CN117670139B

CN117670139B - Intelligent detection method and system for PP line quality

Info

Publication number: CN117670139B
Application number: CN202311724961.2A
Authority: CN
Inventors: 申作锋; 李春斌; 朱洪柱; 申政; 汪霞
Original assignee: Shandong Juli New Material Technology Co ltd; Rizhao Huifeng Nets Co ltd
Current assignee: Shandong Juli New Material Technology Co ltd; Rizhao Huifeng Nets Co ltd
Priority date: 2023-12-15
Filing date: 2023-12-15
Publication date: 2024-05-14
Anticipated expiration: 2043-12-15
Also published as: CN117670139A

Abstract

The disclosure provides an intelligent detection method and system for PP line quality, and relates to the technical field of quality detection, wherein the method comprises the following steps: obtaining M PP line production nodes; constructing a PP line quality detection factor distribution model; obtaining a PP line quality detection evaluation model; performing quality detection on the real-time PP line production node to obtain real-time node quality detection data; obtaining a quality detection evaluation layer of a matching node corresponding to the real-time PP line production node; obtaining a plurality of node factor quality detection evaluation coefficients; obtaining a comprehensive quality detection evaluation coefficient; the method and the device can solve the technical problem that the quality of the PP wire is poor due to lower efficiency and accuracy of the quality detection of the PP wire in the prior art, achieve the aim of improving the efficiency and accuracy of the quality detection of the PP wire, and achieve the technical effect of improving the production quality of the PP wire.

Description

Intelligent detection method and system for PP line quality

Technical Field

The disclosure relates to the technical field of quality detection, in particular to an intelligent detection method and system for PP wire quality.

Background

The PP wire is widely applied to the fields of automobile manufacturing, household appliance manufacturing, communication equipment manufacturing and the like. The manufacturing process of the PP wire comprises the procedures of extrusion, stretching, cutting and the like, and strict quality control is required in the production process of the procedures so as to ensure the quality and performance of the product. At present, the existing PP wire quality detection technology has defects in terms of detection precision, and cannot accurately detect tiny defects or problems. Causing some potential quality problems to be missed, thereby affecting product quality and safety. Meanwhile, the PP line quality detection process still needs to be manually participated, the degree of automation is low, human errors are introduced, and the detection precision is reduced. Accordingly, there is a need for a method to solve the above-mentioned problems.

In summary, in the prior art, the PP line quality detection efficiency and accuracy are low, which results in a technical problem of poor PP line production quality.

Disclosure of Invention

The disclosure provides an intelligent detection method and system for quality of a PP (Polypropylene) wire, which are used for solving the technical problem that the production quality of the PP wire is poor due to lower efficiency and accuracy of quality detection of the PP wire in the prior art.

According to a first aspect of the present disclosure, there is provided an intelligent detection method for PP line quality, including: carrying out production node identification according to the PP line production process characteristic data to obtain M PP line production nodes, wherein M is a positive integer greater than 1; constructing a PP line production chain based on the M PP line production nodes; digging quality detection factors according to the PP line production chain, and constructing a PP line quality detection factor distribution model; deep learning is carried out based on the PP line quality detection factor distribution model, and a PP line quality detection evaluation model is obtained; reading a real-time PP line production node, and carrying out quality detection on the real-time PP line production node according to the PP line quality detection factor distribution model to obtain real-time node quality detection data; positioning the PP line quality detection evaluation model based on the real-time PP line production node to obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, wherein the matching node quality detection evaluation layer comprises a plurality of matching node quality detection evaluation branches and matching node quality evaluation excitation branches; inputting the real-time node quality detection data into the plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients; inputting the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient; and generating a PP line quality detection report according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients and the comprehensive quality detection evaluation coefficient.

According to a second aspect of the present disclosure, there is provided an intelligent detection system for PP line quality, comprising: the PP line production node acquisition module is used for carrying out production node identification according to the PP line production process characteristic data to acquire M PP line production nodes, wherein M is a positive integer greater than 1; the PP line production chain construction module is used for constructing a PP line production chain based on the M PP line production nodes; the PP line quality detection factor distribution model construction module is used for carrying out quality detection factor mining according to the PP line production chain to construct a PP line quality detection factor distribution model; the PP line quality detection evaluation model obtaining module is used for carrying out deep learning based on the PP line quality detection factor distribution model to obtain a PP line quality detection evaluation model; the real-time node quality detection data acquisition module is used for reading real-time PP line production nodes and carrying out quality detection on the real-time PP line production nodes according to the PP line quality detection factor distribution model to acquire real-time node quality detection data; the matching node quality detection evaluation layer obtaining module is used for positioning the PP line quality detection evaluation model based on the real-time PP line production node to obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, wherein the matching node quality detection evaluation layer comprises a plurality of matching node quality detection evaluation branches and matching node quality evaluation excitation branches; the node factor quality detection evaluation coefficient obtaining module is used for inputting the real-time node quality detection data into the plurality of matched node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients; the comprehensive quality detection evaluation coefficient obtaining module is used for inputting the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient; and the PP line quality detection report acquisition module is used for generating a PP line quality detection report according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients and the comprehensive quality detection evaluation coefficient.

One or more technical solutions provided in the present disclosure have at least the following technical effects or advantages: according to the method, M PP line production nodes are obtained by carrying out production node identification according to the PP line production process characteristic data, wherein M is a positive integer greater than 1; constructing a PP line production chain based on the M PP line production nodes; digging quality detection factors according to the PP line production chain, and constructing a PP line quality detection factor distribution model; deep learning is carried out based on the PP line quality detection factor distribution model, and a PP line quality detection evaluation model is obtained; reading a real-time PP line production node, and carrying out quality detection on the real-time PP line production node according to the PP line quality detection factor distribution model to obtain real-time node quality detection data; positioning the PP line quality detection evaluation model based on the real-time PP line production node to obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, wherein the matching node quality detection evaluation layer comprises a plurality of matching node quality detection evaluation branches and matching node quality evaluation excitation branches; inputting the real-time node quality detection data into the plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients; inputting the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient; according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients and the comprehensive quality detection evaluation coefficients, a PP line quality detection report is generated, the technical problem that the quality of the PP line is poor due to low efficiency and accuracy of the quality detection of the PP line in the prior art is solved, the aim of improving the efficiency and accuracy of the quality detection of the PP line is fulfilled, and the technical effect of improving the production quality of the PP line is achieved.

It should be understood that the description of this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

For a clearer description of the present disclosure or of the prior art, the drawings used in the description of the embodiments or of the prior art will be briefly described, it being obvious that the drawings in the description below are only exemplary and that other drawings may be obtained, without inventive effort, by a person skilled in the art, from the provided drawings.

Fig. 1 is a schematic flow chart of an intelligent PP line quality detection method according to an embodiment of the present disclosure;

Fig. 2 is a schematic structural diagram of an intelligent PP line quality detection system according to an embodiment of the present disclosure.

Reference numerals illustrate: the system comprises a PP line production node obtaining module 11, a PP line production chain constructing module 12, a PP line quality detection factor distribution model constructing module 13, a PP line quality detection evaluation model obtaining module 14, a real-time node quality detection data obtaining module 15, a matched node quality detection evaluation layer obtaining module 16, a node factor quality detection evaluation coefficient obtaining module 17, a comprehensive quality detection evaluation coefficient obtaining module 18 and a PP line quality detection report obtaining module 19.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

Embodiment one: an intelligent detection method for PP line quality provided by an embodiment of the present disclosure is described with reference to fig. 1, and the method includes:

carrying out production node identification according to the PP line production process characteristic data to obtain M PP line production nodes, wherein M is a positive integer greater than 1;

Specifically, the PP line production process retrieval is carried out based on the big data, and the PP line production process characteristic data is obtained. For example, PP line production process characterization data includes slurry process, bulk process, or gas phase process. Further, M PP line production nodes are obtained according to the difference of production flows in the PP line production process characteristic data. For example, the M PP line production nodes include raw material processing, extrusion, molding, cooling, and the like. Wherein M is a positive integer greater than 1, since it comprises at least two different production nodes.

Constructing a PP line production chain based on the M PP line production nodes;

Specifically, the production chain construction is carried out on M PP line production nodes according to the production order constraint of the PP line, and the PP line production node distribution chain is obtained. Wherein, the production order constraint of the PP line is the order requirement of the production flow. For example, PP line production order constraints are the production flow sequence requirements for raw material handling, extrusion, forming, and cooling. Accordingly, PP line production order constraints are retrieved based on big data. Further, node value degree identification is carried out on M PP line production nodes, and duty ratio calculation is carried out to obtain M node value proportion coefficients. And marking the M node value proportion coefficients on the PP line production node distribution chain to obtain the PP line production chain.

Digging quality detection factors according to the PP line production chain, and constructing a PP line quality detection factor distribution model;

Specifically, the production nodes in the PP line production chain and the corresponding node value proportion coefficients are extracted. And configuring corresponding node factor mining calculation forces based on the node value proportion coefficients. When the node value is higher than the coefficient of gravity, the greater the mining calculation force of the configured node factors is, and conversely, the smaller the mining calculation force of the configured node factors is. Further, quality detection event backtracking is carried out on the PP line production nodes based on node factor mining calculation force, and node quality detection event records are obtained. When the node factor mining calculation force is larger, more data are acquired in the quality detection process, and conversely, less data are acquired. Further, after backtracking, obtaining an event corresponding to the data as a node quality detection event record. Further, according to the node quality detection event record, a plurality of node sample quality detection factors corresponding to the PP line production nodes are extracted. Further, repeated factor cleaning and decoupling factor cleaning are carried out on the plurality of node sample quality detection factors, a plurality of node quality detection factors are generated, chain storage is carried out on the plurality of node quality detection factors, and a PP line quality detection factor distribution model is obtained.

Deep learning is carried out based on the PP line quality detection factor distribution model, and a PP line quality detection evaluation model is obtained;

Specifically, deep learning is performed according to a plurality of node quality detection factors extracted from a node quality detection factor distribution sub-model, a plurality of factor quality detection estimators are obtained, a node quality detection evaluation branch corresponding to each node quality detection factor is obtained through weighting calculation of the plurality of factor quality detection estimators, and then a plurality of node quality detection evaluation branches are combined to obtain a plurality of node quality detection evaluation branches. And carrying out quality influence degree identification based on the plurality of node quality detection factors to obtain a plurality of factor quality influence degrees. The quality influence degree represents the importance degree of the node quality detection factor in quality detection. Further, duty ratio calculation is performed based on the plurality of factor quality influence degrees, and a plurality of factor quality excitation coefficients are obtained. And respectively carrying out duty ratio calculation of all the factor quality influence degrees on each factor quality influence degree to obtain a plurality of factor quality excitation coefficients. Further, according to the multiple factor quality excitation coefficients, weighting calculation is carried out on multiple node quality detection evaluation branches corresponding to the multiple factor quality influence degrees, and the weighting calculation results are respectively added to the corresponding node quality evaluation excitation branches, so that the node quality evaluation excitation branches are constructed. Further, M node quality detection evaluation branches and node quality evaluation excitation branches are connected to generate M node quality detection evaluation layers. And adding the combined M node quality detection evaluation layers to the PP line quality detection evaluation model.

Reading a real-time PP line production node, and carrying out quality detection on the real-time PP line production node according to the PP line quality detection factor distribution model to obtain real-time node quality detection data;

Specifically, the real-time PP line production node is read by a method for acquiring the working progress. Inputting the real-time PP line production nodes into a PP line quality detection factor distribution model, carrying out production node matching with M node quality detection evaluation layers in the PP line quality detection factor distribution model, and carrying out quality detection on the real-time PP line production nodes to obtain real-time node quality detection data.

Positioning the PP line quality detection evaluation model based on the real-time PP line production node to obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, wherein the matching node quality detection evaluation layer comprises a plurality of matching node quality detection evaluation branches and matching node quality evaluation excitation branches;

specifically, a matching node quality detection evaluation layer corresponding to the real-time PP line production node is obtained based on a node quality detection evaluation layer corresponding to the PP line quality detection evaluation model, wherein the matching node quality detection evaluation layer comprises a plurality of corresponding matching node quality detection evaluation branches and matching node quality evaluation excitation branches.

Inputting the real-time node quality detection data into the plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients;

Specifically, the plurality of matching node quality detection evaluation branches includes a plurality of matching factor quality detection estimators and a matching factor quality detection evaluation output exciter. And carrying out feature recognition on the real-time node quality detection data according to a plurality of matching node quality detection factors corresponding to a plurality of matching factor quality detection estimators in the matching node quality detection and estimation branch to obtain matching corresponding matching node quality detection factors. And carrying out feature recognition on the quality detection factors of the matching nodes and the real-time quality detection data of the nodes to obtain quality detection data of the matching factors. Further, the matching factor quality detection data is input to a plurality of matching factor quality detection estimators, generating a plurality of matching factor quality detection estimation coefficients. Further, a plurality of matching factor quality detection evaluation coefficients are input into a matching factor quality detection evaluation output exciter to generate a plurality of node factor quality detection evaluation coefficients, and the node factor quality detection evaluation coefficients are used for performing duty ratio calculation on the matching factor quality detection evaluation coefficients so as to finish quality detection.

Inputting the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient;

specifically, a plurality of node factor quality detection evaluation coefficients are input into a matched node quality evaluation excitation branch, and the node factor quality detection evaluation coefficients are subjected to duty ratio calculation through the node factor quality detection evaluation coefficients in the matched node quality evaluation excitation branch, so that a comprehensive quality detection evaluation coefficient is obtained.

And generating a PP line quality detection report according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients and the comprehensive quality detection evaluation coefficient.

Specifically, weighting calculation is carried out on a plurality of node factor quality detection evaluation coefficients according to the comprehensive quality detection evaluation coefficients, quality detection weights of the node factors in the real-time node quality detection data are obtained according to the obtained weighting calculation results, data emphasis distribution based on the weights is carried out on the real-time node quality detection data, and a PP line quality detection report is generated.

The technical problem that the quality of the PP wire is poor due to low efficiency and accuracy of the quality detection of the PP wire in the prior art can be solved through the embodiment, the aim of improving the efficiency and accuracy of the quality detection of the PP wire is achieved, and the technical effect of improving the production quality of the PP wire is achieved.

The method provided by the embodiment of the disclosure further comprises the following steps:

Extracting PP line production order constraint according to the PP line production process characteristic data;

Based on the PP line production order constraint, building a PP line production node distribution chain according to the M PP line production nodes;

Carrying out node value recognition based on the M PP line production nodes to obtain M node value degrees;

Performing duty ratio calculation based on the M node value degrees to obtain M node value proportion coefficients;

And identifying the PP line production node distribution chain according to the M node value proportion coefficients, and generating the PP line production chain.

Specifically, the PP line production order constraint is extracted according to the PP line production process characteristic data. Wherein, the PP line production order constraint refers to the order requirement of the production flow. For example, PP line production order constraints are the production flow order requirements for the ready nodes, aggregate nodes, spin nodes, draw nodes, etc.

Further, the M PP line production nodes are subjected to serialization processing according to the PP line production order constraint, and a serialization processing result is obtained and is used as a PP line production node distribution chain.

Further, node value degree identification is carried out based on M PP line production nodes, and M node value degrees are respectively obtained. The node value degree refers to the importance degree of each PP line production node. For example, the method for identifying the value of the node may be to identify the value by the repeatability of the node, or identify the value by the importance degree of the node production result of the node, for example, when the quality influence degree of the cooling result of the cooling node on the PP line is higher, the importance degree of the cooling node is higher, and then the value of the cooling node is higher.

Further, based on the value of the M nodes, the distribution chain of the M PP line production nodes with the respective ratio to the PP line production nodes is calculated, and the value proportion coefficients of the M nodes are obtained. Wherein, the total value of the PP line production node distribution chain is 1. For example, the cooling node has a node value gravity coefficient of 0.3.

Further, the PP line production node distribution chain is identified according to the M node value proportion coefficients, and the PP line production chain is generated. The PP line production chain is provided with a node value specific gravity coefficient mark. The PP line production chain is constructed based on M PP line production nodes, so that the efficiency of PP line quality detection through the PP line production chain can be improved.

According to the PP line production chain, extracting an m-PP line production node and an m node value specific gravity coefficient corresponding to the m-PP line production node;

Configuring an mth node factor mining calculation force based on the mth node value gravity coefficient;

Performing quality detection event backtracking on the m-PP line production node based on the m node factor mining calculation force to obtain an m node quality detection event record;

extracting a plurality of node sample quality detection factors corresponding to the m-PP line production node according to the m-th node quality detection event record;

Based on the plurality of node sample quality detection factors, constructing an mth node quality detection factor distribution sub-model corresponding to the mth-PP line production node;

and adding the m-th node quality detection factor distribution sub-model to the PP line quality detection factor distribution model.

Specifically, an m-PP line production node and an m-node value specific gravity coefficient corresponding to the m-PP line production node are extracted from a PP line production chain. The m-th PP line production node can be the PP line production node which is the foremost or the last in the PP line production chain, and can also be the random PP line production node in the PP line production chain. Accordingly, the order of the node positions of the m-th PP line production node in the PP line production chain has little influence on the construction of the PP line quality detection factor distribution model in the subsequent step, and extraction is performed only as an example.

Further, an mth node factor mining algorithm is configured based on the mth node value gravity coefficient. The m-th node factors mine the calculation force as the calculation processing calculation force. When the value of the mth node is higher than the gravity coefficient, the greater the mining calculation force of the mth node factor is configured, and conversely, the smaller the mining calculation force is.

Further, backtracking the quality detection event of the production node of the m-PP line based on the m node factor mining calculation force, and obtaining an m node quality detection event record. The quality detection event is traced back to perform quality detection on the corresponding node. When the m-th node factor mining calculation force is larger, more data are acquired in the quality detection process, and conversely, less data are acquired. For example, the chemical composition, physical properties, and appearance quality of the raw material processing node are detected to obtain detection data. Further, after backtracking, obtaining an event corresponding to the data as an m-th node quality detection event record. For example, the mth node quality detection event record includes appearance quality detection events in the raw material processing node.

Further, according to the m-th node quality detection event record, a plurality of node sample quality detection factors corresponding to the m-th PP line production node are extracted. The node sample quality detection factor is an index of quality detection in the node. For example, the node sample quality detection factors include indicators of appearance quality detection in raw material processing nodes, such as uniform color, no bubbles, smooth surface, and large particles.

Further, repeated factor cleaning and decoupling factor cleaning are carried out on the plurality of node sample quality detection factors, a plurality of node quality detection factors are generated, chain storage is carried out on the plurality of node quality detection factors, and an mth node quality detection factor distribution submodel is generated.

Further, adding the M-th node quality detection factor distribution sub-model to the PP line quality detection factor distribution model, obtaining M node quality detection factor distribution sub-models corresponding to M PP line production nodes by a method for obtaining the M-th node quality detection factor distribution sub-model, and adding the M node quality detection factor distribution sub-models to the PP line quality detection factor distribution model. The quality detection factor mining is carried out according to the PP line production chain, and a PP line quality detection factor distribution model is constructed, so that the quality detection efficiency and accuracy can be improved.

traversing the plurality of node sample quality detection factors to perform repeated factor cleaning to obtain a plurality of candidate node quality detection factors;

traversing the quality detection factors of the candidate nodes to identify double-factor decoupling degrees, and obtaining multiple factor decoupling degrees;

performing decoupling factor cleaning on the candidate node quality detection factors based on the factor decoupling degrees to generate node quality detection factors;

and carrying out chain storage on the plurality of node quality detection factors to generate the m-th node quality detection factor distribution sub-model.

Specifically, since the greater the node factor mining computation force, the more the range or content of the data of the acquired node sample quality detection factor, and vice versa, the smaller, there may be duplicate data in the acquired data. For example, the repetition factor is a color index and color index parameter that multiple times acquire the appearance quality detection in the raw material processing node. Further, sequentially accessing a plurality of node sample quality detection factors to perform repeated factor cleaning, wherein the repeated factor cleaning method is to extract a random one of the repeated factors as a candidate node quality detection factor, remove the remaining repeated factors, and sequentially perform repeated factor cleaning on the plurality of node sample quality detection factors, so as to obtain a plurality of candidate node quality detection factors.

Further, the degree of re-coupling identification refers to the degree of repetition identification, and when the degree of re-coupling of the candidate node quality detection factor is higher, the degree of repetition of the candidate node quality detection factor is higher, and conversely, the degree of repetition of the candidate node quality detection factor is lower. And sequentially accessing a plurality of candidate node quality detection factors to perform double-coupling degree identification, wherein the double-coupling degree identification method is to randomly extract the two candidate node quality detection factors to perform the double-coupling degree identification until any two candidate node quality detection factors perform the double-coupling degree identification, and a plurality of factor double-coupling degrees are obtained.

Further, the multiple candidate node quality detection factors are subjected to decoupling factor cleaning based on the multiple factor decoupling degrees, and multiple node quality detection factors are generated. In the process of cleaning the decoupling factor, a decoupling degree threshold is preset, and the decoupling degree threshold is obtained by custom setting according to actual conditions by a person skilled in the art, for example, the decoupling degree threshold is 80%. Comparing the multiple factor re-coupling degrees with a re-coupling degree threshold, and when the factor re-coupling degree is larger than or equal to the re-coupling degree threshold, the repeated degree of the candidate node quality detection factors is higher, and randomly extracting one candidate node quality detection factor from the two candidate node quality detection factors which are larger than or equal to the re-coupling degree threshold and adding the candidate node quality detection factors into the multiple node quality detection factors. When the factor re-coupling degree is smaller than the re-coupling degree threshold value, the repetition degree of the candidate node quality detection factors is lower, and two candidate node quality detection factors smaller than the re-coupling degree threshold value are extracted and added to the plurality of node quality detection factors, so that the plurality of node quality detection factors are generated.

Further, the chain storage is a storage structure for storing data by using a group of arbitrary storage units, wherein the storage units can be continuous or discontinuous. Correspondingly, a plurality of node quality detection factors are stored in a chained mode, and an mth node quality detection factor distribution sub-model is generated.

The m-th node quality detection factor distribution sub-model corresponding to the m-th PP line production node is constructed based on the plurality of node sample quality detection factors, so that the quality detection efficiency can be improved.

extracting a plurality of m node quality detection factors according to the m node quality detection factor distribution sub-model;

Deep learning is carried out according to the m-node quality detection factors, and a plurality of m-node quality detection evaluation branches are obtained;

Performing quality influence degree identification based on the m-node quality detection factors to obtain a plurality of factor quality influence degrees;

Performing duty ratio calculation based on the multiple factor quality influence values to obtain multiple factor quality excitation coefficients;

constructing an mth node quality evaluation excitation branch according to the multiple factor quality excitation coefficients;

connecting the m-th node quality detection evaluation branches with the m-th node quality evaluation excitation branches to generate an m-th node quality detection evaluation layer;

And adding the mth node quality detection evaluation layer to the PP line quality detection evaluation model.

Specifically, a plurality of m-node quality detection factors are randomly extracted from an m-node quality detection factor distribution sub-model.

Further, deep learning is performed according to the m-node quality detection factors, a plurality of first factor quality detection estimators corresponding to each m-node quality detection factor are obtained, weighting calculation is performed on the first factor quality detection estimators, a first m-node quality detection estimation branch corresponding to each m-node quality detection factor is obtained, and then the first m-node quality detection estimation branches are combined to obtain a plurality of m-node quality detection estimation branches.

Further, quality influence degree identification is performed based on a plurality of m-node quality detection factors, and a plurality of factor quality influence degrees are obtained. The quality influence degree represents the importance degree of the m-node quality detection factor in quality detection. For example, in the PP line quality detection, the importance of the PP line appearance detection is higher, and the importance of the PP line length detection is lower, so that the factor quality influence of the PP line appearance detection is higher than that of the PP line length detection.

Further, duty ratio calculation is performed based on the plurality of factor quality influence degrees, and a plurality of factor quality excitation coefficients are obtained. And respectively carrying out duty ratio calculation of all the factor quality influence degrees on each factor quality influence degree to obtain a plurality of factor quality excitation coefficients. The factor quality excitation coefficient is the weight corresponding to the factor quality influence degree. For example, if the plurality of factor mass influence degrees are 80%, 90%, and 95%, respectively, the calculated ratio of factor mass excitation coefficients are 80%/80% +90% +95%, 90%/80% +90% +95%, and 95%/80% +90% +95%, respectively, and further, the factor mass excitation coefficients are 0.3, 0.33, and 0.35, respectively.

Further, according to the multiple factor quality excitation coefficients, weighting calculation is carried out on multiple m node quality detection evaluation branches corresponding to the multiple factor quality influence degrees, and the weighting calculation results are respectively added to the corresponding m node quality evaluation excitation branches, so that an m node quality evaluation excitation branch is constructed.

Further, a plurality of m-node quality detection evaluation branches and an m-th node quality evaluation excitation branch are connected to generate an m-th node quality detection evaluation layer. The m-th node quality evaluation excitation branch comprises a plurality of factor quality excitation coefficients corresponding to a plurality of factor quality influence degrees of a plurality of m-th node quality detection evaluation branches, namely weights of the plurality of factor quality influence degrees, and the m-th node quality detection evaluation layer is generated through weighting calculation.

Further, a plurality of node quality detection evaluation branches corresponding to the M PP line production nodes and corresponding node quality evaluation excitation branches generate M node quality detection evaluation layers, and the M node quality detection evaluation layers are combined and added to the PP line quality detection evaluation model.

The PP line quality detection evaluation model is obtained by deep learning based on the PP line quality detection factor distribution model, and accuracy and efficiency of quality detection can be improved by performing quality detection through the PP line quality detection evaluation model.

Traversing the m node quality detection factors, and extracting a first m node quality detection factor;

acquiring PP line quality detection evaluation records according to the first m-node quality detection factors to obtain a first factor quality detection evaluation record library;

respectively performing supervised training on the first factor quality detection evaluation record library according to a plurality of preset deep learning operators to obtain a plurality of first factor quality detection estimators meeting preset output accuracy constraints;

obtaining a plurality of output accuracy rates of the plurality of first factor quality detection estimators;

performing duty ratio calculation according to the output accuracy rates to obtain a plurality of output excitation coefficients;

Adding the plurality of output excitation coefficients to a first factor quality detection assessment output exciter;

And connecting the plurality of first factor quality detection estimators and the first factor quality detection estimation output exciter, generating a first m-node quality detection estimation branch corresponding to the first m-node quality detection factor, and adding the first m-node quality detection estimation branch to the plurality of m-node quality detection estimation branches.

Specifically, a plurality of m-node quality detection factors are sequentially accessed, and a first m-node quality detection factor is extracted. Wherein each m-node quality detection factor is set as a first m-node quality detection factor, respectively.

Further, the PP line quality detection evaluation record corresponding to each first m-node quality detection factor is acquired according to the first m-node quality detection factors, and a plurality of PP line quality detection evaluation records are combined to obtain a first factor quality detection evaluation record library. The PP line quality detection evaluation record comprises quality detection qualified and quality detection unqualified results.

Further, a plurality of preset deep learning operators, for example, a deep learning network such as a BP neural network or a decision tree, is acquired, a plurality of first factor quality detection estimators are constructed, and the plurality of first factor quality detection estimators are deep learning network models. Further, the first factor quality detection evaluation record library is divided into training data and verification data, wherein the dividing ratio is obtained by custom setting by a person skilled in the art according to actual conditions, for example, the dividing ratio is 6:4. and respectively inputting the training data into a plurality of first factor quality detection estimators for supervised training, and outputting to obtain a plurality of output accuracy rates. And comparing the plurality of output accuracy rates with a preset output accuracy rate constraint to obtain a plurality of first factor quality detection estimators meeting the preset output accuracy rate constraint. The preset output accuracy constraint is obtained by custom setting according to practical situations by a person skilled in the art, for example, the preset output accuracy constraint is 80%.

Further, a plurality of output accuracy rates of a plurality of first factor quality detection estimators are extracted. And respectively carrying out duty ratio calculation of all the output accuracy rates on each output accuracy rate to obtain a plurality of output excitation coefficients. The output excitation coefficient is a weight corresponding to the output accuracy. For example, if the output accuracies are 80%, 90%, and 95%, respectively, the output excitation coefficients are 80%/80% +90% +95%, 90%/80% +90% +95%, and 95%/80% +90% +95%, respectively, and the output excitation coefficients are 0.3, 0.33, and 0.35, respectively.

Further, the plurality of output excitation factors are added to the first factor quality detection evaluation output driver for weighting calculation of a plurality of first factor quality detection estimators corresponding to the plurality of output accuracy rates.

Further, a plurality of first factor quality detection estimators and a first factor quality detection estimation output exciter are connected, wherein the first factor quality detection estimation output exciter comprises a plurality of output excitation coefficients of a plurality of output accuracy rates corresponding to the plurality of first factor quality detection estimators, namely weights of the plurality of output accuracy rates. Each first factor quality detection evaluator and the first factor quality detection evaluation output driver are combined to generate a first m-node quality detection evaluation branch corresponding to the first m-node quality detection factor, and the first m-node quality detection evaluation branch is added to the plurality of m-node quality detection evaluation branches.

The method comprises the steps of performing deep learning according to a plurality of m-node quality detection factors to obtain a plurality of m-node quality detection evaluation branches, and performing quality detection through the plurality of m-node quality detection evaluation branches to improve accuracy and efficiency of PP line quality detection.

Extracting a first matching node quality detection evaluation branch according to the plurality of matching node quality detection evaluation branches, wherein the first matching node quality detection evaluation branch comprises a plurality of first matching factor quality detection estimators and a first matching factor quality detection evaluation output exciter;

Performing feature recognition on the real-time node quality detection data according to a first matching node quality detection factor corresponding to the first matching node quality detection evaluation branch to obtain first matching factor quality detection data;

Inputting the first matching factor quality detection data into the plurality of first matching factor quality detection estimators to generate a plurality of first matching factor quality detection estimation coefficients;

inputting the plurality of first matching factor quality detection evaluation coefficients into the first matching factor quality detection evaluation output driver, generating a first node factor quality detection evaluation coefficient, and adding the first node factor quality detection evaluation coefficient to the plurality of node factor quality detection evaluation coefficients.

Specifically, random extraction is performed from a plurality of matching node quality detection evaluation branches to obtain a first matching node quality detection evaluation branch, wherein the first matching node quality detection evaluation branch comprises a plurality of first matching factor quality detection estimators and a first matching factor quality detection evaluation output exciter. The plurality of first matching factor quality detection estimators are for quality detection of the matching factors. The first matching factor quality detection evaluation output driver is used for weighting calculation of quality detection results of the matching factors.

Further, feature recognition is carried out on the real-time node quality detection data according to a plurality of first matching node quality detection factors corresponding to a plurality of first matching factor quality detection estimators in the first matching node quality detection and estimation branch, and the first matching node quality detection factors corresponding to matching are obtained. And performing feature recognition, namely comparison recognition, on the first matching node quality detection factor and the real-time node quality detection data to obtain first matching factor quality detection data.

Further, the first matching factor quality detection data is input to a plurality of first matching factor quality detection estimators, generating a plurality of first matching factor quality detection estimation coefficients. The first matching factor quality detection evaluation coefficient is used for duty ratio calculation, namely weighting calculation is carried out, and quality detection is completed.

Further, a plurality of first matching factor quality detection evaluation coefficients are input into a first matching factor quality detection evaluation output exciter, a first node factor quality detection evaluation coefficient is generated, and the first node factor quality detection evaluation coefficient is used for performing duty ratio calculation on the first matching factor quality detection evaluation coefficient. And obtaining a plurality of node factor quality detection evaluation coefficients by a method for obtaining the first node factor quality detection evaluation coefficient, thereby completing quality detection. The real-time node quality detection data is input into a plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients, so that quality detection efficiency and accuracy can be improved.

Embodiment two:

Based on the same inventive concept as the intelligent detection method of PP line quality in the foregoing embodiment, as shown in fig. 2, the present disclosure further provides an intelligent detection system of PP line quality, where the system includes:

the PP line production node obtaining module 11 is used for carrying out production node identification according to the PP line production process characteristic data to obtain M PP line production nodes, wherein M is a positive integer greater than 1;

A PP line production chain construction module 12, the PP line production chain construction module 12 being configured to construct a PP line production chain based on the M PP line production nodes;

The PP line quality detection factor distribution model construction module 13 is used for carrying out quality detection factor mining according to the PP line production chain to construct a PP line quality detection factor distribution model;

The PP line quality detection evaluation model obtaining module 14 is used for performing deep learning based on the PP line quality detection factor distribution model to obtain a PP line quality detection evaluation model;

The real-time node quality detection data acquisition module 15 is used for reading real-time PP line production nodes and carrying out quality detection on the real-time PP line production nodes according to the PP line quality detection factor distribution model to acquire real-time node quality detection data;

The matching node quality detection evaluation layer obtaining module 16 is configured to locate the PP line quality detection evaluation model based on the real-time PP line production node, and obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, where the matching node quality detection evaluation layer includes a plurality of matching node quality detection evaluation branches and a matching node quality evaluation excitation branch;

The node factor quality detection evaluation coefficient obtaining module 17, where the node factor quality detection evaluation coefficient obtaining module 17 is configured to input the real-time node quality detection data into the plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients;

a comprehensive quality detection evaluation coefficient obtaining module 18, where the comprehensive quality detection evaluation coefficient obtaining module 18 is configured to input the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient;

the PP line quality detection report obtaining module 19 is configured to generate a PP line quality detection report according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients, and the comprehensive quality detection evaluation coefficient.

Further, the system further comprises:

The PP line production order constraint obtaining module is used for extracting PP line production order constraint according to the PP line production process characteristic data;

The PP line production node distribution chain building module is used for building a PP line production node distribution chain according to the M PP line production nodes based on the PP line production order constraint;

The M node value degree obtaining modules are used for carrying out node value degree identification based on the M PP line production nodes to obtain M node value degrees;

the M node value proportion coefficient obtaining modules are used for carrying out duty ratio calculation based on the M node value degrees to obtain M node value proportion coefficients;

The PP line production chain obtaining module is used for marking the PP line production node distribution chain according to the M node value proportion coefficients to generate the PP line production chain.

Further, the system further comprises:

The m-PP line production node extraction module is used for extracting an m-PP line production node and an m node value specific gravity coefficient corresponding to the m-PP line production node according to the PP line production chain;

The mth node factor mining calculation force configuration module is used for configuring the mth node factor mining calculation force based on the mth node value gravity coefficient;

The m-th node quality detection event record obtaining module is used for carrying out quality detection event backtracking on the m-th PP line production node based on the m-th node factor mining calculation force to obtain an m-th node quality detection event record;

The node quality detection factor acquisition module is used for extracting a plurality of node sample quality detection factors corresponding to the m-PP line production node according to the m-th node quality detection event record;

the m-th node quality detection factor distribution sub-model obtaining module is used for constructing an m-th node quality detection factor distribution sub-model corresponding to the m-th PP line production node based on the plurality of node sample quality detection factors;

and the PP line quality detection factor distribution model obtaining module is used for adding the m-th node quality detection factor distribution sub-model to the PP line quality detection factor distribution model.

Further, the system further comprises:

The quality detection factor acquisition module is used for traversing the quality detection factors of the node samples to carry out repeated factor cleaning so as to acquire a plurality of quality detection factors of the candidate nodes;

the factor re-coupling degree obtaining module is used for traversing the candidate node quality detection factors to perform double-coupling degree identification to obtain multiple factor re-coupling degrees;

The node quality detection factor obtaining modules are used for carrying out decoupling factor cleaning on the candidate node quality detection factors based on the factor decoupling degrees to generate node quality detection factors;

And the m-th node quality detection factor distribution sub-model obtaining module is used for storing the plurality of node quality detection factors in a chain manner to generate the m-th node quality detection factor distribution sub-model.

Further, the system further comprises:

The m node quality detection factor obtaining modules are used for extracting a plurality of m node quality detection factors according to the m node quality detection factor distribution sub-model;

the m node quality detection evaluation branch acquisition modules are used for performing deep learning according to the m node quality detection factors to acquire m node quality detection evaluation branches;

the factor quality influence degree obtaining module is used for carrying out quality influence degree identification based on the m-node quality detection factors to obtain a plurality of factor quality influence degrees;

the factor quality excitation coefficient obtaining module is used for carrying out duty ratio calculation based on the factor quality influence values to obtain factor quality excitation coefficients;

The mth node quality evaluation excitation branch construction module is used for constructing an mth node quality evaluation excitation branch according to the multiple factor quality excitation coefficients;

The m-th node quality detection evaluation layer obtaining module is used for connecting the m-th node quality detection evaluation branches and the m-th node quality evaluation excitation branches to generate an m-th node quality detection evaluation layer;

And the PP line quality detection evaluation model generation module is used for adding the m-th node quality detection evaluation layer to the PP line quality detection evaluation model.

Further, the system further comprises:

The first m-node quality detection factor obtaining module is used for traversing the m-node quality detection factors and extracting a first m-node quality detection factor;

The first factor quality detection evaluation record library obtaining module is used for collecting PP line quality detection evaluation records according to the first m-node quality detection factors to obtain a first factor quality detection evaluation record library;

the first factor quality detection evaluator acquisition modules are used for respectively performing supervised training on the first factor quality detection evaluation record library according to a plurality of preset deep learning operators to acquire a plurality of first factor quality detection estimators meeting preset output accuracy constraints;

a plurality of output accuracy rate obtaining modules, configured to obtain a plurality of output accuracy rates of the plurality of first factor quality detection estimators;

the output excitation coefficient acquisition modules are used for performing duty ratio calculation according to the output accuracy rates to acquire output excitation coefficients;

A first factor quality detection evaluation output driver generation module for adding the plurality of output excitation coefficients to a first factor quality detection evaluation output driver;

The m-node quality detection evaluation branch generation modules are used for connecting the first factor quality detection estimators and the first factor quality detection evaluation output exciter, generating a first m-node quality detection evaluation branch corresponding to the first m-node quality detection factor, and adding the first m-node quality detection evaluation branch to the m-node quality detection evaluation branches.

Further, the system further comprises:

The first matching node quality detection evaluation branch extraction module is used for extracting a first matching node quality detection evaluation branch according to the plurality of matching node quality detection evaluation branches, wherein the first matching node quality detection evaluation branch comprises a plurality of first matching factor quality detection estimators and a first matching factor quality detection evaluation output exciter;

The first matching factor quality detection data acquisition module is used for carrying out feature recognition on the real-time node quality detection data according to a first matching node quality detection factor corresponding to the first matching node quality detection evaluation branch to acquire first matching factor quality detection data;

the first matching factor quality detection evaluation coefficient acquisition module is used for inputting the first matching factor quality detection data into the first matching factor quality detection estimators to generate a plurality of first matching factor quality detection evaluation coefficients;

The plurality of node factor quality detection evaluation coefficient obtaining modules are used for inputting the plurality of first matching factor quality detection evaluation coefficients into the first matching factor quality detection evaluation output exciter, generating a first node factor quality detection evaluation coefficient and adding the first node factor quality detection evaluation coefficient to the plurality of node factor quality detection evaluation coefficients.

The specific example of the intelligent detection method for PP line quality in the first embodiment is also applicable to the intelligent detection system for PP line quality in the present embodiment, and by the foregoing detailed description of the intelligent detection method for PP line quality, those skilled in the art can clearly know the intelligent detection system for PP line quality in the present embodiment, so that the details thereof will not be described herein for brevity. The device disclosed in the embodiment corresponds to the method disclosed in the embodiment, so that the description is simpler, and the relevant points refer to the description of the method.

According to an embodiment of the present disclosure, the present disclosure further provides a computer readable storage medium having stored thereon a computer program which, when executed, implements the steps provided by any of the above embodiments.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. An intelligent detection method for PP line quality is characterized by comprising the following steps:

The method comprises the steps of carrying out quality detection factor mining according to the PP line production chain, constructing a PP line quality detection factor distribution model, and comprising the following steps:

Adding the mth node quality detection factor distribution sub-model to the PP line quality detection factor distribution model;

based on the plurality of node sample quality detection factors, constructing an mth node quality detection factor distribution sub-model corresponding to the mth-PP line production node, wherein the method comprises the following steps of:

performing chain storage on the plurality of node quality detection factors to generate an mth node quality detection factor distribution sub-model;

The method comprises the steps of performing deep learning based on the PP line quality detection factor distribution model to obtain a PP line quality detection evaluation model, and comprises the following steps:

Adding the mth node quality detection evaluation layer to the PP line quality detection evaluation model;

2. The method of claim 1, wherein constructing a PP line production chain based on the M PP line production nodes comprises:

3. The method of claim 1, wherein performing deep learning based on the plurality of m-node quality detection factors to obtain a plurality of m-node quality detection evaluation branches comprises:

4. The method of claim 1, wherein inputting the real-time node quality detection data into the plurality of matching node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients, comprising:

5. An intelligent detection system for PP line quality, characterized in that it is used for implementing an intelligent detection method for PP line quality according to any of claims 1-4, said system comprising:

the PP line production node acquisition module is used for carrying out production node identification according to the PP line production process characteristic data to acquire M PP line production nodes, wherein M is a positive integer greater than 1;

The PP line production chain construction module is used for constructing a PP line production chain based on the M PP line production nodes;

The PP line quality detection factor distribution model construction module is used for carrying out quality detection factor mining according to the PP line production chain to construct a PP line quality detection factor distribution model;

the PP line quality detection evaluation model obtaining module is used for carrying out deep learning based on the PP line quality detection factor distribution model to obtain a PP line quality detection evaluation model;

The real-time node quality detection data acquisition module is used for reading real-time PP line production nodes and carrying out quality detection on the real-time PP line production nodes according to the PP line quality detection factor distribution model to acquire real-time node quality detection data;

The matching node quality detection evaluation layer obtaining module is used for positioning the PP line quality detection evaluation model based on the real-time PP line production node to obtain a matching node quality detection evaluation layer corresponding to the real-time PP line production node, wherein the matching node quality detection evaluation layer comprises a plurality of matching node quality detection evaluation branches and matching node quality evaluation excitation branches;

The node factor quality detection evaluation coefficient obtaining module is used for inputting the real-time node quality detection data into the plurality of matched node quality detection evaluation branches to obtain a plurality of node factor quality detection evaluation coefficients;

The comprehensive quality detection evaluation coefficient obtaining module is used for inputting the plurality of node factor quality detection evaluation coefficients into the matching node quality evaluation excitation branch to obtain a comprehensive quality detection evaluation coefficient;

And the PP line quality detection report acquisition module is used for generating a PP line quality detection report according to the real-time node quality detection data, the plurality of node factor quality detection evaluation coefficients and the comprehensive quality detection evaluation coefficient.