision of generating streamline in the dense area of the grid, and automatically increases the searching step size in the sparse area of the grid to avoid a large number of invalid calculations, and improves the generation efficiency of streamline.

The invention provides a method and a system for determining a segmented double-combustion-quick-charge explosive combustion surface of a solid rocket engine. The method comprises the following steps:giving an initial charging three-dimensional configuration of the engine and axial positions of high and low combustion speed interfaces of the initial charging three-dimensional configuration; meshing the cube containing the initial charge three-dimensional configuration of the whole engine, and determining the attribute of each mesh; determining high and low combustion speed initial combustionsurface node sets according to the axial positions of the high and low combustion speed interfaces; carrying out combustion surface interference judgment based on the combustion rate ratio of the high-combustion-rate charge to the low-combustion-rate charge and the minimum distance from the node at the axial position of the high-combustion-rate interface to the high-combustion-rate initial combustion surface point set and the low-combustion-rate initial combustion surface point set; and introducing a virtual combustion surface, correcting the initial combustion surface node sets of the high combustion speed and the low combustion speed, and calculating the combustion surface by adopting a minimum distance function method to respectively obtain the change rules of the charging combustion surfaces of the high combustion speed and the low combustion speed. According to the method, the calculation complexity of the combustion surface is reduced, and invalid calculation of a non-interference area is avoided.

The invention relates to a reservoir group scheduling parallel dynamic planning method considering computing resource economy and feasibility. The method comprises the steps of starting; computing resource recognition; problem scale recognition; import of known data into a transit process and trial; analysis of computing resource economy and feasibility; judgment of computing resource and problemscale matching; import of the known data into peer processes; execution of parallel dynamic planning operation; export of a reservoir group scheduling result; and stop. According to the method, the "dimension disaster" problem of reservoir group scheduling is solved according to serial dynamic planning, distributed computing is adopted to cope with a "time disaster", distributed storage is adoptedto cope with a "memory disaster", and computing efficiency and availability are improved. From the perspective of economy, distributed memories, hard disk fragments and a dynamic access technology are utilized, hard disks replace memory space, and economical expenditure is substantially saved. From the perspective of feasibility, computing time and memory and hard disk space demands are given, matching between computing resources and the problem scale is prejudged and adjusted, and the situation that overloaded operation causes ineffective computing is avoided.

The invention provides a feedback method and a device for a precoding matrix indicator (PMI). The feedback method comprises the following steps: calculation of a first PMI is carried out; reporting of the first PMI is carried out, and when the first PMI can not be acquired timely, a time sequence problem existing in calculation of the first PMI is marked; whether to carry out calculation of a second PMI is decided according to whether the time sequence problem exists in the calculation of the first PMI, and when the time sequence problem exists in the calculation of the first PMI, the calculation of the second PMI is abandoned; and reporting of the second PMI is carried out.

The invention discloses a candidate area extraction method, system, device, medium, and target detection method, and relates to the field of image processing, including: placing several initial candidate areas in an input image; Corresponding to obtaining a first candidate region and a first parameter characterizing the similarity of the first candidate region, combining multiple first candidate regions corresponding to each initial candidate region to obtain a set of first candidate regions; merging each initial candidate region The corresponding first set of candidate regions obtains a set of global candidate regions; sorting the plurality of first candidate regions in the set of global candidate regions in descending order according to the size corresponding to the first parameter to obtain a first sorting result; selecting the first sorting result The first candidate regions in the top ranks are used as the candidate region extraction results of the input image; this method has high calculation efficiency and detection accuracy.

The invention provides a method and a system for determining the combustion surface of a solid rocket motor segmented dual-combustion speed charge. The three-dimensional configuration of the initial charge of the engine and the axial position of the high and low combustion speed interfaces are given; The three-dimensional configuration of the three-dimensional configuration of the drug is divided into meshes to determine the properties of each grid; according to the axial position of the interface between the high and low burning rates, the initial burning surface node set of the high and low burning rates is determined; based on the high and low burning rate charges and the minimum distance from the node at the axial position of the interface between the high and low burning rates to the point set of the high and low burning speed initial burning surfaces to determine the burning surface interference; After the initial burning surface node set is corrected, the minimum distance function method is used to calculate the burning surface, and the change rules of the burning surface of the high and low burning rate charges are obtained respectively. The invention reduces the computational complexity of the combustion surface and avoids the invalid computation of the non-interference area.

The invention discloses an iterative propulsion disturbance domain update method for the simulation of the dynamic aerodynamic characteristics of an aircraft. Aiming at the problems of a large number of invalid calculations in the existing numerical simulation methods of the dynamic aerodynamic characteristics of the aircraft, three types of dynamic calculation domains, namely unsteady, convective and viscous, are adopted. A solution idea that only solves the unconverged unsteady element and only considers the viscous effect in the local area is proposed. The unsteady dynamic calculation domain only includes the grid units whose unsteady effect plays a leading role at each physical moment, the convective dynamic calculation domain only includes the grid units that are disturbed in each iteration step and the solution has not yet converged, and the viscous dynamic calculation domain only includes Containing grid cells that dominate the viscous effects in the flow dynamic computational domain, the three types of dynamic computational domains are adaptively updated at each iteration step. By avoiding invalid calculation in unsteady numerical simulation and reducing the number of iteration steps needed for internal iteration to reach a convergent state, the invention can significantly improve the calculation efficiency of the numerical simulation of the dynamic aerodynamic characteristics of the aircraft.

The invention discloses a multi-grid perturbation domain update acceleration method for the numerical simulation of the flow around the aircraft. Aiming at the problem of a large number of invalid calculations in the existing multi-grid method when simulating the flow around the aircraft, the fine grid and the coarse grid The convective and viscous dynamic calculation domains are used above, so that the flow control equations are solved in the convective dynamic calculation domain that only contains unconverged disturbed units in each iteration step, and the viscous effect is only in the viscous dominant unit. Considered in the viscous dynamic calculation domain, the solution idea of ​​only solving the unconverged disturbed elements and only considering the viscous effect in the local area is realized. By effectively reducing the calculation amount of a single iteration step and the number of iteration steps required to reach the convergence state, the invention can significantly improve the calculation efficiency of the multi-grid method for simulating the flow around the aircraft.

The invention discloses a method for fast triangle block design based on a pipeline and a state machine, which includes the following steps: (1) reading and numbering trimmed triangle data from a scene buffer; (2) performing viewport transformation on the triangle; 3) Calculate the bounding box and boundary function parameters of the triangle; (4) Obtain the tile range intersected with the bounding box; (5) Determine the triangle and tile intersection type according to the range in (4); (6) Set the triangle number and (3) ) (4) The obtained relevant information is written into the intermediate FIFO; (7) The Tile list generation unit reads the relevant data in the FIFO; (8) Analyze the intersection type according to the mask code; (9) Perform the LET intersection test according to the current triangle data, if the intersection Then transfer to (10), otherwise transfer to (11); (10) Calculate the write address of the tile list corresponding to the current tile.

The invention discloses an indoor fall detection method and system based on scene recognition. The method comprises the following steps: collecting video image data of a monitoring area; judging a scene category and cutting a single-person picture; intercepting single-person video data, and carrying out multi-modal processing; according to the multi-modal data and a preset tumble detection network, tumble judgment is carried out; the system comprises an image acquisition unit, a scene classification unit, a multi-mode processing unit and a tumble judgment unit, the image acquisition unit is used for acquiring video image data of a monitoring area; the scene classification unit is used for scene category judgment and single-person picture marking, the multi-modal processing unit is used for multi-modal processing to generate multi-modal data, and the tumble judgment unit is used for tumble judgment. The invention is simple, invalid calculation under a large number of unmanned scenes and multi-person scenes is avoided, and the tumble detection accuracy is improved by using multi-modal data.

The invention relates to a switch grouping method of network reconstruction of a power distribution network. The method comprises I) closing interconnection switches in the power distribution network; II) analyzing basic loops in the power distribution network; and III) grouping operable switches in the basic loops into one group. According to the provided switch grouping method of network reconstruction of the power distribution network, the theoretical foundation is strict, the calculation amount is relatively small and the calculating speed is fast; through combinations of switches in different switch groups, all feasible schemes for the network reconstruction are obtained, so that a large amount of ineffective calculation for the large-scale network reconstruction of the power distribution network is prevented and the calculating speed is improved.

The invention provides a feedback method and a device for a precoding matrix indicator (PMI). The feedback method comprises the following steps: calculation of a first PMI is carried out; reporting of the first PMI is carried out, and when the first PMI can not be acquired timely, a time sequence problem existing in calculation of the first PMI is marked; whether to carry out calculation of a second PMI is decided according to whether the time sequence problem exists in the calculation of the first PMI, and when the time sequence problem exists in the calculation of the first PMI, the calculation of the second PMI is abandoned; and reporting of the second PMI is carried out.

An OpenMP parallel disturbance domain update method for aircraft aerodynamic characteristics simulation, including data reading, static domain decomposition, establishment of dynamic calculation domain, dynamic domain decomposition, storage space allocation, boundary condition processing, residual error estimation, time integration, intra-block increment Large convective dynamic domain, shrinking dynamic calculation domain within a block, increasing viscous dynamic domain within a block, reducing viscous dynamic domain within a block, increasing dynamic calculation domain between blocks, judging whether the solution has reached the convergence condition, and outputting results can be implemented for Existing parallel numerical simulation methods have a large number of invalid calculations when simulating the aerodynamic characteristics of aircraft, and the existing disturbance region update methods may lose computational efficiency when large-scale parallelism is carried out.

The invention discloses a RANS / LES perturbation domain update method for aircraft aerodynamic characteristic simulation. Based on flow parameters such as viscosity coefficient and velocity gradient, dynamic calculation domains of RANS and LES are established to realize dynamic explicit partitioning of the RANS / LES hybrid method. Therefore, while avoiding the non-physical understanding of the "grey area", the RANS / LES hybrid method's dependence on manual experience can be reduced, and its applicability to the flow field around the complex aircraft can be improved; secondly, the unsteady dual time step method and the The solution of the constant time advancing method is unified. By establishing a dynamic calculation domain that increases with the propagation of the disturbance and converges with the solution, the coupling solution of the steady RANS and the unsteady LES is realized, which can not only ensure that there will be no invalid calculations in the numerical simulation, but also Additional boundary condition processing between RANS and LES work areas can be avoided.

The invention discloses an overlapping grid disturbance domain update method for multi-body separation simulation of an aircraft. First, the wall distance algorithm is simplified based on the characteristics of the wall distance marking unit attribute using the implicit assembly method, thereby improving the efficiency of grid assembly; secondly, Under the framework of the disturbance domain update method, the numerical simulation of the overlapping grid is realized, so that only the cells in the unconverged disturbed area are interpolated, thereby reducing the calculation amount of the overlapping grid interpolation operation; again, a new type of overlapping grid is established The information exchange method realizes the transmission of information based on grid deformation and encryption in the large flow gradient area, and the way of transmitting information based on flow gradient interpolation in the small flow gradient area, thereby improving the accuracy of the overlapping area and avoiding the "solitary point" problem. This method can significantly improve the computational efficiency of the numerical simulation of the aircraft multi-body separation problem, and solve the problems of numerical accuracy loss in the overlapping area of ​​the large flow gradient grid and the existence of "solitary points".

The invention relates to the technical field of computational fluid dynamics, and provides a time-parallel disturbance domain update method for aircraft dynamic characteristic simulation. The method includes: reading in data; obtaining the Chebyshev transformation matrix components; initializing the flow field; establishing a dynamic calculation domain; solving the flow control equations of all time layers in parallel in the dynamic calculation domain; updating the dynamic calculation domain; judging whether the parallel solution in the current time period converges, Whether the calculation is completed, if it converges, jump to the dynamic calculation domain to solve all time layers in parallel, and enter the next iterative step of the parallel solution in the current time period; if the solution at the current moment has converged and the calculation is not completed, then jump to flow field initialization , enter the solution of the next time period; otherwise, output the result. The invention avoids the invalid calculation caused by adopting the static calculation field, solves the problem of acceleration bottleneck limited by the time dimension sequential advance solution requirement, and significantly improves the calculation efficiency.

The invention provides a progressive matching design method for the performance of solid rocket motors, which includes: obtaining the mission target of the overall index of the solid rocket motor; According to the initial training sample set, run the combustion surface and internal ballistic calculation model in the initial time interval to obtain N thrust curves, and build a thrust approximate model according to the N thrust curves ; Perform inaccurate sampling to obtain an alternative solution; substitute the alternative solution into the combustion surface calculation and internal ballistic calculation model, calculate the internal ballistic performance within the simulation time, and obtain an alternative thrust curve; when the simulation time is not less than the described work time, the alternative solution is added to the sample set as a new sample point to obtain an updated sample set; when the preset convergence condition is reached, the alternative thrust curve is output as the optimal solution.

The invention discloses a fast HEVC inter-frame coding method, which determines whether to terminate the motion estimation in advance by obtaining the result of the integer pixel search of the Pb1 block in the Rect mode prediction, and the termination of motion estimation in advance refers to: determining whether the integer pixel motion vector of the Pb1 block and the motion vector trend of Pb0 is consistent after completing the integer pixel search of the Pb1 block, if so, determining that the current prediction mode belongs to an invalid calculation, and skipping the sub-pixel search of thePb1 block. The method has the beneficial effects that the pixel search time of the Pb1 block is saved, thereby avoiding a large number of invalid calculations and improving the overall coding speed.

The invention relates to a reservoir group scheduling parallel dynamic planning method considering computing resource economy and feasibility. The method comprises the steps of starting; computing resource recognition; problem scale recognition; import of known data into a transit process and trial; analysis of computing resource economy and feasibility; judgment of computing resource and problemscale matching; import of the known data into peer processes; execution of parallel dynamic planning operation; export of a reservoir group scheduling result; and stop. According to the method, the "dimension disaster" problem of reservoir group scheduling is solved according to serial dynamic planning, distributed computing is adopted to cope with a "time disaster", distributed storage is adoptedto cope with a "memory disaster", and computing efficiency and availability are improved. From the perspective of economy, distributed memories, hard disk fragments and a dynamic access technology are utilized, hard disks replace memory space, and economical expenditure is substantially saved. From the perspective of feasibility, computing time and memory and hard disk space demands are given, matching between computing resources and the problem scale is prejudged and adjusted, and the situation that overloaded operation causes ineffective computing is avoided.

The invention provides a method and a device for performing convolution calculation on an image by using a convolutional neural network, wherein the image is input into a convolutional neural network;wherein the first convolutional layer outputs a first feature map, the first feature map comprises a first control channel, and the first convolutional layer is a middle convolutional layer of the convolutional neural network; the method comprises the following steps: taking an area with a characteristic value smaller than a preset threshold value in a first control channel as a first control area, wherein the characteristic value of any position of the first control channel is used for representing the probability that the corresponding position of a first characteristic graph belongs to target characteristics; determining a redundant area of the first feature map according to the first control area; and other regions except the redundant region in the first feature map enter the next convolution layer of the first convolution layer for convolution operation to obtain a second feature map. In the mode, the redundant area does not perform the next layer of convolution operation, invalid calculation can be prevented, the calculation amount of the neural network is reduced, the convolution calculation efficiency is improved, and the power consumption of convolution calculation is reduced.