CN-121979890-A - Associated index construction method and system for multidimensional heterogeneous data

CN121979890ACN 121979890 ACN121979890 ACN 121979890ACN-121979890-A

Abstract

The invention relates to the technical field of electric digital data processing, in particular to a method and a system for constructing an associated index of multidimensional heterogeneous data. And identifying a target response window of the task main body, carrying out semantic alignment with the output intensity envelope curve of the device, and calculating through time sequence overlapping degree to obtain a correlation coupling degree score. And searching tasks from the candidate pool according to the scores, combining the load limit value, automatically distributing the starting moment and the load quota, and generating an initial execution sequence meeting the double constraint of hardware safety and execution efficiency. When the interruption of the execution flow is monitored, the task state snapshot is synchronously stored, the residual evolution rate is called, the real-time availability characteristic is coupled to carry out secondary semantic deduction, and a dynamically corrected re-index task sequence is generated on the premise of ensuring that the load limit value is not triggered.

Inventors

GU LONG
SU JIANJUN
ZHOU XINGCHEN
Su Xingkang
WANG GUAN
ZHANG LU
ZHANG JUNHAN
LIN WEIQIANG

Assignees

福建睿斯科医疗技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260402

Claims (8)

1. The method for constructing the association index of the multidimensional heterogeneous data is characterized by comprising the following steps of: Collecting an operation state field of a physical source item device, a time-varying evolution field of a task main body and a topology constraint field of an execution space, and carrying out structural encapsulation and normalization processing to obtain a heterogeneous feature vector object; The method comprises the steps of identifying a target response window of a task main body in the heterogeneous feature vector object, carrying out semantic alignment on an expected output intensity envelope curve of a physical source item device on a unified time axis, carrying out time sequence overlapping degree calculation on data subjected to semantic alignment, and obtaining an associated coupling degree score; target retrieval is carried out from the task candidate pool according to the associated coupling degree score, and the execution starting time and the load quota of each task are distributed by combining with the component load limit value field of the physical source item device to obtain a multi-target initial execution sequence; and when the execution flow interruption is monitored, synchronously storing a task state snapshot comprising the accumulated output count and the instantaneous environment parameter, and calling a residual evolution rate field of a task main body and coupling the real-time availability characteristic of a source item device to carry out secondary semantic deduction, thereby obtaining a dynamically corrected re-index task sequence on the premise of ensuring that the component load limit value is not triggered.
2. The method for constructing the association index of the multidimensional heterogeneous data according to claim 1, wherein the method for acquiring the operation status field of the physical source item device specifically comprises the following steps: The method comprises the steps of recording a dynamic scanning track field of an energy beam on the surface of a conversion medium, constructing a space-time evolution model representing the overlapping process of energy deposition and thermal diffusion by utilizing a preset thermal conduction response matrix dynamically distributed along with space coordinates, generating a real-time thermal load mapping field of a physical source item device based on the space-time evolution model, predicting an instantaneous loss rate trend of the conversion medium material by utilizing nonlinear associated data of historical accumulated energy load and environmental temperature rise, predicting the reduction amount of the medium thickness according to the instantaneous loss rate trend, and feedforward calculation of a predicted offset characteristic field of a source item output energy spectrum.
3. The method for constructing the association index of the multidimensional heterogeneous data according to claim 1, wherein the method for obtaining the heterogeneous feature vector object specifically comprises: The method comprises the steps of extracting real-time load intensity and energy spectrum offset parameters in an operation state field of a physical source item device, mapping the parameters to a preset first numerical dimension space, identifying the activity attenuation rate and staged completion degree in a time-varying evolution field of a task main body, and converting the activity attenuation rate and staged completion degree into a time sequence matrix conforming to a second numerical dimension space; And carrying out linear stretching and dimension unified processing on the three numerical dimension spaces by utilizing normalization, and carrying out tensor splicing according to a preset field index sequence to generate the heterogeneous feature vector object.
4. The method for constructing the association index of the multidimensional heterogeneous data according to claim 1, wherein the obtaining the association coupling degree score specifically comprises: The method comprises the steps of extracting trend components from time-varying evolution fields of a task main body, identifying numerical peak value intervals reflecting response sensitivity of the main body in the trend components to obtain target response window fields, mapping expected output intensity envelopes of the target response window fields and a physical source item device to a unified time sequence reference axis, calculating numerical matching degrees of the expected output intensity envelopes and the target response window fields in an overlapping time interval on the unified time sequence reference axis to obtain time sequence overlapping degree calculation results, and carrying out extremum retrieval based on matching degree distribution rules in the overlapping time interval to generate the associated coupling degree scores.
5. The method for constructing the association index of the multidimensional heterogeneous data according to claim 1, wherein the generation of the multi-objective initial execution sequence comprises the following steps: traversing each feature vector in the task candidate pool, defining the associated coupling degree score as a control variable of a main body response weight, defining the component load limit value field as a hardware capacity boundary, constructing a multi-target decision matrix in a discrete time axis, dynamically adjusting the retrieval priority score of each task by calculating the resource load redundancy of each task at different candidate starting moments, retrieving target tasks according to the order of the priority scores from high to low, endowing the target tasks with corresponding execution starting moment fields and load quota fields, and packaging to obtain the multi-target initial execution sequence.
6. The method for constructing an associated index of multidimensional heterogeneous data according to claim 1, wherein the step of synchronously storing the task state snapshot including the accumulated yield count and the instantaneous environment parameter when the execution flow interruption is detected comprises: The method comprises the steps of monitoring the running state of an execution flow in real time, synchronously extracting a current accumulated output count field, an instantaneous environment parameter field and a staged completion percentage field of a task main body when an interrupt signal is captured, carrying out serialization processing on the extracted fields to generate a task state snapshot containing execution context information, carrying out coupling calculation by utilizing the residual task quantity in the task state snapshot and combining with the residual evolution rate field of the task main body, and deducting to obtain an enhanced execution intensity coefficient required by supplementing a target task quantity in the activity period of the current main body.
7. The method for constructing the associated index of the multidimensional heterogeneous data according to claim 6, wherein the obtaining the dynamically modified re-index task sequence specifically comprises: The method comprises the steps of extracting a failure risk time limit field of a task main body and real-time availability characteristics of a physical source item device, comparing an enhanced execution intensity coefficient with a component load limit field, starting a safe load shedding mechanism to adjust down a task output target value recorded in a task state snapshot if a predicted load indicated by the enhanced execution intensity coefficient exceeds a limit value, carrying out load remodeling on task balance data by utilizing the enhanced execution intensity coefficient after verification and coupling an energy conversion efficiency correction item, calculating to obtain corrected load quota of each task unit, resetting priority scores of tasks affected by interruption according to the failure risk time limit field, reinserting a task sequence with the corrected load quota into a queue occupation, and generating a re-index task sequence meeting system comprehensive execution efficiency limit compensation.
8. A system for constructing an associative index of multidimensional heterogeneous data, comprising: The feature encapsulation module is used for acquiring an operation state field of the physical source item device, a time-varying evolution field of the task main body and a topology constraint field of the execution space, and carrying out structural encapsulation and normalization processing to obtain a heterogeneous feature vector object; The semantic alignment module is used for identifying a target response window of the task main body in the heterogeneous feature vector object, carrying out semantic alignment on the target response window and an expected output intensity envelope line of the physical source item device on a unified time axis, and carrying out time sequence overlapping degree calculation on data subjected to semantic alignment to obtain an associated coupling degree score; the task retrieval module is used for carrying out target retrieval from the task candidate pool according to the association coupling degree score, and automatically distributing the execution starting time and the load quota of each task by combining the component load limit value field of the physical source item device to obtain a multi-target initial execution sequence; And the dynamic re-indexing module is used for synchronously storing the task state snapshot comprising the accumulated output count and the instantaneous environment parameter when the execution flow interruption is monitored, extracting the residual evolution rate field of the task main body, coupling the real-time availability characteristic of the source item device, carrying out secondary semantic deduction, and obtaining the dynamically corrected re-indexing task sequence on the premise of ensuring that the component load limit value is not triggered.

Description

Associated index construction method and system for multidimensional heterogeneous data Technical Field The invention relates to the technical field of electric digital data processing, in particular to a method and a system for constructing an associated index of multidimensional heterogeneous data. Background When heterogeneous data with strong time-varying characteristics and complex topological constraints are processed by the existing task scheduling scheme, a static indexing mechanism is generally adopted, and dynamic output fluctuation of a physical source item is difficult to match in real time, so that the system execution efficiency is low. Meanwhile, the existing interrupt recovery logic is mostly dependent on simple check point restarting, so that the instantaneous load risk of a hardware component at the moment of interrupt and the attenuation compensation of a main body active period are ignored, and the hardware overload damage or task failure is easily caused. The method aims to solve the technical problem that the execution efficiency compensation and the hardware intrinsic safety dynamic balance are difficult to realize under the constraint of the execution flow interruption and the hardware load of the multidimensional heterogeneous task. Therefore, a method and a system for constructing the association index of multidimensional heterogeneous data are provided. Disclosure of Invention The invention aims to provide a method and a system for constructing an associated index of multidimensional heterogeneous data, which realize dynamic re-indexing and load reconstruction meeting hardware intrinsic safety constraint by constructing a task state snapshot containing multidimensional environmental characteristics at the moment of interrupt triggering. In order to achieve the above purpose, the present invention provides the following technical solutions: a method for constructing an associated index of multidimensional heterogeneous data comprises the following steps: Collecting an operation state field of a physical source item device, a time-varying evolution field of a task main body and a topology constraint field of an execution space, and carrying out structural encapsulation and normalization processing to obtain a heterogeneous feature vector object; The method comprises the steps of identifying a target response window of a task main body in the heterogeneous feature vector object, carrying out semantic alignment on an expected output intensity envelope curve of a physical source item device on a unified time axis, carrying out time sequence overlapping degree calculation on data subjected to semantic alignment, and obtaining an associated coupling degree score; performing target retrieval from a task candidate pool according to the associated coupling degree score, and automatically distributing the execution starting time and load quota of each task by combining with the component load limit value field of the physical source item device to obtain a multi-target initial execution sequence; and when the execution flow interruption is monitored, synchronously storing a task state snapshot comprising the accumulated output count and the instantaneous environment parameter, and calling a residual evolution rate field of a task main body and coupling the real-time availability characteristic of a source item device to carry out secondary semantic deduction, thereby obtaining a dynamically corrected re-index task sequence on the premise of ensuring that the component load limit value is not triggered. Preferably, the collecting the operation status field of the physical source item device specifically includes: The method comprises the steps of recording a dynamic scanning track field of an energy beam on the surface of a conversion medium, constructing a space-time evolution model representing the overlapping process of energy deposition and thermal diffusion by utilizing a preset thermal conduction response matrix dynamically distributed along with space coordinates, generating a real-time thermal load mapping field of a physical source item device based on the space-time evolution model, predicting an instantaneous loss rate trend of the conversion medium material by utilizing nonlinear associated data of historical accumulated energy load and environmental temperature rise, predicting the reduction amount of the medium thickness according to the instantaneous loss rate trend, and feedforward calculation of a predicted offset characteristic field of a source item output energy spectrum. Preferably, the obtaining the heterogeneous feature vector object specifically includes: The method comprises the steps of extracting real-time load intensity and energy spectrum offset parameters in an operation state field of a physical source item device, mapping the parameters to a preset first numerical dimension space, identifying the activity attenuation rate and staged completion degree in a time-