16 results about "Fiber clustering" patented technology

The invention relates to a detection and in-situ repair method and system for spinning yarn knots, and the method comprises the steps: applying high-frequency vibration to a twisting triangular area of a spinning frame, and separating loose fiber clusters attached to the surfaces of yarns through axial vibration; carrying out polarized light scanning on the surface of the yarn to obtain three-dimensional structure characteristics of the yarn, and identifying a candidate knot area with a sudden diameter change; an instantaneous stress pulse is applied to the candidate knot area, the local elastic response of the yarn is detected, the authenticity of the candidate knot area is judged, real knots show rigidity sudden change, and loose fibers show elastic attenuation; directionally spraying an atomized solvent to the area which is confirmed as a real knot to soften fibers, and meanwhile, applying radial compression force to redirect the fibers to recover the continuity of the yarn; the local temperature distribution of the repaired yarn is detected through infrared thermal imaging, and if the temperature is abnormal, secondary repairing is triggered or the yarn is marked as an unrepairable section. The method effectively solves the problems that a traditional method is high in misjudgment rate and poor in repairing capacity.

The invention relates to a method for characterizing the size of an irregular proppant-fiber cluster unit in a slit, and the method comprises the steps: S1, carrying out a proppant-fiber mixed conveying physical model experiment, obtaining an experiment image, carrying out the preprocessing and threshold segmentation of the experiment image, focusing on a region of interest (ROI), and obtaining a proppant-fiber cluster unit size; dividing the pixel value of the image into continuous intervals according to a preset threshold parameter, and generating a multi-category segmentation mask; s2, performing multi-scale morphological operation on the segmented masks, removing small noise cluster units and filling internal holes of the cluster units; s3, identifying each connected region in the image, counting the area of the connected region, and dividing cluster units according to a preset level; and S4, performing color gradient rendering, contour labeling and area numerical value labeling on the cluster units of different levels to enhance the visual discrimination, and outputting an image. According to the experimental image processing method provided by the invention, the information of each cluster unit can be more intuitively obtained, and the action mechanism of the proppant-fiber cluster unit can be clearly known.

The invention provides an in-slit amorphous proppant-fiber cluster unit identification and size distribution characterization method, which comprises the following steps: S1, carrying out a proppant-fiber conveying experiment, carrying out local distribution division on an experiment picture according to a preset method, and respectively carrying out threshold segmentation on each local block, obtaining an overall mask matrix and dark pixel point coordinates of the experimental picture; s2, each dark pixel point is taken as an initial center, expansion is carried out around, a cluster unit index matrix is formed, and cluster unit segmentation is realized; s3, filling holes in the cluster units, identifying a hole area completely surrounded by a single cluster unit by scanning zero-value elements in an index matrix of the cluster units, and filling the holes with indexes of peripheral clusters; and S4, carrying out statistics on all cluster units, converting a pixel area into a real area, and outputting size distribution parameters of the cluster units. According to the invention, the automatic identification of the amorphous proppant-fiber cluster unit in the slit and the characteristic parameter output of the cluster unit can be realized.

The invention provides a human brain superficial layer white matter fiber structure map construction method and system, and the method comprises the steps: generating a finely divided whole brain function fiber cluster map in an unsupervised manner based on whole brain diffusion magnetic resonance imaging data; in combination with a geometric feature threshold value and a deep learning classifier, identifying superficial layer white matter fibers from the whole brain function fiber cluster and carrying out hierarchical clustering to generate a superficial layer white matter fiber sub-atlas with the optimal clustering number; and analyzing the functional attributes of the superficial layer white matter fibers through a trajectory weighted dynamic functional connection analysis method to obtain a superficial layer white matter fiber structure chart of the human brain. According to the method, the technical blank in the field is filled, and a high-quality data acquisition channel is provided for research based on superficial layer white matter.

The present disclosure belongs to the technical field of oil exploitation, and specifically relates to a temporary plugging fiber cluster, and a preparation method therefor, a usage method therefor and the use thereof. The temporary plugging fiber cluster provided by the present disclosure comprises degradable fiber filaments and a degradation promoter, wherein the degradable fiber filaments are wound around the outside of the degradation promoter.

The application provides a fiber bundle division method and system based on fiber similarity and a storage medium, and relates to the technical field of image processing; the method comprises the following steps: performing interval division on each fiber of a target fiber bundle, performing cluster point sampling on the fibers in each interval, calculating the similarity between the fibers based on the cluster points, generating an initial similarity matrix, performing hierarchical clustering on the fiber bundle based on the similarity matrix, generating a preliminary clustering result, calculating the similarity of fiber clusters between intervals based on the cluster points, generating a clustering similarity matrix, performing hierarchical clustering on the fiber clusters based on the clustering similarity matrix, and performing hierarchical clustering on the clustered result to generate a fiber clustering result of the target fiber bundle. The fiber bundle is clustered multiple times, so that the fibers in the clusters have tightness, and the clusters have separation degree, thereby improving the fiber bundle division precision and efficiency.

The application discloses a cross-subject electroencephalogram emotion calculation method based on manifold dynamics, belongs to the field of electroencephalogram emotion calculation, and comprises the following steps: modeling an electroencephalogram observation space as a fiber cluster structure, realizing decoupling of emotional semantics and individual styles through instance normalization, and generating an enhanced sample set by using random style perturbation. A consistent autoencoder is constructed to map the samples to a low-dimensional latent space, and the geometric barycenter of multi-source domain data is mined as a canonical attractor. On this basis, a conditional flow matching technology is adopted to learn a time-dependent vector field, the source domain distribution is continuously evolved to the canonical attractor along an optimal transmission path by driving the source domain distribution through a neural ordinary differential equation, and the aligned features are obtained to complete emotion classification. The application realizes effective elimination of individual differences under the premise of maintaining the integrity of the topological structure, and significantly improves the accuracy and robustness of cross-subject emotion recognition.

The invention relates to the technical field of equipment control, and discloses a laboratory equipment control system and method based on data processing, and the system comprises a multi-scale feature analysis module, a topological manifold evolution module, a fiber cross section solving module, a singular point standard resolution module, a leaf-shaped structure discretization module and a semantic instruction coding module. Analyzing the original data to obtain multi-level stability features; carrying out topological structure evolution on the dynamic operation state to obtain a state evolution manifold; performing relation mapping on the state evolution manifold, and performing constraint derivation on the fiber bundle structure to obtain a continuous control section; performing abnormal point digestion on the continuous control section to obtain an ideal control manifold; carrying out leafy structure segmentation on the ideal control manifold, and carrying out cross discretization on the basic leafy structure to obtain a control action sequence; performing semantic coding on the control action sequence to obtain an operation instruction set; according to the invention, the reference information generation efficiency based on artificial intelligence and smart home can be improved.

The present application relates to a kind of surface flowing fiber cluster topology optimization method;The present application uses material interpolation method to determine surface flowing model, implicit two-dimensional manifold of surface flowing is defined on bottom flow form, porous medium is filled on implicit two-dimensional manifold, while artificial Darcy friction coefficient is introduced to surface Navier-Stokes equation, and surface flowing fiber cluster topology optimization is realized by filling porous medium to fixed two-dimensional manifold.

The invention discloses an image instance segmentation method for overlapping and separating fiber agglomerates in a papermaking process, and the method comprises the following steps: constructing a papermaking fiber suspension microscopic image data set, and carrying out the data preprocessing; constructing an image instance segmentation network model, and dynamically generating a thermodynamic diagram prior graph of a fiber category by using a Transform decoder output feature; carrying out down-sampling on a prior graph of the thermodynamic diagram and then fusing with original mask attention, and constructing an attention enhancement mechanism guided by thermodynamic diagram saliency; and constructing a thermodynamic diagram saliency loss function containing coverage degree and separation degree constraints, performing joint optimization with the native loss, and finally outputting a fiber mass image instance segmentation result. The method can effectively improve the separation precision of overlapped fiber agglomerates and the segmentation performance of small-scale instances, and is suitable for intelligent analysis of high-flux and high-precision fiber morphology in the papermaking industry.

The invention discloses a method and system for cross-species whole-brain homologous cortex partitioning, and the method comprises the steps: respectively constructing a vertex-fiber cluster feature matrix and a vertex-white matter beam incidence matrix based on diffusion magnetic resonance imaging data of human and other species, horizontally splicing the feature matrixes of human and other species along the vertex dimension, and carrying out the cross-species whole-brain homologous cortex partitioning. Generating a cross-species joint feature matrix; taking the cross-species joint feature matrix as a clustering feature, and performing clustering to obtain a super vertex public partition template; averaging the vertex-fiber cluster feature matrix and the vertex-white matter beam incidence matrix, optimizing average features through a low-rank tensor consensus graph to obtain consensus graphs, and fusing the consensus graphs; and finally partitioning the optimized fusion consensus graph based on a spectral clustering algorithm, and judging a homologous region by combining the hyper-vertex overlapping rate in the partitions of each pair of human and other species to obtain a cross-species homologous corticogram. According to the method, homologous partition of cerebral cortex of human beings and macaque is realized, and cross-species consistency and biological correlation are remarkably improved.

The present invention relates to a method, system, electronic device and medium for generating a random distribution of fibers, and particularly relates to the technical field of composite materials. The method includes taking the first fiber as the center, and repeatedly using the artificial fish swarm algorithm to generate an initial fiber cluster within the fiber generation window; moving a search box within the fiber generation window where the initial fiber cluster is generated, and when the fiber volume fraction within the search box after each movement is less than a preset volume fraction, then based on the hardcore model, scatter points within the search box after each movement to obtain new fibers until the search box finishes moving within the fiber generation window, and deleting the fibers within the fiber generation window according to the preset volume fraction to obtain the finally generated fibers. The present invention can effectively generate a random distribution structure of fibers and can solve the cluster problem.

The present invention relates to a method and system for detecting and in-situ repairing knots of textile yarns. High-frequency vibration is applied in the twist triangle area of a ring spinning frame to separate loose fiber clusters attached to the yarn surface through axial vibration. The yarn surface is scanned with polarized light to obtain its three-dimensional structural characteristics and identify candidate knot areas with sudden diameter changes. An instantaneous stress pulse is applied to the candidate knot areas, and the local elastic response of the yarn is detected to judge the authenticity of the candidate knot areas, where: real knots show a rigid mutation, while loose fibers exhibit elastic attenuation. For the areas confirmed to be real knots, atomized solvent is sprayed directionally to soften the fibers, and at the same time, a radial compression force is applied to reorient the fibers and restore the continuity of the yarn. The repaired yarn is detected for the local temperature distribution through infrared thermal imaging. If there is a temperature anomaly, secondary repair is triggered or it is marked as an irreparable section. The present invention effectively solves the problems of high false positive rate and lack of repair ability in traditional methods.

The present invention relates to a numerical simulation method for flexible fiber bundles in a cement matrix, which is used to generate a random distribution model of flexible fibers in the cement matrix. This method first generates a fiber cluster containing multiple curved fibers. The fibers in the fiber cluster do not intersect with each other. Only by calculating the distances between the fibers at the centers of each fiber cluster can it be determined whether the fibers intersect. Therefore, the operation rate is greatly improved, the program running time is reduced, and it conforms to the existing state of flexible fibers in the cement matrix.

The application discloses a kind of multi-working condition motion control method and system of movable platform, belong to multi-working condition motion control technical field;The method specifically includes: collection multi-working condition motion sequence set, constructs multi-working condition motion fiber cluster space, and is searched by particle swarm optimization algorithm transition mapping parameter optimization, obtains multi-working condition control benchmark set;Based on multi-working condition control benchmark set, motion semantic reduction operator fragment is screened out, and normal form rewriting reduction is carried out, constructs heading velocity coupling change diagram, obtains unified motion target set;Based on unified motion target set, generates multi-working condition motion joint control instruction sequence, carries out multi-working condition motion control to movable platform;The application improves the control continuity of multi-working condition motion control, enhances the stability of multi-working condition motion control.

The application belongs to the field of atomic emission spectrum analysis equipment and analysis technology, and particularly relates to a three-dimensional network fiber cluster micro-plasma excitation source and a spectrum detection device and method. The excitation source comprises a fiber tip discharge cluster working electrode and a counter electrode. The fiber tip discharge cluster working electrode and the counter electrode are horizontally arranged opposite to each other and a preset discharge gap is kept in the middle. The fiber tip discharge cluster working electrode and the counter electrode are electrically connected with the positive pole and the negative pole of a high-voltage direct current power supply respectively to form a discharge loop. The fiber tip discharge cluster working electrode comprises a fiber cluster conductor and an insulating support tube. The fiber cluster conductor is assembled in a strip or columnar form in the inner cavity of the insulating support tube. The fiber cluster conductor is a carbon felt strip. The three-dimensional isotropic random interweaving characteristics of the carbon felt fiber are utilized to construct a high-density micro-discharge site in a millimeter-level limited space. The field intensity interference between parallel fibers is effectively broken, and a cluster-shaped micro-plasma with high energy density is induced at a lower direct current voltage.