Search

CN-121434274-B - Big data analysis management method and system for machine manufacturing production data

CN121434274BCN 121434274 BCN121434274 BCN 121434274BCN-121434274-B

Abstract

The invention discloses a big data analysis management method and a big data analysis management system for machine manufacturing production data, and relates to the technical field of machine manufacturing production data management. The method comprises the steps of firstly collecting multi-source production data, then carrying out standardized pretreatment on the multi-source production data to obtain a standardized data sequence, and carrying out quantitative analysis on the process time sequence causal strength of dependent variable data and fruit variable data in the standardized data sequence based on an industrial time sequence causal entropy algorithm, and outputting causal weights as analysis results in step 200. According to the invention, by constructing a causal-relation-based structured index system and self-learning adaptive process time sequence semantic library and combining an industrial time sequence causal entropy algorithm and real-time pre-judging retrieval, the problems of single index dimension and fixed system parameters of traditional production data are solved, and the high-efficiency positioning of causal-relation data, the dynamic adaptation process change of core parameters and the rapid response of production abnormality are realized, so that support is provided for the accurate management and high-efficiency processing of the production data.

Inventors

  • WANG YINGRUI
  • WANG WEIZHEN
  • MIAO YANWEI
  • Shang Xinghua
  • WU GANG

Assignees

  • 中科力祥电气(山东)有限公司

Dates

Publication Date
20260508
Application Date
20251104

Claims (8)

  1. 1. The big data analysis and management method for the machine manufacturing production data is characterized by comprising the following steps of: Step 100, firstly, multi-source production data are collected, and then standardized pretreatment is carried out on the multi-source production data to obtain a standardized data sequence, so that the whole flow control of the data is realized; step 200, based on an industrial time sequence causal entropy algorithm, carrying out quantitative analysis on the process time sequence causal strength of dependent variable data and fruit variable data in the standardized data sequence, and outputting causal weights as analysis results to provide quantitative basis for subsequent management; Step 300, based on the causal weight, building a four-dimensional dynamic index key integrating the process attribute, the equipment running state, the data time sequence range and the causal weight, and building a time sequence causal tree index according to causal association strength layering to realize index system structured management; step 400, based on the historical data, analysis result and real-time feedback data of the standardized data sequence, constructing and self-learning updating process time sequence semantic library, synchronously calibrating algorithm core parameters, and guaranteeing dynamic adaptation of analysis and management; step 500, monitoring the standardized data sequence in real time, acquiring real-time abnormal data triggering a fluctuation threshold value, triggering pre-judging search based on an updated process time sequence semantic library and the time sequence factor fruit tree index, loading corresponding associated data and outputting the corresponding associated data, and realizing quick response management of the abnormal data; wherein, the step 200 further comprises: step 210, time sequence association degree quantization, calculating dependent variable data sequence Sequence of data on fruit variables Wave cooperative degree of association of (1) Wherein In order to fluctuate the degree of cooperative association, Is of lag time and meets Capturing the association relation of abnormal fluctuation; Step 220, process constraint correction, introducing a process causal constraint factor To the degree of association of the fluctuation collaboration Correcting to obtain the process adaptation association degree Distinguishing statistical association from process cause and effect; step 230, calculating industrial time sequence causal entropy, and adapting association degree based on the process Calculating causal entropy Quantifying causal association certainty; step 240, causal weight mapping, mapping the causal entropy Conversion to the causal weights The total causal weight is 1, and a priority basis is provided for index hierarchical management; In the step 210, the degree of association of the collaboration property is fluctuated The calculation formula of (2) is as follows: ; in the formula, As the start time of the timing window, For a sequence of dependent variable data The average value over the time-series window, For fruit variable data sequences The average value over the time-series window, For the dependent variable data fluctuation threshold, For the fluctuation threshold of the fruit variable data Or (b) In the fluctuation cooperativity association degree calculation formula molecule, corresponding time points Product term of (2) Taking 0; in the step 220, the process adaptation association degree The calculation formula of (2) is as follows: ; in the formula, In order to adapt the coefficients of the process, As a sign function only When it is, go to step 230; In the step 230, causal entropy The calculation formula of (2) is as follows: ; in the formula, Is a fluctuation smoothing factor and , Is a fluctuation amplification factor and , As a function of the index of the values, , In order to be a strong cause and effect, Is a weak causal agent which is used for the treatment of the heart disease, Is non-causal; In the step 240, causal weights The calculation formula of (2) is as follows: ; in the formula, For data sequences of the dependent variables There is a sequence of all fruit variable data that are time-series correlated, Is the number of fruit variable data sequences.
  2. 2. The method of big data analysis management of machine-made production data according to claim 1, wherein the step 100 further comprises: Step 110, collecting equipment operation data, process parameter data and quality inspection data to form an original data set; Step 120, preprocessing operation of noise filtering and missing value complement is sequentially carried out on the original data set, and invalid data is removed; step 130, performing standardization processing on the preprocessed original data, classifying according to data types and mapping to a preset numerical value interval to unify the data sizes of the same type to obtain a dependent variable data sequence With at least one fruit variable data sequence Wherein The data of the dependent variable is represented, The data representing the fruit variable is displayed, Is a time stamp, and the time stamps of the dependent variable data sequence and the fruit variable data sequence are consistent, the time span and the process time window are same The two kinds of materials are matched with each other, The effective time range of the causal relationship of the characterization process.
  3. 3. The method according to claim 2, wherein in the step 130, the normalization process is performed by classifying the preprocessed raw data according to a predetermined data type classification rule, each data type corresponds to a unique predetermined value interval, mapping each type of data to a corresponding interval by linear transformation, and ensuring the dependent variable data sequence Sequence of data on fruit variables The time sequence alignment of the sequence and cause and effect analysis of the subsequent process is satisfied.
  4. 4. The method for managing big data analysis of machine-made production data according to claim 1, wherein the step 300 further comprises: Step 310, based on process attributes, equipment operating status, data timing ranges and causal weights The four-dimensional dynamic index key is constructed, the index core identification dimension is definitely indexed, and the data time sequence range corresponds to ; Step 320, taking the dependent variable data of strong causal effect as a root node, taking the dependent variable data of strong causal effect as a first-level child node, taking the dependent variable data of weak causal effect as a second-level child node, excluding non-causal effect, constructing the time sequence causal tree index with distinct hierarchy, and realizing the structural management of the associated data, wherein the strong causal effect corresponds to the first-level child node Weak causal correspondence Non-causal correspondence ; The step 400 further includes: step 410, the process timing semantic library stores Causal constraint factor of process Process adaptation coefficient Wave smoothing factor Wave amplification factor Dependent variable data fluctuation threshold Fruit variable data fluctuation threshold When the process parameter is adjusted or the process is newly added, the parameter is automatically calibrated; Step 420, updating the process time sequence semantic library based on the historical accumulated data of the standardized data sequence and the retrieval accuracy data fed back in real time, wherein the self-learning adaptation time is less than or equal to 24 hours, and the suitability of index construction and pre-judging retrieval is ensured.
  5. 5. The method of big data analysis management of machine-made production data of claim 1, wherein the step 500 further comprises: step 510, monitoring the standardized data sequence in real time, wherein the fluctuation threshold of the dependent variable data is The fluctuation threshold of the fruit variable data is Triggering an abnormal signal when any one data in the standardized data sequence exceeds a corresponding fluctuation threshold value, wherein the abnormal signal is associated with the corresponding data exceeding the threshold value; step 520, calling an updated process time sequence semantic library based on dependent variable data corresponding to the abnormal signal, and prejudging associated primary sub-node data and secondary sub-node data from the time sequence dependent variable data index; And 530, loading indexes corresponding to the pre-judging associated data in advance, and synchronously outputting root node data and total associated child node data corresponding to the abnormal signals after receiving the search request, wherein the search response time is less than or equal to 1 second, so that efficient management of the abnormal data is realized.
  6. 6. A big data analysis management system for machine-made production data, comprising: The data acquisition management module is used for acquiring multi-source production data, carrying out standardized pretreatment on the multi-source production data to obtain a standardized data sequence, and realizing the overall flow control of the data; The causal entropy analysis module is used for quantitatively analyzing the process time sequence causal strength of the dependent variable data and the effect variable data in the standardized data sequence based on an industrial time sequence causal entropy algorithm, outputting causal weights as analysis results and providing quantitative basis for subsequent management; The index construction management module is used for constructing four-dimensional dynamic index keys integrating process attributes, equipment running states, data time sequence ranges and causal weights based on the causal weights, constructing time sequence causal tree indexes in a layered mode according to causal association strength, and realizing index system structural management; The self-learning adaptation module is used for constructing and self-learning updating a process time sequence semantic library based on historical data, analysis results and real-time feedback data of the standardized data sequence, synchronously calibrating algorithm core parameters and guaranteeing dynamic adaptation of analysis and management; The pre-judging retrieval management module is used for monitoring the standardized data sequence in real time, acquiring real-time abnormal data triggering a fluctuation threshold value, triggering pre-judging retrieval based on an updated process time sequence semantic library and the time sequence fruit tree index, loading and outputting corresponding associated data, and realizing quick response management of the abnormal data; The monitoring module is used for monitoring the index construction state, the algorithm parameter accuracy, the prejudgement hit rate and the data flow integrity, so that the stable operation of the system is ensured; wherein the causal entropy analysis module is further configured to: time sequence association degree quantization, calculating dependent variable data sequence Sequence of data on fruit variables Wave cooperative degree of association of (1) Wherein In order to fluctuate the degree of cooperative association, Is of lag time and meets Capturing the association relation of abnormal fluctuation; Process constraint correction, introducing a process causal constraint factor To the degree of association of the fluctuation collaboration Correcting to obtain the process adaptation association degree Distinguishing statistical association from process cause and effect; Industrial time sequence causal entropy calculation based on the process adaptation association degree Calculating causal entropy Quantifying causal association certainty; Causal weight mapping, mapping the causal entropy Conversion to the causal weights The total causal weight is 1, and a priority basis is provided for index hierarchical management; the fluctuation cooperativity association degree The calculation formula of (2) is as follows: ; in the formula, As the start time of the timing window, For a sequence of dependent variable data The average value over the time-series window, For fruit variable data sequences The average value over the time-series window, For the dependent variable data fluctuation threshold, For the fluctuation threshold of the fruit variable data Or (b) In the fluctuation cooperativity association degree calculation formula molecule, corresponding time points Product term of (2) Taking 0; the process adaptation association degree The calculation formula of (2) is as follows: ; in the formula, In order to adapt the coefficients of the process, As a sign function only When the method is used, industrial time sequence causal entropy calculation is carried out; The causal entropy The calculation formula of (2) is as follows: ; in the formula, Is a fluctuation smoothing factor and , Is a fluctuation amplification factor and , As a function of the index of the values, In order to be a strong cause and effect, Is a weak causal agent which is used for the treatment of the heart disease, Is non-causal The causal weight The calculation formula of (2) is as follows: ; in the formula, For data sequences of the dependent variables There is a sequence of all fruit variable data that are time-series correlated, Is the number of fruit variable data sequences.
  7. 7. An electronic device, the electronic device comprising: the device comprises a processor and a memory, wherein the memory is in communication connection with the processor; The memory is configured to store executable instructions that are executed by at least one of the processors, the processor configured to execute the executable instructions to implement the large data analysis management method of machine-made production data as claimed in any one of claims 1 to 5.
  8. 8. A computer-readable storage medium, in which a computer program is stored, which when executed by a processor implements the large data analysis management method of machine-made production data according to any one of claims 1 to 5.

Description

Big data analysis management method and system for machine manufacturing production data Technical Field The invention relates to the technical field of machine manufacturing production data management, in particular to a big data analysis management method and system of machine manufacturing production data. Background In the mechanical manufacturing production process, along with the deep fusion of the internet of things technology and intelligent manufacturing, multi-source heterogeneous data such as equipment operation data (such as rotating speed and temperature), process parameter data (such as cutting depth and welding current), quality inspection data (such as dimensional accuracy and qualification rate) and the like can be continuously generated. The data has huge data quantity and high sampling frequency, and has close time sequence causal relation with production quality control and equipment fault investigation, namely enterprises need to quickly trace the root through accurately identifying causal relation among the data when abnormal data occur, so that the stability of production flow is ensured, and unqualified products are reduced, which becomes one of the core demands of industrial data management. Currently, the mainstream production data management system has obvious limitations in data organization and parameter configuration, namely, on one hand, data indexes are constructed by relying on single dimension (such as time stamp and equipment ID), time ordering of data or classification according to equipment can only be realized, the data cannot be combined with back process causal association, related information is required to be screened one by one in massive data when abnormality occurs, key influencing factors are difficult to quickly locate, on the other hand, core parameters (such as time sequence analysis window and data fluctuation threshold value) for supporting data processing and causal analysis in the system are mainly statically set, manual reconfiguration according to process adjustment (such as process material replacement and process parameter optimization) is required, adjustment hysteresis exists, and parameter setting deviation is easy to be caused due to manual experience difference. The mode directly causes that the existing system cannot adapt to process change efficiently, and is difficult to meet the actual requirement of quick response of production abnormality, and becomes a key bottleneck for restricting industrial data value mining. Disclosure of Invention The invention aims to provide a big data analysis management method and system for machine manufacturing production data, so as to solve the problems in the background technology. In order to achieve the above purpose, the invention provides a big data analysis management method of machine manufacturing production data, which comprises the following steps: Step 100, firstly, multi-source production data are collected, and then standardized pretreatment is carried out on the multi-source production data to obtain a standardized data sequence, so that the whole flow control of the data is realized; step 200, based on an industrial time sequence causal entropy algorithm, carrying out quantitative analysis on the process time sequence causal strength of dependent variable data and fruit variable data in the standardized data sequence, and outputting causal weights as analysis results to provide quantitative basis for subsequent management; Step 300, based on the causal weight, building a four-dimensional dynamic index key integrating the process attribute, the equipment running state, the data time sequence range and the causal weight, and building a time sequence causal tree index according to causal association strength layering to realize index system structured management; step 400, based on the historical data, analysis result and real-time feedback data of the standardized data sequence, constructing and self-learning updating process time sequence semantic library, synchronously calibrating algorithm core parameters, and guaranteeing dynamic adaptation of analysis and management; and 500, monitoring the standardized data sequence in real time, acquiring real-time abnormal data triggering a fluctuation threshold value, triggering pre-judging search based on the updated process time sequence semantic library and the time sequence fruit tree index, loading corresponding associated data and outputting the corresponding associated data, and realizing quick response management of the abnormal data. Preferably, the step 100 further includes: Step 110, collecting equipment operation data, process parameter data and quality inspection data to form an original data set; Step 120, preprocessing operation of noise filtering and missing value complement is sequentially carried out on the original data set, and invalid data is removed; step 130, performing standardization processing on the preprocessed original data, c