Search

CN-121996987-A - Data acquisition optimization method and system for Internet of things equipment based on deep learning

CN121996987ACN 121996987 ACN121996987 ACN 121996987ACN-121996987-A

Abstract

The invention relates to the technical field of intelligent data acquisition, in particular to an Internet of things equipment data acquisition optimization method and system based on deep learning, comprising the steps of acquiring multidimensional real-time operation data of Internet of things equipment, preprocessing the real-time operation data based on equipment identification, and obtaining standard operation data; the method comprises the steps of carrying out parameter importance analysis on standard operation data based on a deep learning algorithm to obtain parameter acquisition priority, carrying out dynamic strategy adjustment and verification on the parameter acquisition priority based on a working condition change rule to obtain dynamic acquisition strategy data, extracting working condition switching and parameter abnormal event sets from the standard operation data, and carrying out strategy iteration optimization and effect verification on the parameter abnormal event sets based on the dynamic acquisition strategy data to obtain strategy iteration data. According to the scheme, the accuracy and the effectiveness of equipment data acquisition are improved through a dynamic acquisition scheme.

Inventors

  • LI XIAOCHUN

Assignees

  • 杭州友成科技有限公司

Dates

Publication Date
20260508
Application Date
20260407

Claims (10)

  1. 1. The data acquisition optimization method for the Internet of things equipment based on deep learning is characterized by comprising the following steps of: collecting multidimensional real-time operation data of the Internet of things equipment, and preprocessing the real-time operation data based on equipment identification to obtain standard operation data; Carrying out parameter importance analysis on the standard operation data based on a deep learning algorithm to obtain parameter acquisition priority; Carrying out dynamic strategy adjustment and verification on the parameter acquisition priority based on the working condition change rule to obtain dynamic acquisition strategy data; Extracting working condition switching and parameter abnormal event sets from standard operation data, and carrying out strategy iteration optimization and effect verification on the parameter abnormal event sets based on dynamic acquisition strategy data to obtain strategy iteration data; And generating a visual interface corresponding to the dynamic acquisition strategy data, the strategy iteration data and the real-time operation data based on the system interaction logic, and carrying out linkage display on the visual interface.
  2. 2. The data acquisition optimization method for the internet of things equipment based on deep learning according to claim 1, wherein the specific acquisition process of the multidimensional real-time operation data is as follows: the method comprises the steps of performing multidimensional data acquisition on the running state and core parameters of the Internet of things equipment to obtain original acquired data; Performing equipment identification association and data encapsulation on the original acquired data based on equipment service logic to obtain encapsulated operation data; uploading the packaging operation data to a preset central data platform based on a standardized interface of the Internet of things system; Receiving a real-time operation message of a central data platform in a message subscription mode, and carrying out message analysis on the real-time operation message to obtain analysis operation data; and carrying out equipment operation context association on the analysis operation data to obtain multidimensional real-time operation data.
  3. 3. The data acquisition optimization method for the internet of things equipment based on deep learning according to claim 2, wherein the specific obtaining process of the standard operation data is as follows: classifying equipment and manufacturers of the multidimensional real-time operation data, and adding equipment labels for the multidimensional real-time operation data based on the classification result; Carrying out integrity check on the multidimensional real-time operation data based on the equipment tag, and carrying out data deduplication on the real-time operation data after the integrity check based on the equipment identifier and a data time difference threshold value to obtain deduplication operation data; The continuous parameter missing values in the duplication removal operation data are complemented by using a linear interpolation method, and the discrete parameter missing values are complemented by using a mode complementation method to obtain the complemented operation data; based on 3 Carrying out abnormal value detection on the complement operation data in principle to obtain abnormal data, and replacing the abnormal data by using the average value of the normal data in the same time period to obtain denoising operation data; And uniformly reserving two decimal places for all numerical parameters in the denoising operation data by adopting a rounding method, and carrying out data structure standardization operation on the denoising operation data based on a preset field template to obtain standard operation data.
  4. 4. The data acquisition optimization method for the internet of things equipment based on deep learning according to claim 3, wherein the specific acquisition process of the parameter acquisition priority is as follows: Extracting parameter operation data sets corresponding to all the devices from the standard operation data by taking the device identification as a main key, and extracting parameter time sequence events from all the parameter operation data sets to obtain a parameter event group set; Sequencing the parameter event group set according to the time sequence to obtain a parameter event sequence set, and calculating the correlation coefficient of each parameter and the service index based on the service scene core target; Counting the update frequency of each parameter in unit time under normal production working conditions, and obtaining the update frequency duty ratio of the parameters through linear normalization processing; Calculating the weighted value of the frequency of the update of the parameters based on the correlation coefficient to obtain a parameter importance score, and classifying the parameter importance score by adopting a K-means clustering algorithm to obtain a core parameter and non-core parameter classification result; And distributing acquisition priorities for different types of parameters according to the service scene requirements, and generating parameter acquisition priority data.
  5. 5. The method for optimizing data acquisition of the internet of things device based on deep learning according to claim 4, wherein the specific acquisition process of the dynamic acquisition strategy data is as follows: extracting working condition switching events from each parameter operation data set to obtain a working condition event set; sequencing the working condition event group set according to the time sequence to obtain a working condition event sequence set, and calculating the change frequency and fluctuation amplitude of each parameter under different working conditions; determining a frequency correction coefficient corresponding to the working condition based on the change frequency and the fluctuation amplitude, and calculating the adjusted acquisition frequency of each parameter under different working conditions by combining a frequency adjustment coefficient corresponding to the parameter acquisition priority; Selecting a plurality of devices of different manufacturers as test point devices, collecting real-time data according to the regulated collection frequency, and calculating the effective rate and transmission delay of the collected data; and checking the acquisition strategy based on a preset effective rate threshold and a transmission delay threshold, and fine-tuning the frequency correction coefficient and retrying points when the threshold requirement is not met until the requirement is met, so as to obtain dynamic acquisition strategy data.
  6. 6. The method for optimizing data collection of internet of things equipment based on deep learning according to claim 5, wherein the specific obtaining process of the strategy iteration data is as follows: carrying out multidimensional time sequence feature extraction on the standard operation data based on the dynamic acquisition strategy data and the parameter abnormal event set to obtain parameter time sequence features; Carrying out graph structure modeling and abnormal context modeling on the standard operation data based on the parameter time sequence characteristics and the parameter abnormal event set to obtain a parameter association graph structure; Taking the parameter anomaly event set as an anomaly tag, taking the anomaly tag as a supervision signal, and training the LSTM neural network model to obtain a parameter anomaly analysis model; forward propagation is carried out on the parameter association graph structure based on the parameter anomaly analysis model, so that anomaly contribution weights are obtained; And carrying out iterative adjustment on the parameter acquisition strategy based on the abnormal contribution weight, and carrying out validity verification on the adjusted strategy to obtain strategy iterative data.
  7. 7. The method for optimizing data collection of internet of things equipment based on deep learning according to claim 6, wherein the strategy iterative optimization process comprises a strategy version management and failure rollback mechanism, and the specific obtaining process is as follows: Identifying each version acquisition strategy by adopting a semantical version number, and recording the creation time, the modification content, the responsible person and the effective equipment range of each version strategy to form a complete version iteration record; the collection strategy which is currently in effect is automatically backed up before strategy iteration and is used as a rollback standby version; if the iterative strategy still does not meet the preset index after the continuous acquisition and verification for a plurality of times, triggering an automatic rollback mechanism, and immediately switching to the original strategy version of the backup to ensure that the data acquisition service is not interrupted; after the rollback operation is completed, iteration failure reasons, rollback time and influence equipment information are recorded into a system iteration log, and an alarm notification is pushed to operation and maintenance personnel for manual intervention and investigation.
  8. 8. The method for optimizing data acquisition of the internet of things equipment based on deep learning according to claim 7, wherein the strategy iterative optimization process further comprises an equipment offline data complement and safety protection mechanism, and the specific acquisition process is as follows: Automatically starting a local cache function when the equipment is offline, and collecting and storing core parameter data and corresponding time stamps according to a preset frequency; after the equipment is on line again, the buffer data is uploaded in sequence increment according to the time stamp, and the normal acquisition rhythm is recovered after the uploading is finished; If the offline time exceeds the threshold value, adopting a fragmentation uploading mechanism to avoid data transmission congestion, and carrying out standardized processing and validity check on the supplementary data again; The data transmission process adopts TLS1.3 encryption protocol, the data storage adopts AES-256 encryption algorithm, the system adopts RBAC authority control model to distinguish the operation authorities of the manager, the operation and maintenance personnel and the common user, and the key strategy adjustment and the data query operation need to carry out secondary identity verification and record operation logs.
  9. 9. The data acquisition optimization system of the Internet of things equipment based on deep learning is characterized in that the system is used for executing the data acquisition optimization method of the Internet of things equipment based on deep learning, which is described in any one of claims 1-8.
  10. 10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program, the computer program being executed by a processor to implement a deep learning based data acquisition optimization method of an internet of things device according to any one of the preceding claims 1-8.

Description

Data acquisition optimization method and system for Internet of things equipment based on deep learning Technical Field The invention relates to the technical field of data intelligent acquisition, in particular to an Internet of things equipment data acquisition optimization method and system based on deep learning. Background At the moment of the rapid development of the industrial Internet of things, the Internet of things equipment is widely applied to various links of industrial production, and the acquisition and analysis of multi-dimensional operation data of the equipment become core means for guaranteeing production stability and improving operation and maintenance efficiency. Especially in the scenes of precision manufacturing equipment such as a forming machine, the real-time performance and accuracy of equipment operation parameters (such as injection pressure, mold temperature, cycle period and the like) directly influence the forming quality and the production efficiency of products. Therefore, the accuracy and pertinence of data acquisition are extremely high. The existing data acquisition technology of the equipment of the Internet of things mostly adopts a fixed frequency acquisition mode, and the difference of parameter change rules under different working conditions of the equipment is not fully considered, so that obvious limitations exist. When data acquisition is carried out, the non-differential acquisition of all parameters leads to insufficient acquisition resources of core parameters and redundant acquisition of non-core parameters, so that network bandwidth and storage resources are wasted, and meanwhile, the data processing pressure is increased. Moreover, the adopted acquisition strategy lacks dynamic adjustment capability, can not adapt to the switching of different working conditions such as equipment production/standby, and is difficult to respond to sudden scenes such as abnormal fluctuation of parameters, so that the effectiveness and timeliness of data acquisition are insufficient, and reliable data support can not be provided for the equipment operation and maintenance and production optimization. In addition, the prior art has the problems of non-uniform format, inaccurate abnormal value filtering and the like in the data preprocessing link, the deviation of the subsequent data analysis result is easy to cause, a scientific quantization model is lacking in parameter importance assessment, the acquisition priority is subjectively and randomly divided, and the practical value of the acquired data is further reduced. Disclosure of Invention According to the method, the device operation data are subjected to feature extraction and strategy optimization based on deep learning, so that a dynamic acquisition scheme is generated, and the accuracy and the effectiveness of device data acquisition are improved. The technical scheme provided by the invention is that the data acquisition optimization method of the Internet of things equipment based on deep learning comprises the following steps: collecting multidimensional real-time operation data of the Internet of things equipment, and preprocessing the real-time operation data based on equipment identification to obtain standard operation data; Carrying out parameter importance analysis on the standard operation data based on a deep learning algorithm to obtain parameter acquisition priority; Carrying out dynamic strategy adjustment and verification on the parameter acquisition priority based on the working condition change rule to obtain dynamic acquisition strategy data; Extracting working condition switching and parameter abnormal event sets from standard operation data, and carrying out strategy iteration optimization and effect verification on the parameter abnormal event sets based on dynamic acquisition strategy data to obtain strategy iteration data; And generating a visual interface corresponding to the dynamic acquisition strategy data, the strategy iteration data and the real-time operation data based on the system interaction logic, and carrying out linkage display on the visual interface. Preferably, the specific process of obtaining the multidimensional real-time operation data is as follows: the method comprises the steps of performing multidimensional data acquisition on the running state and core parameters of the Internet of things equipment to obtain original acquired data; Performing equipment identification association and data encapsulation on the original acquired data based on equipment service logic to obtain encapsulated operation data; uploading the packaging operation data to a preset central data platform based on a standardized interface of the Internet of things system; Receiving a real-time operation message of a central data platform in a message subscription mode, and carrying out message analysis on the real-time operation message to obtain analysis operation data; and carrying out equipment oper