16 results about "Packing algorithm" patented technology

The invention provides a numerical control multi-objective optimization energy-saving method and system based on an improved wolf pack algorithm, and the method comprises the steps: constructing a mixed integer linear programming model of multi-objective flexible job shop scheduling, taking machine tool distribution, a machining sequence, tool distribution and a machining speed as decision variables, and taking total weighted tardiness and total energy consumption as optimization targets; the method comprises the following steps: constructing an MOGWO-ALNS algorithm for solving a mixed integer linear programming model based on an MOGWO algorithm and an adaptive large neighborhood search mechanism, constructing a three-stage collaborative optimization architecture by the MOGWO-ALNS algorithm through a dual-threshold trigger mechanism, solving the mixed integer linear programming model based on the MOGWO-ALNS algorithm, obtaining a Pareto optimal solution set of the mixed integer linear programming model, decoding the Pareto optimal solution set, and finally obtaining the mixed integer linear programming model. And generating a multi-target balanced energy-saving scheduling scheme. According to the method, collaborative optimization of the delivery date and the energy consumption in numerical control machining is achieved, the solving efficiency and the optimization effect are improved, and the method has good engineering application value.

The invention relates to the technical field of control system data processing, in particular to a ship intelligent workshop crane hoisting path planning method, which comprises the following steps: collecting crane coordinates, obstacle distribution, load swing angle and material weight, and inputting a physical constraint model to generate a structured environment state data set; initializing wolf pack algorithm parameters based on the structured environment state data set, calculating a compressed detection wolf walking step length according to an obstacle distance, generating a global task queue by an industrial cloud platform according to a dynamic parameter set, and outputting an execution state data set by a crane end; counting emergency braking times from an execution state data set to optimize a walk step length function, comparing task time to optimize a wolf investigation proportion, analyzing historical swing data to establish a step length and swing mapping library, and optimizing a parameter set to feed back and update a physical constraint model. According to the invention, cooperation of environment perception, dynamic path planning and closed-loop verification is realized, and accident risks and production delay in a ship workshop narrow space dynamic obstacle scene are reduced.

The application provides a multi-unmanned aerial vehicle task allocation system based on a two-stage multi-population wolf swarm algorithm, comprising: an initialization module, which defines the individual wolf position in the two-stage multi-population wolf swarm algorithm as a multi-unmanned aerial vehicle task allocation scheme, divides the wolf swarm into multiple sub-populations, and comprises an ordinary group and a leader group; a wandering module: the ordinary group and the leader group explore wolves and adopt different wandering number calculation methods to reorder task numbers; a calling module: the ordinary group and the leader group use different convergence speeds to learn the scheme of the current alpha wolf; a siege module: except for the alpha wolf, the individual wolves of the ordinary group and the leader group use different threshold values to mutate or learn the scheme of the individual wolf with low fitness; an individual wolf reconstruction module: a roulette mechanism is adopted for reconstruction; and a sub-population adjustment module: after reconstruction is completed, the multi-population mechanism is started, and the leader group is reconstituted. The application improves the wolf swarm algorithm through multiple mechanisms, and improves the solving efficiency and accuracy.

This application relates to efficient convolution in tiled environments. One method includes: receiving an input tensor, the input tensor having a domain consisting of [n1, ..., n]. k The shape is defined by [t1, ...,t], where k is equal to the number of dimensions representing the input tensor; it receives tile tensor metadata, which includes: [t1, ...,t] k The tile tensor shape is defined, along with information indicating the interleaving stride to be applied relative to each dimension of the tile tensor; an output tensor comprising multiple tile tensors is constructed by applying a packing algorithm that maps each element of the input tensor to at least one slot position of one of the multiple tile tensors based on the tile tensor shape and the interleaving stride, wherein the interleaving stride results in a discontinuous mapping of the elements of the input tensor such that each of the tile tensors comprises a subset of the elements of the input tensor that are spaced apart within the input tensor according to the interleaving stride.

The application provides a sparse matrix accelerator based on a block-based packing algorithm, comprising an algorithm preprocessing module, a calculation module and a sorting module. The algorithm preprocessing module can obtain a sparse matrix comprising a flow matrix and a stationary matrix, block the flow matrix and the stationary matrix according to a preset numerical value, then perform shift and merge processing on the first non-zero element, the second non-zero element and the zero element in the matrix respectively to obtain a dense flow matrix and a dense stationary matrix, and save the address of the first non-zero element before shifting and the address of the second non-zero element before and after shifting. The calculation module can perform multiplication and addition calculation on the dense flow matrix and the dense stationary matrix to obtain a partial sum result and a first address result of the partial sum result. The sorting module can sort the partial sum result according to the first address result to obtain a result matrix. The above accelerator can improve the density of the matrix, at the same time, correct column addition can be realized, and the calculation efficiency can be improved.

The invention discloses a low-consumption and high-credibility text analysis method and device based on a large language model, and relates to the technical field of natural language processing. The method comprises the following steps: based on a preset context efficiency coefficient, performing calculation according to a context capacity maximum threshold value and a single request interception threshold value to obtain an effective load safety threshold value; on the basis of the effective load safety threshold value, according to the system instruction template, the JSON protocol template, the service text data, the current text data and the preset safety spare token length, a dynamic packing algorithm is used for carrying out request aggregation, and a JSONL request is obtained; and performing text analysis by using a large language model according to a JSONL request based on a three-level analysis mechanism of consistency verification to obtain a comprehensive judgment result and a comprehensive judgment evidence. The invention relates to a lightweight text analysis method for overcoming model illusion and resource bottleneck based on a large language model.

The invention discloses a college-bus route planning method based on a multi-constraint clustering and combined heuristic algorithm, and relates to the technical field of route planning, and the method comprises the steps: generating an information table of students needing to be picked up; according to the time difference between the school arrival time and the school departure time, the students in the student information table needing to be picked up are batched, and a batch student information table is obtained; generating initialized boarding points for students in the batch student information table, and iteratively optimizing the initialized boarding points according to constraint conditions, so as to generate a concentrated boarding point information table of a corresponding batch; generating an initial route scheme table according to the centralized boarding points in the centralized boarding point information table of each batch by using a greedy boxing algorithm; optimizing routes in the initial route scheme table by using a tabu search algorithm to obtain an optimized route scheme table; and a multi-route connection strategy of a single college bus is formulated according to the optimized route scheme table of each batch, and then a college bus connection scheduling table is obtained. According to the invention, the college scheduling line can be reasonably planned.

The invention relates to the technical field of boxing algorithms, in particular to a boxing method, device and equipment and a computer storage medium. According to the boxing method, high-quality candidate placement points are dynamically generated and maintained through a two-step point increasing method, and the comprehensiveness and tightness of space exploration are ensured; business constraints such as cargo grouping and loading sequence are converted into allowable areas in geometric space through a projection filtering point method, and deep fusion of business rules and spatial layout is realized; and through an orientation search method, intelligent preferential selection is carried out in multiple orientations, the space utilization rate is maximized, and the placement stability is ensured. The three collaboratively form a closed loop, so that the loading space utilization rate, the loading accuracy and the calculation efficiency are finally improved, and business rules are met.

The application provides a distributed index construction method based on improved iSAX coding, first, aiming at the digital feature selection problem, in order to increase the coding similarity of similar data, a similarity digital coding is designed, and through the transposition of the matrix, the problem of different weights of high bits and low bits of each digital coding in the whole digital coding is solved. Secondly, aiming at the problem of high time consumption in the traversal process caused by the too high index tree, a B+ index tree based on the similarity digital coding is designed, the number of child nodes is increased, the height of the tree is reduced, and the access speed of adjacent nodes is improved. And a leaf partition snake-shaped packing algorithm is designed, which guarantees the load balance, shortens the leaf node packing time, improves the index construction speed, and improves the overall index construction speed compared with the traditional distributed index construction algorithm, and provides a more efficient index framework for the similarity query process.

The invention discloses a two-stage three-dimensional boxing method based on a double-population strategy, and provides a two-stage boxing optimization strategy, a hybrid heuristic algorithm is used as a first stage, and a splitting and replacing strategy with smaller calculated amount is used as a second stage of the algorithm. Meanwhile, an improved angle occupying method is used as a heuristic boxing algorithm, and a hybrid heuristic algorithm is improved by adopting a double-population strategy, a goods pre-combination strategy and an improved simulated annealing algorithm dynamic temperature rise strategy; the strategies of the second stage further comprise a four-corner tightening strategy, a recombination strategy and a large object sinking strategy, the calculation amount can be remarkably reduced under the condition that the quality of solutions is improved, and the diversity of populations is guaranteed. The boxing method provided by the invention has a wide prospect in the field of logistics.

The invention provides an online fault line selection method and system based on an improved wolf pack algorithm and a storage medium, and the method comprises the steps: 1, collecting a zero-sequence current component through a current sensor, and carrying out the variational mode decomposition to obtain four variables [delta] E1, [delta] E2, [delta] E3 and [delta] E4; 2, the four variables delta E1, delta E2, delta E3 and delta E4 after variational mode decomposition are optimized through an improved wolf pack algorithm, and optimal solutions are obtained; and step 3, the four variational mode decomposed variables delta E1, delta E2, delta E3 and delta E4 and the optimized solution of the improved wolf pack algorithm are sent to an RBF neural network for training, and a fault line selection result is obtained. According to the online fault line selection method and system based on the improved wolf pack algorithm and the storage medium, rapid positioning of a fault line can be realized, the neural network learning efficiency is higher, and conditions are created for realizing accurate, rapid and intelligent targets of fault line selection.

This invention discloses a method for predicting the throughput capacity of a ship lock. The method involves obtaining the total cargo volume, ship type, and the proportion of cargo volume carried by each ship type for the predicted year; calculating the total number of ship trips for each ship type in the predicted year; randomly combining ship types and numbers into the same lock throughput batch; determining the platooning pattern of ships within the same throughput batch using a two-dimensional packing algorithm based on the lock's length and width and the ship's length and width; calculating the average deadweight tonnage per lock trip for each ship type in the platooning pattern that completes the predicted annual cargo volume according to the platooning pattern; and taking the maximum value among the calculated average deadweight tonnage per lock trip for several platooning patterns as the predicted annual lock throughput capacity. The advantages of this invention are its clear principle, simple calculation process, accurate results, and high efficiency, providing more precise data support for ship lock engineering design.

This invention proposes a parking space planning method based on dynamic programming and bin packing algorithm, comprising: Step 1, constructing an abstract layout of the apron, wherein the abstract layout includes setting the area of ​​the apron and defining obstacles and service lanes in the apron area; Step 2, combining the aircraft types to be parked according to their available taxiing methods to form multiple combinations of aircraft types and taxiing methods, calculating the footprint of each combination of aircraft types and taxiing methods, and calculating the width of the right taxiway based on the footprint; Step 3, treating the parking space planning problem as a complete knapsack problem; each combination of aircraft types is regarded as an item to be packed, the value of each aircraft type is regarded as the item value, the available area of ​​the apron is regarded as the knapsack capacity, determining whether each combination of aircraft types can be parked on the apron according to the parking conditions, and obtaining all possible parking space planning results in the order of each row.

The invention discloses a large fragment generator power field simulation method and a computer product, and the method comprises the steps: carrying out the discretization of the boundary of a closed space surrounding a fragment layer through employing a discrete surface element, and carrying out the discretization of the internal space of the closed space through employing a voxel method; the internal space of the closed space is iteratively filled with spherical fragments through a random confidential arrangement algorithm, and initial arrangement of the fragments is completed; the initial arrangement of the fragments is optimized through a dynamic relaxation method, fragment compaction state simulation is completed, and the spatial position and radius information of the fragments is output; generating a discrete element model of the fragments based on the spatial position and radius information of the fragments; and based on the discrete element model of the fragments, combining a finite element method and an ALE method to realize modeling of a warhead containing a large number of fragments and rapid simulation evaluation of a power field. According to the method, a more accurate and detailed simulation result is achieved, and technical support can be provided for fragment type ammunition design and rapid performance evaluation.

The application discloses a kind of 3D irregular object packing methods and systems for subtractive manufacturing, and relates to subtractive manufacturing technical field.The method includes the following steps: obtaining container raw materials and 3D objects to be packed;According to container and 3D object, collision-free space is calculated, and collision-free region is initially generated;According to the operation distance of cutting tool, the collision-free region is further filtered to form candidate placement position;According to the Euclidean distance, the candidate position is sorted to form packing layout;According to the accessible region of all 3D objects, cutting shell is formed, and cutting path is planned according to cutting shell;According to packing layout, cutting shell and cutting path, 3D object is cut.The application solves the limitations of traditional packing algorithm in subtractive manufacturing, and also provides a new solution for efficient and accurate manufacturing process.

The application discloses a multi-request processing method, device, system, storage medium and product, and the method comprises the following steps: when receiving a plurality of query requests sent by target terminals, the plurality of query requests are processed in batches by calling a suppression interface of a target attack handling platform; the classification of the batch processing results is performed according to the identification information of each thread; the classification processing results are fed back to the corresponding target terminals; in the above manner, when receiving a plurality of query requests sent by target terminals, it indicates a high concurrency scenario, at this time, without changing the existing user capability access mode, the single suppression interface and the multi-batch suppression interface of the attack handling platform are called by using a packing algorithm to process the plurality of query requests in the high concurrency scenario in batches, so that the processing of the plurality of query requests is completed under the condition of limited performance, and then the processing results are fed back to the corresponding target terminals, thereby effectively improving the efficiency of processing multiple requests and improving network security.